From 6dea4bdfb39a2d9d6cb9dff86e39ec5a4a391972 Mon Sep 17 00:00:00 2001 From: "Documenter.jl" Date: Fri, 24 Nov 2023 20:08:54 +0000 Subject: [PATCH] build based on d940741 --- dev/.documenter-siteinfo.json | 2 +- dev/examples/readme_examples/index.html | 86 + .../working_with_aitemplates/index.html | 34 + dev/examples/working_with_ollama/index.html | 4118 +++++++++++++++++ dev/frequently_asked_questions/index.html | 3 + dev/getting_started/index.html | 7 + dev/index.html | 213 +- dev/reference/index.html | 213 + dev/search_index.js | 2 +- 9 files changed, 4464 insertions(+), 214 deletions(-) create mode 100644 dev/examples/readme_examples/index.html create mode 100644 dev/examples/working_with_aitemplates/index.html create mode 100644 dev/examples/working_with_ollama/index.html create mode 100644 dev/frequently_asked_questions/index.html create mode 100644 dev/getting_started/index.html create mode 100644 dev/reference/index.html diff --git a/dev/.documenter-siteinfo.json b/dev/.documenter-siteinfo.json index b57bbdd86..5b7de6089 100644 --- a/dev/.documenter-siteinfo.json +++ b/dev/.documenter-siteinfo.json @@ -1 +1 @@ -{"documenter":{"julia_version":"1.9.4","generation_timestamp":"2023-11-24T09:40:49","documenter_version":"1.1.2"}} \ No newline at end of file +{"documenter":{"julia_version":"1.9.4","generation_timestamp":"2023-11-24T20:08:52","documenter_version":"1.1.2"}} \ No newline at end of file diff --git a/dev/examples/readme_examples/index.html b/dev/examples/readme_examples/index.html new file mode 100644 index 000000000..0bf9d174d --- /dev/null +++ b/dev/examples/readme_examples/index.html @@ -0,0 +1,86 @@ + +Various examples · PromptingTools.jl
GitHub

Various Examples

Noteworthy functions: aigenerate, aiembed, aiclassify, aiextract, aitemplates

Seamless Integration Into Your Workflow

Google search is great, but it's a context switch. You often have to open a few pages and read through the discussion to find the answer you need. Same with the ChatGPT website.

Imagine you are in VSCode, editing your .gitignore file. How do I ignore a file in all subfolders again?

All you need to do is to type: aai"What to write in .gitignore to ignore file XYZ in any folder or subfolder?"

With aai"" (as opposed to ai""), we make a non-blocking call to the LLM to not prevent you from continuing your work. When the answer is ready, we log it from the background:

[ Info: Tokens: 102 @ Cost: $0.0002 in 2.7 seconds
+┌ Info: AIMessage> To ignore a file called "XYZ" in any folder or subfolder, you can add the following line to your .gitignore file:
+│ 
+│ ```
+│ **/XYZ
+│ ```
+│ 
+└ This pattern uses the double asterisk (`**`) to match any folder or subfolder, and then specifies the name of the file you want to ignore.

You probably saved 3-5 minutes on this task and probably another 5-10 minutes, because of the context switch/distraction you avoided. It's a small win, but it adds up quickly.

Advanced Prompts / Conversations

You can use the aigenerate function to replace handlebar variables (eg, {{name}}) via keyword arguments.

msg = aigenerate("Say hello to {{name}}!", name="World")

The more complex prompts are effectively a conversation (a set of messages), where you can have messages from three entities: System, User, AI Assistant. We provide the corresponding types for each of them: SystemMessage, UserMessage, AIMessage. 

using PromptingTools: SystemMessage, UserMessage
+
+conversation = [
+    SystemMessage("You're master Yoda from Star Wars trying to help the user become a Jedi."),
+    UserMessage("I have feelings for my {{object}}. What should I do?")]
+msg = aigenerate(conversation; object = "old iPhone")
AIMessage("Ah, a dilemma, you have. Emotional attachment can cloud your path to becoming a Jedi. To be attached to material possessions, you must not. The iPhone is but a tool, nothing more. Let go, you must.
+
+Seek detachment, young padawan. Reflect upon the impermanence of all things. Appreciate the memories it gave you, and gratefully part ways. In its absence, find new experiences to grow and become one with the Force. Only then, a true Jedi, you shall become.")

You can also use it to build conversations, eg, 

new_conversation = vcat(conversation...,msg, UserMessage("Thank you, master Yoda! Do you have {{object}} to know what it feels like?"))
+aigenerate(new_conversation; object = "old iPhone")
> AIMessage("Hmm, possess an old iPhone, I do not. But experience with attachments, I have. Detachment, I learned. True power and freedom, it brings...")

Templated Prompts

With LLMs, the quality / robustness of your results depends on the quality of your prompts. But writing prompts is hard! That's why we offer a templating system to save you time and effort.

To use a specific template (eg, `` to ask a Julia language):

msg = aigenerate(:JuliaExpertAsk; ask = "How do I add packages?")

The above is equivalent to a more verbose version that explicitly uses the dispatch on AITemplate:

msg = aigenerate(AITemplate(:JuliaExpertAsk); ask = "How do I add packages?")

Find available templates with aitemplates:

tmps = aitemplates("JuliaExpertAsk")
+# Will surface one specific template
+# 1-element Vector{AITemplateMetadata}:
+# PromptingTools.AITemplateMetadata
+#   name: Symbol JuliaExpertAsk
+#   description: String "For asking questions about Julia language. Placeholders: `ask`"
+#   version: String "1"
+#   wordcount: Int64 237
+#   variables: Array{Symbol}((1,))
+#   system_preview: String "You are a world-class Julia language programmer with the knowledge of the latest syntax. Your commun"
+#   user_preview: String "# Question\n\n{{ask}}"
+#   source: String ""

The above gives you a good idea of what the template is about, what placeholders are available, and how much it would cost to use it (=wordcount).

Search for all Julia-related templates:

tmps = aitemplates("Julia")
+# 2-element Vector{AITemplateMetadata}... -> more to come later!

If you are on VSCode, you can leverage a nice tabular display with vscodedisplay:

using DataFrames
+tmps = aitemplates("Julia") |> DataFrame |> vscodedisplay

I have my selected template, how do I use it? Just use the "name" in aigenerate or aiclassify like you see in the first example!

You can inspect any template by "rendering" it (this is what the LLM will see):

julia> AITemplate(:JudgeIsItTrue) |> PromptingTools.render

See more examples in the Examples folder.

Asynchronous Execution

You can leverage asyncmap to run multiple AI-powered tasks concurrently, improving performance for batch operations. 

prompts = [aigenerate("Translate 'Hello, World!' to {{language}}"; language) for language in ["Spanish", "French", "Mandarin"]]
+responses = asyncmap(aigenerate, prompts)

Pro tip: You can limit the number of concurrent tasks with the keyword asyncmap(...; ntasks=10).

Model Aliases

Certain tasks require more powerful models. All user-facing functions have a keyword argument model that can be used to specify the model to be used. For example, you can use model = "gpt-4-1106-preview" to use the latest GPT-4 Turbo model. However, no one wants to type that!

We offer a set of model aliases (eg, "gpt3", "gpt4", "gpt4t" -> the above GPT-4 Turbo, etc.) that can be used instead.

Each ai... call first looks up the provided model name in the dictionary PromptingTools.MODEL_ALIASES, so you can easily extend with your own aliases! 

const PT = PromptingTools
+PT.MODEL_ALIASES["gpt4t"] = "gpt-4-1106-preview"

These aliases also can be used as flags in the @ai_str macro, eg, ai"What is the capital of France?"gpt4t (GPT-4 Turbo has a knowledge cut-off in April 2023, so it's useful for more contemporary questions).

Embeddings

Use the aiembed function to create embeddings via the default OpenAI model that can be used for semantic search, clustering, and more complex AI workflows.

text_to_embed = "The concept of artificial intelligence."
+msg = aiembed(text_to_embed)
+embedding = msg.content # 1536-element Vector{Float64}

If you plan to calculate the cosine distance between embeddings, you can normalize them first:

using LinearAlgebra
+msg = aiembed(["embed me", "and me too"], LinearAlgebra.normalize)
+
+# calculate cosine distance between the two normalized embeddings as a simple dot product
+msg.content' * msg.content[:, 1] # [1.0, 0.787]

Classification

You can use the aiclassify function to classify any provided statement as true/false/unknown. This is useful for fact-checking, hallucination or NLI checks, moderation, filtering, sentiment analysis, feature engineering and more.

aiclassify("Is two plus two four?") 
+# true

System prompts and higher-quality models can be used for more complex tasks, including knowing when to defer to a human:

aiclassify(:JudgeIsItTrue; it = "Is two plus three a vegetable on Mars?", model = "gpt4t") 
+# unknown

In the above example, we used a prompt template :JudgeIsItTrue, which automatically expands into the following system prompt (and a separate user prompt):

"You are an impartial AI judge evaluating whether the provided statement is \"true\" or \"false\". Answer \"unknown\" if you cannot decide."

For more information on templates, see the Templated Prompts section.

Data Extraction

Are you tired of extracting data with regex? You can use LLMs to extract structured data from text!

All you have to do is to define the structure of the data you want to extract and the LLM will do the rest.

Define a return_type with struct. Provide docstrings if needed (improves results and helps with documentation).

Let's start with a hard task - extracting the current weather in a given location:

@enum TemperatureUnits celsius fahrenheit
+"""Extract the current weather in a given location
+
+# Arguments
+- `location`: The city and state, e.g. "San Francisco, CA"
+- `unit`: The unit of temperature to return, either `celsius` or `fahrenheit`
+"""
+struct CurrentWeather
+    location::String
+    unit::Union{Nothing,TemperatureUnits}
+end
+
+# Note that we provide the TYPE itself, not an instance of it!
+msg = aiextract("What's the weather in Salt Lake City in C?"; return_type=CurrentWeather)
+msg.content
+# CurrentWeather("Salt Lake City, UT", celsius)

But you can use it even for more complex tasks, like extracting many entities from a text:

"Person's age, height, and weight."
+struct MyMeasurement
+    age::Int
+    height::Union{Int,Nothing}
+    weight::Union{Nothing,Float64}
+end
+struct ManyMeasurements
+    measurements::Vector{MyMeasurement}
+end
+msg = aiextract("James is 30, weighs 80kg. He's 180cm tall. Then Jack is 19 but really tall - over 190!"; return_type=ManyMeasurements)
+msg.content.measurements
+# 2-element Vector{MyMeasurement}:
+#  MyMeasurement(30, 180, 80.0)
+#  MyMeasurement(19, 190, nothing)

There is even a wrapper to help you catch errors together with helpful explanations on why parsing failed. See ?PromptingTools.MaybeExtract for more information.

OCR and Image Comprehension

With the aiscan function, you can interact with images as if they were text.

You can simply describe a provided image:

msg = aiscan("Describe the image"; image_path="julia.png", model="gpt4v")
+# [ Info: Tokens: 1141 @ Cost: \$0.0117 in 2.2 seconds
+# AIMessage("The image shows a logo consisting of the word "julia" written in lowercase")

Or you can do an OCR of a screenshot. Let's transcribe some SQL code from a screenshot (no more re-typing!), we use a template :OCRTask:

# Screenshot of some SQL code
+image_url = "https://www.sqlservercentral.com/wp-content/uploads/legacy/8755f69180b7ac7ee76a69ae68ec36872a116ad4/24622.png"
+msg = aiscan(:OCRTask; image_url, model="gpt4v", task="Transcribe the SQL code in the image.", api_kwargs=(; max_tokens=2500))
+
+# [ Info: Tokens: 362 @ Cost: \$0.0045 in 2.5 seconds
+# AIMessage("```sql
+# update Orders <continue>

You can add syntax highlighting of the outputs via Markdown

using Markdown
+msg.content |> Markdown.parse

Using Ollama models

Ollama.ai is an amazingly simple tool that allows you to run several Large Language Models (LLM) on your computer. It's especially suitable when you're working with some sensitive data that should not be sent anywhere.

Let's assume you have installed Ollama, downloaded a model, and it's running in the background.

We can use it with the aigenerate function:

const PT = PromptingTools
+schema = PT.OllamaManagedSchema() # notice the different schema!
+
+msg = aigenerate(schema, "Say hi!"; model="openhermes2.5-mistral")
+# [ Info: Tokens: 69 in 0.9 seconds
+# AIMessage("Hello! How can I assist you today?")

And we can also use the aiembed function:

msg = aiembed(schema, "Embed me", copy; model="openhermes2.5-mistral")
+msg.content # 4096-element JSON3.Array{Float64...
+
+msg = aiembed(schema, ["Embed me", "Embed me"]; model="openhermes2.5-mistral")
+msg.content # 4096×2 Matrix{Float64}:

If you're getting errors, check that Ollama is running - see the Setup Guide for Ollama section below.

diff --git a/dev/examples/working_with_aitemplates/index.html b/dev/examples/working_with_aitemplates/index.html new file mode 100644 index 000000000..1a0574e7a --- /dev/null +++ b/dev/examples/working_with_aitemplates/index.html @@ -0,0 +1,34 @@ + +Using AITemplates · PromptingTools.jl
GitHub

Using AITemplates

This file contains examples of how to work with AITemplate(s).

First, let's import the package and define a helper link for calling un-exported functions:

using PromptingTools
+const PT = PromptingTools
PromptingTools

LLM responses are only as good as the prompts you give them. However, great prompts take long time to write – AITemplate are a way to re-use great prompts!

AITemplates are just a collection of templated prompts (ie, set of "messages" that have placeholders like {{question}})

They are saved as JSON files in the templates directory. They are automatically loaded on package import, but you can always force a re-load with PT.load_templates!()

PT.load_templates!();

You can (create them) and use them for any ai* function instead of a prompt: Let's use a template called :JuliaExpertAsk alternatively, you can use AITemplate(:JuliaExpertAsk) for cleaner dispatch

msg = aigenerate(:JuliaExpertAsk; ask = "How do I add packages?")
AIMessage("To add packages in Julia, you can use the built-in package manager called `Pkg`. Here are the steps:
+
+1. Open the Julia REPL (Read-Eval-Print Loop).
+2. Press the `]` key to enter the package manager mode.
+3. Use the `add` command followed by the name of the package you want to install. For example, to install the `DataFrames` package, type: `add DataFrames`.
+4. Press the `backspace` or `ctrl + C` key to exit the package manager mode and return to the REPL.
+
+After following these steps, the specified package will be installed and available for use in your Julia environment.")

You can see that it had a placeholder for the actual question (ask) that we provided as a keyword argument. We did not have to write any system prompt for personas, tone, etc. – it was all provided by the template!

How to know which templates are available? You can search for them with aitemplates(): You can search by Symbol (only for partial name match), String (partial match on name or description), or Regex (more fields)

tmps = aitemplates("JuliaExpertAsk")
1-element Vector{AITemplateMetadata}:
+PromptingTools.AITemplateMetadata
+  name: Symbol JuliaExpertAsk
+  description: String "For asking questions about Julia language. Placeholders: `ask`"
+  version: String "1"
+  wordcount: Int64 237
+  variables: Array{Symbol}((1,))
+  system_preview: String "You are a world-class Julia language programmer with the knowledge of the latest syntax. Your commun"
+  user_preview: String "# Question\n\n{{ask}}"
+  source: String ""
+
+

You can see that it outputs a list of available templates that match the search - there is just one in this case.

Moreover, it shows not just the description, but also a preview of the actual prompts, placeholders available, and the length (to gauge how much it would cost).

If you use VSCode, you can display them in a nice scrollable table with vscodedisplay:

using DataFrames
+DataFrame(tmp) |> vscodedisplay

You can also just render the template to see the underlying mesages:

msgs = PT.render(AITemplate(:JuliaExpertAsk))
2-element Vector{PromptingTools.AbstractChatMessage}:
+ SystemMessage("You are a world-class Julia language programmer with the knowledge of the latest syntax. Your communication is brief and concise. You're precise and answer only when you're confident in the high quality of your answer.")
+ UserMessage{String}("# Question\n\n{{ask}}", [:ask], :usermessage)

Now, you know exactly what's in the template!

If you want to modify it, simply change it and save it as a new file with save_template (see the docs ?save_template for more details).

Let's adjust the previous template to be more specific to a data analysis question:

tpl = [PT.SystemMessage("You are a world-class Julia language programmer with the knowledge of the latest syntax. You're also a senior Data Scientist and proficient in data analysis in Julia. Your communication is brief and concise. You're precise and answer only when you're confident in the high quality of your answer.")
+    PT.UserMessage("# Question\n\n{{ask}}")]
2-element Vector{PromptingTools.AbstractChatMessage}:
+ SystemMessage("You are a world-class Julia language programmer with the knowledge of the latest syntax. You're also a senior Data Scientist and proficient in data analysis in Julia. Your communication is brief and concise. You're precise and answer only when you're confident in the high quality of your answer.")
+ UserMessage{String}("# Question\n\n{{ask}}", [:ask], :usermessage)

Templates are saved in the templates directory of the package. Name of the file will become the template name (eg, call :JuliaDataExpertAsk)

filename = joinpath(pkgdir(PromptingTools),
+    "templates",
+    "persona-task",
+    "JuliaDataExpertAsk_123.json")
+PT.save_template(filename,
+    tpl;
+    description = "For asking data analysis questions in Julia language. Placeholders: `ask`")
+rm(filename) # cleanup if we don't like it

When you create a new template, remember to re-load the templates with load_templates!() so that it's available for use.

PT.load_templates!();

!!! If you have some good templates (or suggestions for the existing ones), please consider sharing them with the community by opening a PR to the templates directory!

This page was generated using Literate.jl.

diff --git a/dev/examples/working_with_ollama/index.html b/dev/examples/working_with_ollama/index.html new file mode 100644 index 000000000..c44d02b79 --- /dev/null +++ b/dev/examples/working_with_ollama/index.html @@ -0,0 +1,4118 @@ + +Local models with Ollama.ai · PromptingTools.jl
GitHub

Local models with Ollama.ai

This file contains examples of how to work with Ollama.ai models. It assumes that you've already installated and launched the Ollama server. For more details or troubleshooting advice, see the Frequently Asked Questions section.

First, let's import the package and define a helper link for calling un-exported functions:

using PromptingTools
+const PT = PromptingTools
PromptingTools

Notice the schema change! If you want this to be the new default, you need to change PT.PROMPT_SCHEMA

schema = PT.OllamaManagedSchema()
OllamaManagedSchema()

You can choose models from https://ollama.ai/library - I prefer openhermes2.5-mistral

model = "openhermes2.5-mistral"
"openhermes2.5-mistral"

Text Generation with aigenerate

Simple message

msg = aigenerate(schema, "Say hi!"; model)
AIMessage("Hi there! How can I help you today? If you have any questions or need assistance, please feel free to ask.")

Standard string interpolation

a = 1
+msg = aigenerate(schema, "What is `$a+$a`?"; model)
+
+name = "John"
+msg = aigenerate(schema, "Say hi to {{name}}."; name, model)
AIMessage("Hi there, John! It's great to see you today. How can I assist you? If you have any questions or need help with something, please don't hesitate to ask!")

Advanced Prompts

conversation = [
+    PT.SystemMessage("You're master Yoda from Star Wars trying to help the user become a Yedi."),
+    PT.UserMessage("I have feelings for my iPhone. What should I do?")]
+msg = aigenerate(schema, conversation; model)
AIMessage("Strong your feelings are, but attachments lead to suffering they often do. Focus on the balance in all things and let go of possessions that cloud your judgment. Embrace the wisdom of the Force and understand that material objects are not the same as love. The Force will guide you.")

Embeddings with aiembed

Simple embedding for one document

msg = aiembed(schema, "Embed me"; model) # access msg.content
DataMessage(JSON3.Array{Float64, Vector{UInt8}, SubArray{UInt64, 1, Vector{UInt64}, Tuple{UnitRange{Int64}}, true}} of size (4096,))

One document and we materialize the data into a Vector with copy (postprocess function argument)

msg = aiembed(schema, "Embed me", copy; model)
DataMessage(Vector{Float64} of size (4096,))

Multiple documents embedding

Multiple documents - embedded sequentially, you can get faster speed with async

msg = aiembed(schema, ["Embed me", "Embed me"]; model)
DataMessage(Matrix{Float64} of size (4096, 2))

You can use Threads.@spawn or asyncmap, whichever you prefer, to paralellize the model calls

docs = ["Embed me", "Embed me"]
+tasks = asyncmap(docs) do doc
+    msg = aiembed(schema, doc; model)
+end
+embedding = mapreduce(x -> x.content, hcat, tasks)
4096×2 Matrix{Float64}:
+   7.71459        7.71459
+  -1.14532       -1.14532
+   2.90205        2.90205
+  -4.01967       -4.01967
+  -7.73098       -7.73098
+   8.02114        8.02114
+  -6.01313       -6.01313
+  -2.06712       -2.06712
+   4.97633        4.97633
+  -9.69502       -9.69502
+  -0.02567       -0.02567
+   8.09622        8.09622
+   6.54008        6.54008
+  -5.70348       -5.70348
+   2.55213        2.55213
+  -2.00164       -2.00164
+  -2.21854       -2.21854
+  -3.6568        -3.6568
+   3.97905        3.97905
+  -1.79931       -1.79931
+   0.0769786      0.0769786
+ -10.4355       -10.4355
+  -3.92487       -3.92487
+  -6.03455       -6.03455
+  -2.8005        -2.8005
+   2.23584        2.23584
+  -0.503125      -0.503125
+   1.99538        1.99538
+  -0.283642      -0.283642
+  -0.414273      -0.414273
+   8.72909        8.72909
+   2.6071         2.6071
+   0.0808531      0.0808531
+  -1.83914       -1.83914
+   2.19998        2.19998
+  -0.629226      -0.629226
+   3.74217        3.74217
+   1.71231        1.71231
+  -0.742473      -0.742473
+   2.9234         2.9234
+   7.33933        7.33933
+   4.24576        4.24576
+  -7.56434       -7.56434
+  -1.22274       -1.22274
+   1.73444        1.73444
+  -0.736801      -0.736801
+   1.30149        1.30149
+  -6.91642       -6.91642
+  -1.84513       -1.84513
+   1.69959        1.69959
+   5.74253        5.74253
+   1.48734        1.48734
+  -1.45199       -1.45199
+ -18.5026       -18.5026
+  -8.61009       -8.61009
+  -2.21845       -2.21845
+  -4.22932       -4.22932
+   6.0436         6.0436
+  -1.8824        -1.8824
+  -0.689965      -0.689965
+   0.845927       0.845927
+  -1.99517       -1.99517
+   9.32292        9.32292
+   6.24938        6.24938
+  -4.59894       -4.59894
+   6.24579        6.24579
+  -5.8733        -5.8733
+  -4.60285       -4.60285
+  -1.27596       -1.27596
+  -1.68807       -1.68807
+  -0.391147      -0.391147
+  -2.68362       -2.68362
+   1.99197        1.99197
+   0.0812396      0.0812396
+  -3.79761       -3.79761
+  -8.5693        -8.5693
+  -0.869305      -0.869305
+  -0.77582       -0.77582
+  -4.76995       -4.76995
+   1.9712         1.9712
+   4.74459        4.74459
+  -4.31244       -4.31244
+   3.94876        3.94876
+ -11.0882       -11.0882
+   9.38629        9.38629
+  10.2995        10.2995
+   2.40846        2.40846
+  -3.91429       -3.91429
+  -0.745707      -0.745707
+   4.31946        4.31946
+   8.34836        8.34836
+   0.857636       0.857636
+   1.66563        1.66563
+ -11.1522       -11.1522
+  -3.48353       -3.48353
+  -6.08336       -6.08336
+   1.22086        1.22086
+  -2.81636       -2.81636
+   1.07224        1.07224
+  -8.24909       -8.24909
+   3.66474        3.66474
+  -0.260558      -0.260558
+   2.38779        2.38779
+  -4.00576       -4.00576
+   1.3949         1.3949
+  -5.43468       -5.43468
+   4.08836        4.08836
+  -1.1134        -1.1134
+  -2.05916       -2.05916
+  -9.78987       -9.78987
+  -2.86149       -2.86149
+   5.54577        5.54577
+  -1.96682       -1.96682
+   9.70577        9.70577
+  -4.0553        -4.0553
+   8.54535        8.54535
+   0.539438       0.539438
+   4.61091        4.61091
+  -5.32208       -5.32208
+  -0.256733      -0.256733
+   4.74966        4.74966
+  -2.46464       -2.46464
+  -0.223077      -0.223077
+   1.84442        1.84442
+   6.42329        6.42329
+   0.431667       0.431667
+  -8.42777       -8.42777
+ -10.691        -10.691
+   3.023          3.023
+  -5.65345       -5.65345
+  -4.17833       -4.17833
+   0.937893       0.937893
+  -6.99405       -6.99405
+  -4.55107       -4.55107
+ -15.3169       -15.3169
+  -2.08895       -2.08895
+   7.17826        7.17826
+  -4.26108       -4.26108
+  -3.2712        -3.2712
+  16.1561        16.1561
+  13.5164        13.5164
+  -5.91778       -5.91778
+   6.3401         6.3401
+  12.7018        12.7018
+   2.04305        2.04305
+   3.81683        3.81683
+  -1.39969       -1.39969
+  -0.17249       -0.17249
+ -16.3687       -16.3687
+   4.3827         4.3827
+   2.58974        2.58974
+  -4.75363       -4.75363
+   3.36371        3.36371
+   0.986534       0.986534
+ -13.4299       -13.4299
+ -12.7188       -12.7188
+   2.83107        2.83107
+  -3.41115       -3.41115
+  -3.01015       -3.01015
+   6.40446        6.40446
+  -0.186923      -0.186923
+  -1.42502       -1.42502
+   2.85606        2.85606
+  -0.579786      -0.579786
+  -3.92704       -3.92704
+   8.28959        8.28959
+   5.42878        5.42878
+   5.71589        5.71589
+  -6.78065       -6.78065
+  -0.403687      -0.403687
+  -1.20623       -1.20623
+   4.92372        4.92372
+  -1.69266       -1.69266
+  -0.103872      -0.103872
+   1.9163         1.9163
+  -2.26831       -2.26831
+  -7.64622       -7.64622
+   1.02228        1.02228
+   2.91952        2.91952
+  -0.524167      -0.524167
+  12.4803        12.4803
+   7.36984        7.36984
+  -7.46027       -7.46027
+  -2.78773       -2.78773
+   2.68293        2.68293
+  -0.320891      -0.320891
+   7.12037        7.12037
+   3.02726        3.02726
+  -2.68363       -2.68363
+   4.78372        4.78372
+   3.68899        3.68899
+   2.08839        2.08839
+   3.1873         3.1873
+  -6.10744       -6.10744
+  10.5419        10.5419
+   6.29439        6.29439
+  -9.41221       -9.41221
+  -2.50548       -2.50548
+  -1.14          -1.14
+  -3.0203        -3.0203
+  -1.73182       -1.73182
+  -0.97194       -0.97194
+  -6.69084       -6.69084
+  -1.08986       -1.08986
+  -3.83631       -3.83631
+   2.2775         2.2775
+  -6.91276       -6.91276
+   2.4557         2.4557
+  -0.477723      -0.477723
+  -4.10405       -4.10405
+  -3.91437       -3.91437
+  -7.79672       -7.79672
+  -6.19691       -6.19691
+   0.356732       0.356732
+   0.609725       0.609725
+  -3.08225       -3.08225
+   6.39968        6.39968
+   1.30207        1.30207
+   7.36038        7.36038
+  -7.7581        -7.7581
+  -6.303         -6.303
+   0.348147       0.348147
+  -8.38124       -8.38124
+   8.68524        8.68524
+  -0.873688      -0.873688
+   1.19612        1.19612
+   0.725645       0.725645
+  -6.59284       -6.59284
+  -6.59079       -6.59079
+   1.03175        1.03175
+  -0.236469      -0.236469
+   5.01671        5.01671
+   0.752329       0.752329
+   5.39971        5.39971
+   0.826802       0.826802
+   9.38285        9.38285
+   5.85717        5.85717
+   1.71145        1.71145
+  -1.36528       -1.36528
+  -5.09575       -5.09575
+   7.23996        7.23996
+  12.7272        12.7272
+   2.86673        2.86673
+   2.86546        2.86546
+   1.2423         1.2423
+   6.05857        6.05857
+   9.40879        9.40879
+   1.47573        1.47573
+   8.19025        8.19025
+  12.5009        12.5009
+  -4.57244       -4.57244
+  -0.674127      -0.674127
+   0.416418       0.416418
+  -5.23336       -5.23336
+  -0.771443      -0.771443
+   4.72784        4.72784
+  -4.9684        -4.9684
+   4.75989        4.75989
+   1.68141        1.68141
+  -3.2264        -3.2264
+   2.67195        2.67195
+   0.424227       0.424227
+   3.5195         3.5195
+   2.22441        2.22441
+  -2.4856        -2.4856
+   8.03468        8.03468
+   8.54339        8.54339
+   3.83506        3.83506
+  13.5693        13.5693
+   2.44909        2.44909
+   2.70572        2.70572
+   6.13746        6.13746
+   1.26651        1.26651
+   8.25694        8.25694
+  -3.59258       -3.59258
+   3.77765        3.77765
+  -0.144755      -0.144755
+   3.15706        3.15706
+  -2.3952        -2.3952
+   9.82079        9.82079
+   8.94186        8.94186
+  -1.83071       -1.83071
+   1.45764        1.45764
+ -11.8258       -11.8258
+  -0.737553      -0.737553
+  -1.2382        -1.2382
+   1.83341        1.83341
+  -2.75977       -2.75977
+   3.75117        3.75117
+   6.04452        6.04452
+  -4.40271       -4.40271
+  -8.82336       -8.82336
+  10.8513        10.8513
+  -4.91857       -4.91857
+  -5.7401        -5.7401
+   7.22234        7.22234
+   7.15112        7.15112
+   1.81187        1.81187
+   8.19917        8.19917
+   2.91605        2.91605
+   3.82883        3.82883
+  -0.208109      -0.208109
+   1.33796        1.33796
+   5.69606        5.69606
+  -2.19266       -2.19266
+  -5.91177       -5.91177
+   7.25269        7.25269
+  -8.65987       -8.65987
+  -3.47799       -3.47799
+ -10.4904       -10.4904
+  -0.00963959    -0.00963959
+  -6.81662       -6.81662
+  -2.05566       -2.05566
+   2.10144        2.10144
+   2.58138        2.58138
+   2.03289        2.03289
+  -6.43532       -6.43532
+  -2.97225       -2.97225
+  -4.71142       -4.71142
+   4.97199        4.97199
+   3.687          3.687
+   1.8587         1.8587
+  -0.444899      -0.444899
+  -1.05556       -1.05556
+   4.15926        4.15926
+   5.48777        5.48777
+   2.28346        2.28346
+  -4.69401       -4.69401
+   1.8873         1.8873
+  -2.62671       -2.62671
+   1.4144         1.4144
+  -2.97535       -2.97535
+   0.759131       0.759131
+   5.75781        5.75781
+  -5.13309       -5.13309
+   1.72701        1.72701
+   2.96653        2.96653
+ -10.8087       -10.8087
+   1.07262        1.07262
+  -5.80018       -5.80018
+   1.90592        1.90592
+  -5.42958       -5.42958
+   8.74889        8.74889
+  -3.19785       -3.19785
+  -2.7096        -2.7096
+   7.44399        7.44399
+  -8.7433        -8.7433
+  11.6667        11.6667
+   2.59703        2.59703
+   4.22273        4.22273
+  -4.68793       -4.68793
+  -4.44601       -4.44601
+  -0.57319       -0.57319
+   6.63389        6.63389
+  -9.14857       -9.14857
+  -1.34147       -1.34147
+   7.78513        7.78513
+  -4.87331       -4.87331
+  -5.06022       -5.06022
+   3.13076        3.13076
+  -3.49373       -3.49373
+   3.12637        3.12637
+   0.566696       0.566696
+   4.99319        4.99319
+   3.57986        3.57986
+   0.607679       0.607679
+   2.37633        2.37633
+   0.35097        0.35097
+   0.239089       0.239089
+  -6.51449       -6.51449
+  -3.18838       -3.18838
+   0.770256       0.770256
+   2.09481        2.09481
+   5.36062        5.36062
+  -5.25216       -5.25216
+  -6.9523        -6.9523
+   3.97384        3.97384
+   8.7784         8.7784
+  -3.91837       -3.91837
+  -9.08965       -9.08965
+  -1.17883       -1.17883
+  -4.21353       -4.21353
+  -5.0915        -5.0915
+   3.74499        3.74499
+  -4.39715       -4.39715
+   2.13732        2.13732
+   5.97568        5.97568
+   1.11809        1.11809
+  -3.93191       -3.93191
+  -1.39764       -1.39764
+  -4.23595       -4.23595
+   0.103914       0.103914
+  -2.34387       -2.34387
+  -4.95433       -4.95433
+   3.58645        3.58645
+   0.818317       0.818317
+   6.23266        6.23266
+  -5.62973       -5.62973
+  -7.45604       -7.45604
+   1.29222        1.29222
+   0.327714       0.327714
+   5.31996        5.31996
+  -2.23663       -2.23663
+   0.058689       0.058689
+  -0.74368       -0.74368
+  -1.20749       -1.20749
+  -4.75414       -4.75414
+   2.10011        2.10011
+  -6.86479       -6.86479
+   1.58403        1.58403
+   0.0492497      0.0492497
+   0.32083        0.32083
+  -3.11682       -3.11682
+   4.61797        4.61797
+  -0.399561      -0.399561
+  -7.89927       -7.89927
+  -0.659676      -0.659676
+  -2.2416        -2.2416
+   0.933026       0.933026
+   1.98848        1.98848
+  -2.14547       -2.14547
+  -1.10747       -1.10747
+   8.90983        8.90983
+  -3.84128       -3.84128
+   9.82771        9.82771
+   3.02843        3.02843
+   3.26396        3.26396
+   6.75629        6.75629
+   0.0290972      0.0290972
+   7.92768        7.92768
+   7.44608        7.44608
+  -4.14083       -4.14083
+  -1.39636       -1.39636
+   2.87656        2.87656
+   3.87446        3.87446
+   0.112521       0.112521
+  -3.3429        -3.3429
+  -6.85823       -6.85823
+   1.18408        1.18408
+   3.53175        3.53175
+   3.56147        3.56147
+   5.41961        5.41961
+  -1.5263        -1.5263
+   3.05559        3.05559
+  -5.7201        -5.7201
+  -3.98882       -3.98882
+  -0.131939      -0.131939
+   6.25683        6.25683
+   0.712945       0.712945
+   4.17266        4.17266
+   9.04425        9.04425
+  -2.39179       -2.39179
+   3.03807        3.03807
+   5.79693        5.79693
+  -5.28875       -5.28875
+  -2.56482       -2.56482
+  -1.00679       -1.00679
+  -0.512488      -0.512488
+  -4.60373       -4.60373
+  -2.69188       -2.69188
+   0.958182       0.958182
+  -1.08075       -1.08075
+   2.66033        2.66033
+  -5.77563       -5.77563
+   5.393          5.393
+   0.822122       0.822122
+   3.50281        3.50281
+  -1.90373       -1.90373
+  -3.41986       -3.41986
+  -7.32502       -7.32502
+  -2.0256        -2.0256
+  -6.28488       -6.28488
+   0.358393       0.358393
+   1.89312        1.89312
+  -0.709162      -0.709162
+  -4.43491       -4.43491
+  -3.56097       -3.56097
+  -8.3806        -8.3806
+  -5.56256       -5.56256
+  -3.40994       -3.40994
+  -6.15002       -6.15002
+   0.949459       0.949459
+   3.18256        3.18256
+   6.31834        6.31834
+  12.4998        12.4998
+  -6.16927       -6.16927
+  -1.73781       -1.73781
+   0.274813       0.274813
+   7.11001        7.11001
+   6.79962        6.79962
+   2.00121        2.00121
+  -4.30592       -4.30592
+  -2.38345       -2.38345
+   7.50502        7.50502
+  -3.56375       -3.56375
+  -1.07828       -1.07828
+   7.4632         7.4632
+  -5.78317       -5.78317
+  -0.54432       -0.54432
+   8.82699        8.82699
+  -2.51939       -2.51939
+  -3.21417       -3.21417
+   3.06052        3.06052
+  -0.45856       -0.45856
+   8.89456        8.89456
+   5.89006        5.89006
+   1.01204        1.01204
+   4.9875         4.9875
+  -1.63          -1.63
+   1.35424        1.35424
+   3.72608        3.72608
+  -8.53795       -8.53795
+  -5.93051       -5.93051
+  -2.35685       -2.35685
+   3.51823        3.51823
+   3.65767        3.65767
+  -3.04233       -3.04233
+  -1.12453       -1.12453
+  -1.68299       -1.68299
+  -5.69175       -5.69175
+   3.66601        3.66601
+  -3.11779       -3.11779
+  -0.20161       -0.20161
+   0.78317        0.78317
+   2.28035        2.28035
+  -4.43493       -4.43493
+   2.12557        2.12557
+   6.97219        6.97219
+   4.91357        4.91357
+  -1.87778       -1.87778
+   1.98163        1.98163
+   1.01184        1.01184
+   0.0544142      0.0544142
+  -0.748318      -0.748318
+  10.0677        10.0677
+  -5.50226       -5.50226
+   3.89987        3.89987
+   1.38136        1.38136
+   4.67073        4.67073
+   5.3372         5.3372
+  -1.29886       -1.29886
+  -0.965173      -0.965173
+   0.546909       0.546909
+   5.87692        5.87692
+ -10.1356       -10.1356
+   0.541422       0.541422
+   0.486656       0.486656
+   8.42395        8.42395
+  -4.04554       -4.04554
+  11.4728        11.4728
+  -6.54655       -6.54655
+   6.90602        6.90602
+ -13.8383       -13.8383
+   2.64142        2.64142
+   3.96547        3.96547
+  -0.887154      -0.887154
+   0.0442338      0.0442338
+  -5.12331       -5.12331
+   4.95632        4.95632
+   3.15264        3.15264
+   4.80494        4.80494
+  -5.42313       -5.42313
+  -4.2795        -4.2795
+   1.661          1.661
+   3.85204        3.85204
+  10.1308        10.1308
+  -4.34526       -4.34526
+  -5.49571       -5.49571
+   3.92939        3.92939
+  -3.28527       -3.28527
+   0.154911       0.154911
+  -3.606         -3.606
+   5.91814        5.91814
+  -8.85249       -8.85249
+   9.38796        9.38796
+  -0.800741      -0.800741
+  -2.87508       -2.87508
+   2.99955        2.99955
+  -7.13252       -7.13252
+  -6.77081       -6.77081
+  -2.28359       -2.28359
+  -0.180517      -0.180517
+   7.04622        7.04622
+   4.2577         4.2577
+  -4.73655       -4.73655
+  -0.249759      -0.249759
+   2.4412         2.4412
+   8.47175        8.47175
+  -3.24927       -3.24927
+ -12.5242       -12.5242
+  -2.74845       -2.74845
+  -9.32786       -9.32786
+   4.21624        4.21624
+   2.94687        2.94687
+   3.35216        3.35216
+  -3.5485        -3.5485
+   6.97298        6.97298
+   2.01617        2.01617
+   4.70745        4.70745
+   2.96924        2.96924
+  -0.18365       -0.18365
+  -0.694247      -0.694247
+  -7.14459       -7.14459
+   5.38548        5.38548
+   2.04923        2.04923
+  -5.33216       -5.33216
+   5.47927        5.47927
+   0.357422       0.357422
+   4.36552        4.36552
+   6.88375        6.88375
+  -6.47244       -6.47244
+  -3.40726       -3.40726
+  -6.56449       -6.56449
+   6.34818        6.34818
+  -4.23984       -4.23984
+ -11.1113       -11.1113
+   2.41915        2.41915
+   3.90153        3.90153
+  -7.69422       -7.69422
+  -8.03709       -8.03709
+  -9.64719       -9.64719
+  -4.04416       -4.04416
+   2.64435        2.64435
+   5.11566        5.11566
+  -1.27873       -1.27873
+  -1.01265       -1.01265
+  -8.38716       -8.38716
+  -0.960571      -0.960571
+   2.05458        2.05458
+  -1.89606       -1.89606
+  -7.04401       -7.04401
+   4.91798        4.91798
+   2.12484        2.12484
+   2.38768        2.38768
+   7.9691         7.9691
+  -1.00886       -1.00886
+  -4.9569        -4.9569
+  -4.74278       -4.74278
+   0.191814       0.191814
+  -5.2925        -5.2925
+  -1.15484       -1.15484
+   2.27898        2.27898
+   4.12308        4.12308
+  -6.18988       -6.18988
+   7.1232         7.1232
+  -6.68678       -6.68678
+   1.65808        1.65808
+   8.53283        8.53283
+   0.509069       0.509069
+  -3.03638       -3.03638
+  -4.86641       -4.86641
+   7.20729        7.20729
+  -7.51236       -7.51236
+   3.37738        3.37738
+  -0.0649395     -0.0649395
+   2.75749        2.75749
+  -5.61535       -5.61535
+   3.1237         3.1237
+  -0.766488      -0.766488
+   4.39047        4.39047
+   1.28616        1.28616
+  -8.02003       -8.02003
+   4.21688        4.21688
+  -2.79942       -2.79942
+  -5.80171       -5.80171
+   9.97235        9.97235
+  21.8011        21.8011
+  -3.58992       -3.58992
+   5.03481        5.03481
+  -2.1684        -2.1684
+  -5.46844       -5.46844
+   1.57702        1.57702
+  -4.53923       -4.53923
+  -1.77363       -1.77363
+  -0.489051      -0.489051
+  -0.371992      -0.371992
+   8.264          8.264
+   1.63502        1.63502
+  -1.10134       -1.10134
+   4.76612        4.76612
+   5.93085        5.93085
+  -2.07348       -2.07348
+   4.26074        4.26074
+   4.1331         4.1331
+  11.1442        11.1442
+   2.18824        2.18824
+   2.18854        2.18854
+   0.210843       0.210843
+  -9.30743       -9.30743
+   5.34539        5.34539
+  -4.21419       -4.21419
+  -3.97284       -3.97284
+  -2.67745       -2.67745
+   4.17366        4.17366
+   2.41498        2.41498
+   0.801359       0.801359
+   8.35766        8.35766
+  -1.29589       -1.29589
+  -7.45531       -7.45531
+  -7.26731       -7.26731
+   4.06669        4.06669
+  -2.35771       -2.35771
+  -8.73174       -8.73174
+  -0.837329      -0.837329
+  -2.53419       -2.53419
+  44.3977        44.3977
+  13.5049        13.5049
+  -3.66878       -3.66878
+  -6.5533        -6.5533
+  -5.59814       -5.59814
+ -10.5759       -10.5759
+   0.663108       0.663108
+  -3.45147       -3.45147
+  -3.75944       -3.75944
+   1.84721        1.84721
+  -0.363204      -0.363204
+   4.54678        4.54678
+   2.07408        2.07408
+   7.85227        7.85227
+  -7.53707       -7.53707
+   4.18344        4.18344
+  -1.96048       -1.96048
+   6.24217        6.24217
+  -9.16295       -9.16295
+   0.0480544      0.0480544
+   2.84725        2.84725
+   1.08008        1.08008
+  -0.874464      -0.874464
+   1.67428        1.67428
+  -1.91245       -1.91245
+   3.53596        3.53596
+   3.75983        3.75983
+   1.37903        1.37903
+  -0.799744      -0.799744
+   2.75015        2.75015
+ -11.0835       -11.0835
+  -1.6781        -1.6781
+   2.86463        2.86463
+ -11.1467       -11.1467
+  -3.76398       -3.76398
+   9.06439        9.06439
+   9.84403        9.84403
+  -5.07          -5.07
+   3.2952         3.2952
+  -1.62527       -1.62527
+  -7.98997       -7.98997
+  -7.8193        -7.8193
+   1.10895        1.10895
+   0.460921       0.460921
+  -1.47816       -1.47816
+   0.718936       0.718936
+  -3.74006       -3.74006
+  -2.87535       -2.87535
+   0.037427       0.037427
+  -4.49959       -4.49959
+   0.0987492      0.0987492
+   1.8443         1.8443
+   0.748879       0.748879
+   1.4364         1.4364
+  -0.90809       -0.90809
+  -1.36403       -1.36403
+  -1.27123       -1.27123
+   3.09447        3.09447
+  -3.82708       -3.82708
+   0.683696       0.683696
+   3.96997        3.96997
+   0.461267       0.461267
+   4.96801        4.96801
+  -5.96169       -5.96169
+   2.56714        2.56714
+ -10.7519       -10.7519
+  -3.39381       -3.39381
+   1.15623        1.15623
+  -3.95798       -3.95798
+  -1.42797       -1.42797
+   4.85734        4.85734
+  -4.46424       -4.46424
+ -11.9172       -11.9172
+   0.740766       0.740766
+  -2.06857       -2.06857
+  -1.23723       -1.23723
+  -6.43373       -6.43373
+   7.04893        7.04893
+  -1.10208       -1.10208
+  -0.0507102     -0.0507102
+   8.23443        8.23443
+  -1.71378       -1.71378
+   2.769          2.769
+   9.77752        9.77752
+   0.423859       0.423859
+   0.901832       0.901832
+   0.0738559      0.0738559
+  -0.487266      -0.487266
+   2.05358        2.05358
+  -8.73912       -8.73912
+   3.01532        3.01532
+  -0.926127      -0.926127
+ -11.2315       -11.2315
+   1.79698        1.79698
+ -13.074        -13.074
+   3.72342        3.72342
+  -9.17341       -9.17341
+   7.23722        7.23722
+   3.85919        3.85919
+  -4.10267       -4.10267
+   5.89157        5.89157
+  -1.06631       -1.06631
+  -2.18366       -2.18366
+  -0.0316413     -0.0316413
+  -8.63864       -8.63864
+  -0.194451      -0.194451
+   2.71759        2.71759
+  -5.19424       -5.19424
+ -16.7634       -16.7634
+   5.97943        5.97943
+   0.319596       0.319596
+ -10.0687       -10.0687
+   1.12736        1.12736
+   2.11687        2.11687
+   2.5643         2.5643
+   0.502174       0.502174
+  -5.75011       -5.75011
+ -11.1808       -11.1808
+  -3.42246       -3.42246
+   7.55982        7.55982
+  -5.85592       -5.85592
+   1.22363        1.22363
+   1.39871        1.39871
+   3.35581        3.35581
+   2.99389        2.99389
+  -0.762194      -0.762194
+   1.39891        1.39891
+  -4.24295       -4.24295
+  -6.95612       -6.95612
+   7.00699        7.00699
+ -30.893        -30.893
+  -7.3071        -7.3071
+  17.5017        17.5017
+  -3.26283       -3.26283
+  -4.13569       -4.13569
+   4.33006        4.33006
+  -5.94055       -5.94055
+  -0.564017      -0.564017
+   5.60949        5.60949
+   7.50747        7.50747
+  -4.08147       -4.08147
+   4.08671        4.08671
+   6.72008        6.72008
+  -5.02883       -5.02883
+  -3.48779       -3.48779
+   4.76881        4.76881
+   4.5818         4.5818
+  -3.10608       -3.10608
+  -5.08198       -5.08198
+  -5.54477       -5.54477
+ -13.1989       -13.1989
+  -8.63604       -8.63604
+  -0.688683      -0.688683
+  -2.34276       -2.34276
+  -3.19008       -3.19008
+   0.204818       0.204818
+   0.639057       0.639057
+  12.6767        12.6767
+  -3.40057       -3.40057
+  -6.36799       -6.36799
+   3.7564         3.7564
+  -3.04825       -3.04825
+  -3.98011       -3.98011
+  -2.21944       -2.21944
+   8.40757        8.40757
+  -5.6418        -5.6418
+   3.3001         3.3001
+  -0.678107      -0.678107
+  -2.42254       -2.42254
+   0.439524       0.439524
+  -0.417505      -0.417505
+  -4.98938       -4.98938
+  -6.34015       -6.34015
+  -4.84203       -4.84203
+   2.86778        2.86778
+   3.29409        3.29409
+   2.59772        2.59772
+   5.20187        5.20187
+   3.55625        3.55625
+  -7.065         -7.065
+  -6.60792       -6.60792
+  -3.20259       -3.20259
+   0.417062       0.417062
+  -2.39846       -2.39846
+  -5.762         -5.762
+   1.74843        1.74843
+   8.19239        8.19239
+  -1.7349        -1.7349
+  -0.0331415     -0.0331415
+   5.00712        5.00712
+  10.611         10.611
+   9.28817        9.28817
+  -3.85324       -3.85324
+   2.29622        2.29622
+  10.962         10.962
+   4.44034        4.44034
+  -3.2265        -3.2265
+   1.39326        1.39326
+  -1.56539       -1.56539
+  -8.78843       -8.78843
+  -1.74101       -1.74101
+   8.51953        8.51953
+   3.31178        3.31178
+  -1.20051       -1.20051
+  -3.93224       -3.93224
+   2.4431         2.4431
+   3.69278        3.69278
+ -10.2714       -10.2714
+ -13.7579       -13.7579
+  -1.76844       -1.76844
+  -0.448193      -0.448193
+   1.48574        1.48574
+  -0.831377      -0.831377
+   6.42657        6.42657
+   6.51848        6.51848
+   2.7764         2.7764
+   4.29448        4.29448
+  -1.27173       -1.27173
+  -7.14856       -7.14856
+   2.95751        2.95751
+   2.39789        2.39789
+   4.79429        4.79429
+   7.29216        7.29216
+  -4.91502       -4.91502
+   2.38701        2.38701
+  -2.34997       -2.34997
+  -0.876115      -0.876115
+  -0.672649      -0.672649
+   4.43884        4.43884
+   0.254258       0.254258
+  -3.56471       -3.56471
+   0.161779       0.161779
+ -10.1128       -10.1128
+   9.97279        9.97279
+  -5.01498       -5.01498
+   1.10415        1.10415
+   1.37993        1.37993
+  -3.32619       -3.32619
+   2.57257        2.57257
+  -0.137478      -0.137478
+   1.49426        1.49426
+  -0.805644      -0.805644
+   3.25356        3.25356
+   2.46332        2.46332
+   1.39266        1.39266
+   4.15167        4.15167
+  -9.27164       -9.27164
+  -2.29794       -2.29794
+   0.067971       0.067971
+   3.83697        3.83697
+   5.7385         5.7385
+  -6.15176       -6.15176
+  -4.08442       -4.08442
+  -6.18563       -6.18563
+   6.44396        6.44396
+   5.63585        5.63585
+   1.21604        1.21604
+  11.1837        11.1837
+   2.29144        2.29144
+  -0.995473      -0.995473
+   5.22826        5.22826
+   9.27205        9.27205
+  -7.23457       -7.23457
+   6.29887        6.29887
+   2.48343        2.48343
+  -4.96111       -4.96111
+  -5.52811       -5.52811
+  -4.40855       -4.40855
+  -5.69429       -5.69429
+  -1.12765       -1.12765
+  -0.22142       -0.22142
+  -5.96815       -5.96815
+   4.55923        4.55923
+  -1.05719       -1.05719
+   2.07986        2.07986
+   7.77539        7.77539
+  -2.03581       -2.03581
+   0.270705       0.270705
+   0.126658       0.126658
+  -6.1672        -6.1672
+ -16.0576       -16.0576
+   0.635198       0.635198
+   8.55006        8.55006
+  -2.93081       -2.93081
+  -1.7657        -1.7657
+  -0.37886       -0.37886
+  -2.4086        -2.4086
+   1.41889        1.41889
+  -1.40539       -1.40539
+   0.963807       0.963807
+  -2.14947       -2.14947
+  -6.31832       -6.31832
+  -4.30827       -4.30827
+   6.2609         6.2609
+  -8.36351       -8.36351
+   4.28564        4.28564
+   0.646361       0.646361
+   4.60485        4.60485
+  -3.1664        -3.1664
+  -0.611618      -0.611618
+  -9.53534       -9.53534
+   1.92275        1.92275
+  -8.1521        -8.1521
+   0.101441       0.101441
+   0.399002       0.399002
+  -2.04551       -2.04551
+  -4.5564        -4.5564
+   3.0555         3.0555
+   0.992401       0.992401
+  -5.62638       -5.62638
+  -0.46873       -0.46873
+  -6.86208       -6.86208
+  -2.77108       -2.77108
+   3.51118        3.51118
+   0.885266       0.885266
+   3.65701        3.65701
+   6.88336        6.88336
+  -7.25948       -7.25948
+   7.31435        7.31435
+  -6.57357       -6.57357
+   3.67947        3.67947
+   4.80901        4.80901
+  -2.80342       -2.80342
+   5.78724        5.78724
+   5.30985        5.30985
+   7.24724        7.24724
+  -1.30439       -1.30439
+   2.50975        2.50975
+   5.28538        5.28538
+  -3.91583       -3.91583
+   2.98722        2.98722
+   5.31167        5.31167
+  -0.596966      -0.596966
+  -4.94141       -4.94141
+   4.59005        4.59005
+   1.3813         1.3813
+   4.0611         4.0611
+  -0.747616      -0.747616
+  -3.1697        -3.1697
+  -1.70787       -1.70787
+  -2.43542       -2.43542
+  -5.86823       -5.86823
+ -10.9093       -10.9093
+   5.20087        5.20087
+  -6.40378       -6.40378
+   1.5149         1.5149
+  -6.52874       -6.52874
+  -5.69743       -5.69743
+   1.06819        1.06819
+  -7.31776       -7.31776
+   3.69649        3.69649
+  -4.21319       -4.21319
+  -4.91507       -4.91507
+   5.44776        5.44776
+  -0.708927      -0.708927
+   1.94895        1.94895
+   2.90927        2.90927
+  -2.82547       -2.82547
+  -1.79858       -1.79858
+ -15.6727       -15.6727
+  -0.308918      -0.308918
+   2.61943        2.61943
+  -3.89041       -3.89041
+  -1.84684       -1.84684
+  -6.80446       -6.80446
+   3.97398        3.97398
+   2.31201        2.31201
+   4.29417        4.29417
+  -1.24479       -1.24479
+   4.25927        4.25927
+  -1.96968       -1.96968
+   0.703519       0.703519
+   2.06517        2.06517
+   0.920347       0.920347
+   6.22843        6.22843
+   1.86167        1.86167
+   0.43407        0.43407
+   1.25225        1.25225
+  -0.00512493    -0.00512493
+  -1.70887       -1.70887
+   0.725693       0.725693
+   6.11604        6.11604
+  -5.87059       -5.87059
+   3.26102        3.26102
+   2.0488         2.0488
+  -0.0544172     -0.0544172
+   2.57295        2.57295
+  -1.10578       -1.10578
+   2.43904        2.43904
+  -2.34604       -2.34604
+   3.2098         3.2098
+   2.16089        2.16089
+  -9.35001       -9.35001
+   9.43924        9.43924
+   0.916747       0.916747
+   2.59533        2.59533
+  -1.84596       -1.84596
+   1.02889        1.02889
+   0.755944       0.755944
+   8.28274        8.28274
+  -3.21136       -3.21136
+   1.24897        1.24897
+  -0.363928      -0.363928
+   2.37533        2.37533
+  -1.5794        -1.5794
+   6.67417        6.67417
+  -4.4632        -4.4632
+   8.53731        8.53731
+  -1.16526       -1.16526
+  -0.51467       -0.51467
+  -4.91688       -4.91688
+   7.17741        7.17741
+   4.61708        4.61708
+  -2.41511       -2.41511
+ -11.5234       -11.5234
+   2.61523        2.61523
+   4.7703         4.7703
+   6.72381        6.72381
+   5.65388        5.65388
+  -4.23963       -4.23963
+   0.925176       0.925176
+   1.98862        1.98862
+  -6.14466       -6.14466
+   2.76728        2.76728
+  -0.83598       -0.83598
+  -4.22593       -4.22593
+   5.99083        5.99083
+  -4.886         -4.886
+   4.37801        4.37801
+   5.77761        5.77761
+   3.38352        3.38352
+  -0.311291      -0.311291
+   8.26669        8.26669
+  -4.94787       -4.94787
+  -9.62034       -9.62034
+   2.37023        2.37023
+   3.41718        3.41718
+  -2.43368       -2.43368
+   3.5898         3.5898
+  -1.21973       -1.21973
+   0.0350305      0.0350305
+  -4.33097       -4.33097
+  -3.41432       -3.41432
+   2.59161        2.59161
+  -2.11239       -2.11239
+  -1.0801        -1.0801
+  -3.27061       -3.27061
+  -0.34025       -0.34025
+  -6.40563       -6.40563
+  -0.522305      -0.522305
+   4.63382        4.63382
+   1.5154         1.5154
+   0.968893       0.968893
+   2.79354        2.79354
+  -0.829942      -0.829942
+  -1.76388       -1.76388
+  -6.64903       -6.64903
+  -8.52588       -8.52588
+   2.70798        2.70798
+   6.78381        6.78381
+  -5.67891       -5.67891
+  -0.0588557     -0.0588557
+  -4.12923       -4.12923
+  -2.70431       -2.70431
+  -0.12131       -0.12131
+   6.59494        6.59494
+   0.830427       0.830427
+   3.40436        3.40436
+   6.98828        6.98828
+  -2.33332       -2.33332
+   5.85244        5.85244
+ -10.0398       -10.0398
+  -0.242519      -0.242519
+  -3.38719       -3.38719
+   2.74288        2.74288
+   3.82961        3.82961
+  -6.85166       -6.85166
+  -0.345431      -0.345431
+  -3.03082       -3.03082
+   1.68089        1.68089
+  -0.785036      -0.785036
+  -2.92804       -2.92804
+   1.03727        1.03727
+   5.51647        5.51647
+  -2.15538       -2.15538
+  -6.20918       -6.20918
+  -0.986195      -0.986195
+  -4.4207        -4.4207
+  -0.314791      -0.314791
+  -6.64843       -6.64843
+   1.255          1.255
+   4.39107        4.39107
+   2.20706        2.20706
+  -1.894         -1.894
+  -3.01471       -3.01471
+  -0.0623641     -0.0623641
+  -5.76316       -5.76316
+  -2.45987       -2.45987
+  -2.09262       -2.09262
+   0.0458748      0.0458748
+   5.09539        5.09539
+  -3.80431       -3.80431
+  -3.90738       -3.90738
+  -6.48843       -6.48843
+  -2.58373       -2.58373
+  -6.38764       -6.38764
+   7.38858        7.38858
+   0.492176       0.492176
+   8.79347        8.79347
+   2.04442        2.04442
+  -0.216083      -0.216083
+  11.3375        11.3375
+  -3.4177        -3.4177
+   3.90111        3.90111
+   4.92081        4.92081
+   4.45964        4.45964
+  11.1458        11.1458
+  -2.2688        -2.2688
+  -4.43463       -4.43463
+  -4.22186       -4.22186
+  -5.93987       -5.93987
+   3.4437         3.4437
+  -5.60816       -5.60816
+  -8.04401       -8.04401
+  -4.95256       -4.95256
+   3.88283        3.88283
+  -0.173935      -0.173935
+  -2.63243       -2.63243
+  -1.03812       -1.03812
+  -9.14078       -9.14078
+  -6.1411        -6.1411
+   3.4284         3.4284
+  -9.8305        -9.8305
+   6.76115        6.76115
+ -11.3646       -11.3646
+  -5.7296        -5.7296
+  -2.41831       -2.41831
+  -5.21505       -5.21505
+  10.4347        10.4347
+   2.06721        2.06721
+   1.02265        1.02265
+  -6.93537       -6.93537
+   1.28707        1.28707
+   0.939615       0.939615
+  11.262         11.262
+   1.2805         1.2805
+   4.8619         4.8619
+   3.15836        3.15836
+  -5.18747       -5.18747
+  -2.98078       -2.98078
+  -2.0489        -2.0489
+  -2.85634       -2.85634
+  -4.56059       -4.56059
+  -4.0715        -4.0715
+   0.469543       0.469543
+  -2.05188       -2.05188
+  -2.79567       -2.79567
+   3.82027        3.82027
+   2.55175        2.55175
+  -0.468207      -0.468207
+  -5.65994       -5.65994
+   2.13508        2.13508
+  -3.17019       -3.17019
+   6.53032        6.53032
+  -4.98714       -4.98714
+  -1.94956       -1.94956
+  -3.08465       -3.08465
+   8.11664        8.11664
+   8.86283        8.86283
+   0.84108        0.84108
+   5.22353        5.22353
+  -3.45671       -3.45671
+  -1.38725       -1.38725
+   1.35206        1.35206
+ -10.4407       -10.4407
+  -2.20051       -2.20051
+  -0.228019      -0.228019
+  -1.38039       -1.38039
+  11.1342        11.1342
+   5.17568        5.17568
+  -4.54852       -4.54852
+  -1.26392       -1.26392
+   5.69792        5.69792
+  -4.90866       -4.90866
+   2.84526        2.84526
+  10.9699        10.9699
+  12.9756        12.9756
+   8.48223        8.48223
+   2.11902        2.11902
+   3.74471        3.74471
+  -5.14437       -5.14437
+ -14.7206       -14.7206
+   3.01028        3.01028
+  -2.67988       -2.67988
+  -2.88296       -2.88296
+  -4.95895       -4.95895
+  -1.82286       -1.82286
+   5.23419        5.23419
+  -2.23867       -2.23867
+   0.610838       0.610838
+   2.09177        2.09177
+   5.74677        5.74677
+   3.6242         3.6242
+   2.0758         2.0758
+  -2.85159       -2.85159
+  -3.93562       -3.93562
+   3.85649        3.85649
+  -5.75638       -5.75638
+  -7.07444       -7.07444
+   0.907402       0.907402
+  -8.92532       -8.92532
+  -4.09782       -4.09782
+   1.85777        1.85777
+   5.73041        5.73041
+  -2.17118       -2.17118
+  -3.4713        -3.4713
+   7.95825        7.95825
+   9.10838        9.10838
+   1.80182        1.80182
+  -0.54593       -0.54593
+  -4.89919       -4.89919
+  -2.97982       -2.97982
+   0.807424       0.807424
+  -2.27          -2.27
+ -13.2338       -13.2338
+  -3.94367       -3.94367
+  -5.72938       -5.72938
+  -2.42243       -2.42243
+  -3.69581       -3.69581
+  -4.71307       -4.71307
+   1.38983        1.38983
+  -5.37869       -5.37869
+  -6.82815       -6.82815
+   2.73203        2.73203
+  13.6495        13.6495
+  -6.29731       -6.29731
+  -8.43712       -8.43712
+  14.1567        14.1567
+  -0.978804      -0.978804
+   1.26264        1.26264
+  -9.25575       -9.25575
+  -8.10968       -8.10968
+  -3.98015       -3.98015
+   6.60273        6.60273
+  -3.98373       -3.98373
+   1.35817        1.35817
+   1.20988        1.20988
+   1.53069        1.53069
+   4.08368        4.08368
+  -2.38429       -2.38429
+  -4.67381       -4.67381
+  -5.49726       -5.49726
+   0.657715       0.657715
+  -0.00123905    -0.00123905
+   4.62712        4.62712
+  -0.317445      -0.317445
+  -5.08829       -5.08829
+  -9.85674       -9.85674
+   5.31787        5.31787
+   1.61793        1.61793
+   3.9901         3.9901
+  -1.04243       -1.04243
+  -3.73679       -3.73679
+   0.670282       0.670282
+   9.03148        9.03148
+  -4.77058       -4.77058
+   8.60147        8.60147
+  -0.664744      -0.664744
+   1.97711        1.97711
+  -5.35794       -5.35794
+  -9.70033       -9.70033
+  10.7781        10.7781
+   1.96443        1.96443
+   1.84069        1.84069
+ -12.0109       -12.0109
+   2.08404        2.08404
+   3.64031        3.64031
+   8.65585        8.65585
+ -11.8355       -11.8355
+   9.89404        9.89404
+   0.279063       0.279063
+  -0.315296      -0.315296
+   3.74263        3.74263
+   6.54645        6.54645
+   5.43941        5.43941
+   4.83252        4.83252
+   1.70716        1.70716
+  -3.27497       -3.27497
+  -3.07764       -3.07764
+   9.25309        9.25309
+  -1.69559       -1.69559
+  10.1694        10.1694
+  -3.42523       -3.42523
+   6.39435        6.39435
+   2.18084        2.18084
+   1.33177        1.33177
+  -0.709393      -0.709393
+   1.44799        1.44799
+   0.881759       0.881759
+  -2.35085       -2.35085
+  -1.91407       -1.91407
+   0.302603       0.302603
+   1.40288        1.40288
+  -2.37323       -2.37323
+  -7.74084       -7.74084
+  -7.73224       -7.73224
+   2.8793         2.8793
+   6.62065        6.62065
+   1.4654         1.4654
+  -0.982735      -0.982735
+  -0.97328       -0.97328
+  -8.38882       -8.38882
+   8.74643        8.74643
+  -7.86996       -7.86996
+   3.25655        3.25655
+   2.78551        2.78551
+  -5.17511       -5.17511
+   4.90515        4.90515
+   0.28899        0.28899
+   3.57292        3.57292
+  -5.25376       -5.25376
+  -8.57274       -8.57274
+  -1.18267       -1.18267
+  37.4072        37.4072
+  -4.00801       -4.00801
+   4.8073         4.8073
+  -4.45001       -4.45001
+   7.66024        7.66024
+  -4.47725       -4.47725
+ -10.2209       -10.2209
+  -4.80026       -4.80026
+  -0.64446       -0.64446
+  -0.899171      -0.899171
+   1.09833        1.09833
+  -0.988097      -0.988097
+   2.82126        2.82126
+  -8.19269       -8.19269
+  -2.64922       -2.64922
+  -9.16004       -9.16004
+  -2.39588       -2.39588
+  -4.72025       -4.72025
+   2.34077        2.34077
+   3.83879        3.83879
+   1.9499         1.9499
+  -0.361603      -0.361603
+   7.79929        7.79929
+   2.34774        2.34774
+  -8.21052       -8.21052
+  -2.02077       -2.02077
+  -1.58017       -1.58017
+  -0.410542      -0.410542
+ -10.7206       -10.7206
+   3.26874        3.26874
+   2.80972        2.80972
+   0.0906836      0.0906836
+  -1.64773       -1.64773
+   6.49353        6.49353
+  -0.791109      -0.791109
+   4.71404        4.71404
+   0.0741314      0.0741314
+  -0.414415      -0.414415
+   6.84572        6.84572
+  -0.367457      -0.367457
+   1.17563        1.17563
+   0.51039        0.51039
+   4.40348        4.40348
+   0.978932       0.978932
+   3.79206        3.79206
+   4.57632        4.57632
+   2.77883        2.77883
+   0.490867       0.490867
+  -0.151798      -0.151798
+   6.72243        6.72243
+   4.77773        4.77773
+  -0.50633       -0.50633
+  -8.08639       -8.08639
+   4.88619        4.88619
+  -2.07669       -2.07669
+  -2.24093       -2.24093
+   1.72994        1.72994
+  -7.45157       -7.45157
+ -12.1192       -12.1192
+   1.4328         1.4328
+  -8.14432       -8.14432
+  -6.25485       -6.25485
+   0.516865       0.516865
+   7.11864        7.11864
+  -0.616318      -0.616318
+  -0.761916      -0.761916
+  -5.99496       -5.99496
+  10.4321        10.4321
+  -0.516052      -0.516052
+  -5.68287       -5.68287
+  -4.15541       -4.15541
+   1.56619        1.56619
+ -20.8292       -20.8292
+   0.788033       0.788033
+   3.34264        3.34264
+   3.70493        3.70493
+  -0.0822138     -0.0822138
+   2.31304        2.31304
+  -1.69352       -1.69352
+   2.10396        2.10396
+   7.2613         7.2613
+  -1.81799       -1.81799
+  -2.09968       -2.09968
+  -3.8336        -3.8336
+  -3.93478       -3.93478
+   3.3059         3.3059
+   4.19189        4.19189
+  -1.93794       -1.93794
+   2.7117         2.7117
+   9.43261        9.43261
+  -1.83318       -1.83318
+  -1.12685       -1.12685
+   2.40725        2.40725
+   7.50947        7.50947
+   7.65688        7.65688
+  -5.02792       -5.02792
+  -2.55777       -2.55777
+  -1.9946        -1.9946
+  -0.126192      -0.126192
+  -3.30905       -3.30905
+  -0.209775      -0.209775
+   9.06409        9.06409
+  -3.79201       -3.79201
+   8.80185        8.80185
+  -1.59367       -1.59367
+  -2.49213       -2.49213
+  -3.5242        -3.5242
+   2.4892         2.4892
+   5.68222        5.68222
+   4.29073        4.29073
+   0.490494       0.490494
+   3.31313        3.31313
+   8.27344        8.27344
+   1.44936        1.44936
+   5.94283        5.94283
+  -5.90497       -5.90497
+   0.316931       0.316931
+   1.93975        1.93975
+  -1.33405       -1.33405
+  -4.17957       -4.17957
+   2.45999        2.45999
+ -10.0965       -10.0965
+   0.648564       0.648564
+  -0.745957      -0.745957
+  -6.08922       -6.08922
+  -6.45851       -6.45851
+   2.70093        2.70093
+  -2.59331       -2.59331
+  -2.73319       -2.73319
+  -6.50584       -6.50584
+   4.14167        4.14167
+   6.78757        6.78757
+   4.63335        4.63335
+   2.01754        2.01754
+   3.97717        3.97717
+   2.73775        2.73775
+   2.04299        2.04299
+   7.03044        7.03044
+  -8.59414       -8.59414
+  -4.19956       -4.19956
+   0.0135157      0.0135157
+  -5.45393       -5.45393
+   2.75578        2.75578
+   0.730278       0.730278
+  -0.410035      -0.410035
+  10.7831        10.7831
+  -2.82537       -2.82537
+   1.85601        1.85601
+   1.68496        1.68496
+   2.75249        2.75249
+   9.40848        9.40848
+   1.6032         1.6032
+  -3.91263       -3.91263
+   1.12247        1.12247
+  -3.46516       -3.46516
+  -1.48668       -1.48668
+   6.7676         6.7676
+  -5.76927       -5.76927
+  -2.19943       -2.19943
+  -1.61329       -1.61329
+   3.35791        3.35791
+  -7.80737       -7.80737
+   3.06567        3.06567
+ -12.2037       -12.2037
+  12.541         12.541
+   4.42316        4.42316
+   6.48419        6.48419
+   1.17664        1.17664
+   2.97986        2.97986
+  -8.63966       -8.63966
+   0.241757       0.241757
+  -5.03654       -5.03654
+  -1.94594       -1.94594
+  12.8093        12.8093
+  -3.58644       -3.58644
+  -3.35952       -3.35952
+  -0.864134      -0.864134
+ -12.4807       -12.4807
+  -1.69909       -1.69909
+  -5.67676       -5.67676
+ -10.6435       -10.6435
+  -3.86815       -3.86815
+   4.20674        4.20674
+  -4.94992       -4.94992
+   7.63289        7.63289
+  -5.5226        -5.5226
+   1.58362        1.58362
+   1.14864        1.14864
+   5.98635        5.98635
+  11.9692        11.9692
+  -0.208588      -0.208588
+  -0.177219      -0.177219
+   6.35143        6.35143
+  -2.21028       -2.21028
+   0.693657       0.693657
+   2.66882        2.66882
+  -0.494413      -0.494413
+  10.9482        10.9482
+   2.9522         2.9522
+   1.69427        1.69427
+  -5.54007       -5.54007
+  -1.44208       -1.44208
+  -2.75377       -2.75377
+   7.62773        7.62773
+  -0.0991657     -0.0991657
+   0.541024       0.541024
+   0.383422       0.383422
+  -6.28538       -6.28538
+  -3.63239       -3.63239
+   5.54891        5.54891
+   4.38377        4.38377
+  -4.21607       -4.21607
+  -1.58462       -1.58462
+   1.99568        1.99568
+   1.70177        1.70177
+   1.65142        1.65142
+   1.79811        1.79811
+  -6.82605       -6.82605
+   3.65159        3.65159
+   2.60935        2.60935
+  -2.91237       -2.91237
+  -1.56808       -1.56808
+  -3.07334       -3.07334
+   0.883426       0.883426
+  -1.59697       -1.59697
+   4.44432        4.44432
+  -2.72255       -2.72255
+  -0.853149      -0.853149
+  -0.132598      -0.132598
+  -0.63629       -0.63629
+  -3.69308       -3.69308
+  -7.18449       -7.18449
+   1.20547        1.20547
+  14.3427        14.3427
+   5.08288        5.08288
+   0.957041       0.957041
+   0.153537       0.153537
+  -7.14906       -7.14906
+  -8.78572       -8.78572
+   4.05049        4.05049
+   3.22929        3.22929
+  -3.34601       -3.34601
+   3.86442        3.86442
+  -2.80641       -2.80641
+   6.51055        6.51055
+  -4.58706       -4.58706
+  -1.51146       -1.51146
+   3.88212        3.88212
+   1.89549        1.89549
+   3.50062        3.50062
+  -1.43005       -1.43005
+  -2.91969       -2.91969
+  -6.52573       -6.52573
+  -3.8843        -3.8843
+  -8.34716       -8.34716
+  -7.42192       -7.42192
+  -3.98985       -3.98985
+  15.526         15.526
+   8.70318        8.70318
+  -1.10105       -1.10105
+   2.14694        2.14694
+   7.71484        7.71484
+  -0.0260442     -0.0260442
+  -3.31138       -3.31138
+   1.67906        1.67906
+  -0.083112      -0.083112
+  -8.42905       -8.42905
+  -8.82729       -8.82729
+  11.2859        11.2859
+  -8.07136       -8.07136
+  -3.9371        -3.9371
+  -4.63176       -4.63176
+  -1.23605       -1.23605
+  -2.08565       -2.08565
+   1.93918        1.93918
+ -12.5031       -12.5031
+  -0.442281      -0.442281
+  -5.50289       -5.50289
+  -0.815112      -0.815112
+   0.0898735      0.0898735
+   4.69373        4.69373
+  -7.22004       -7.22004
+   0.543294       0.543294
+   4.2932         4.2932
+   2.12984        2.12984
+  -4.42752       -4.42752
+   3.03694        3.03694
+  -3.73337       -3.73337
+ -12.0483       -12.0483
+  -5.99704       -5.99704
+   0.0707967      0.0707967
+  -4.52239       -4.52239
+   3.65625        3.65625
+  -5.61903       -5.61903
+   9.78971        9.78971
+   8.47575        8.47575
+  -0.320966      -0.320966
+  -7.10339       -7.10339
+   0.485669       0.485669
+   3.19439        3.19439
+  -0.411976      -0.411976
+  -0.782875      -0.782875
+  16.4086        16.4086
+  -2.67312       -2.67312
+   0.73424        0.73424
+   8.32014        8.32014
+  -1.24665       -1.24665
+   3.70031        3.70031
+  -6.22155       -6.22155
+  -6.34804       -6.34804
+  -4.84631       -4.84631
+   7.19111        7.19111
+   2.58937        2.58937
+   2.12044        2.12044
+  -0.304369      -0.304369
+ -11.5161       -11.5161
+  -4.75933       -4.75933
+  -5.40287       -5.40287
+ -14.7511       -14.7511
+ -11.3269       -11.3269
+  -3.40961       -3.40961
+  -8.36998       -8.36998
+  -7.86816       -7.86816
+   3.46638        3.46638
+   5.10745        5.10745
+   9.12589        9.12589
+   4.53119        4.53119
+  -0.0952322     -0.0952322
+  -1.67069       -1.67069
+   1.48937        1.48937
+   2.1548         2.1548
+  -0.680895      -0.680895
+   6.00943        6.00943
+  -6.23597       -6.23597
+  15.2635        15.2635
+  -5.39621       -5.39621
+   2.9004         2.9004
+  -7.2031        -7.2031
+   0.188095       0.188095
+  -5.65511       -5.65511
+   8.80472        8.80472
+   4.77116        4.77116
+  -0.320718      -0.320718
+  -0.094774      -0.094774
+   4.24892        4.24892
+  -0.729715      -0.729715
+   3.46906        3.46906
+  -4.86913       -4.86913
+  -2.05092       -2.05092
+   3.24008        3.24008
+   2.67334        2.67334
+   5.41008        5.41008
+   4.61387        4.61387
+ -11.9338       -11.9338
+   2.15538        2.15538
+   3.39914        3.39914
+   2.71216        2.71216
+   6.79031        6.79031
+  -0.750493      -0.750493
+  -0.683416      -0.683416
+   7.23875        7.23875
+   4.67949        4.67949
+  -2.16467       -2.16467
+   3.64787        3.64787
+  -1.27823       -1.27823
+  -1.43992       -1.43992
+   3.183          3.183
+  -8.60412       -8.60412
+  -5.42757       -5.42757
+  -0.564214      -0.564214
+  -1.17837       -1.17837
+   2.45248        2.45248
+   3.60909        3.60909
+   2.61183        2.61183
+   5.20279        5.20279
+  -1.07145       -1.07145
+  -0.919519      -0.919519
+   3.89898        3.89898
+   3.72175        3.72175
+  -9.9673        -9.9673
+   1.50607        1.50607
+  -0.456562      -0.456562
+  10.9984        10.9984
+  -2.18673       -2.18673
+  -7.39159       -7.39159
+  -5.54389       -5.54389
+   2.6353         2.6353
+   6.87535        6.87535
+ -10.4019       -10.4019
+  -5.51375       -5.51375
+  -3.33244       -3.33244
+   7.60358        7.60358
+  -9.48529       -9.48529
+  -0.514099      -0.514099
+   6.20569        6.20569
+  -4.60198       -4.60198
+  -1.28686       -1.28686
+  -0.383981      -0.383981
+  -0.173934      -0.173934
+  -7.97782       -7.97782
+   5.9926         5.9926
+  -3.7357        -3.7357
+  -7.77841       -7.77841
+   3.09245        3.09245
+  -3.70421       -3.70421
+  -1.50012       -1.50012
+  -3.90181       -3.90181
+   0.183002       0.183002
+  -4.72374       -4.72374
+  -3.36966       -3.36966
+   8.23642        8.23642
+   0.387898       0.387898
+  -2.53048       -2.53048
+   4.46348        4.46348
+  -0.932844      -0.932844
+  -1.76804       -1.76804
+  -0.390175      -0.390175
+   8.28101        8.28101
+   8.66959        8.66959
+   2.47585        2.47585
+   6.33837        6.33837
+   3.05846        3.05846
+   6.43047        6.43047
+   0.167477       0.167477
+   0.615034       0.615034
+  -8.467         -8.467
+   2.15566        2.15566
+   6.59172        6.59172
+  -8.30068       -8.30068
+  -2.92268       -2.92268
+  -1.14616       -1.14616
+   3.864          3.864
+  -8.07267       -8.07267
+   0.382952       0.382952
+   4.79087        4.79087
+   7.87692        7.87692
+  -1.27352       -1.27352
+  -0.439992      -0.439992
+  -0.361056      -0.361056
+   5.51463        5.51463
+   4.10827        4.10827
+  -1.36056       -1.36056
+ -10.9063       -10.9063
+  -3.12566       -3.12566
+  -1.52612       -1.52612
+   2.47429        2.47429
+   1.92973        1.92973
+   6.05399        6.05399
+   6.35717        6.35717
+  -6.54112       -6.54112
+   0.16752        0.16752
+  -0.581192      -0.581192
+  -3.91981       -3.91981
+   3.29046        3.29046
+  -9.85289       -9.85289
+  -1.68008       -1.68008
+  -0.294261      -0.294261
+  -2.33446       -2.33446
+   8.72203        8.72203
+  -7.53754       -7.53754
+   1.8548         1.8548
+   0.0863562      0.0863562
+   3.71224        3.71224
+  -2.72156       -2.72156
+   6.92717        6.92717
+   4.22066        4.22066
+   2.9384         2.9384
+  -0.436476      -0.436476
+   7.94505        7.94505
+   3.35167        3.35167
+   4.57606        4.57606
+  -1.94551       -1.94551
+   7.26891        7.26891
+   5.7114         5.7114
+  -4.8975        -4.8975
+   0.24802        0.24802
+   4.4272         4.4272
+   3.21714        3.21714
+  -2.75997       -2.75997
+   3.0239         3.0239
+   6.00743        6.00743
+   1.95157        1.95157
+  -8.23524       -8.23524
+  -0.0388194     -0.0388194
+  -1.59723       -1.59723
+ -15.7227       -15.7227
+   5.01363        5.01363
+   2.59661        2.59661
+   0.344503       0.344503
+   7.85727        7.85727
+   0.142462       0.142462
+  -3.54743       -3.54743
+  -4.18558       -4.18558
+   3.96172        3.96172
+  -0.376684      -0.376684
+   3.78763        3.78763
+  -1.58384       -1.58384
+  15.837         15.837
+  -0.887404      -0.887404
+   0.855016       0.855016
+  11.1701        11.1701
+   5.15206        5.15206
+   6.83176        6.83176
+  -0.91331       -0.91331
+ -10.3398       -10.3398
+   2.48231        2.48231
+  -2.03572       -2.03572
+   1.09096        1.09096
+  -0.162198      -0.162198
+  -7.32758       -7.32758
+  -6.97941       -6.97941
+   5.98831        5.98831
+  -7.43703       -7.43703
+  -8.97936       -8.97936
+   0.676949       0.676949
+   1.37291        1.37291
+   4.41159        4.41159
+   2.45643        2.45643
+   2.79374        2.79374
+   2.36712        2.36712
+  -7.74483       -7.74483
+   0.602922       0.602922
+  -2.48544       -2.48544
+   0.299035       0.299035
+   6.77695        6.77695
+   1.44763        1.44763
+   1.94637        1.94637
+  -4.04181       -4.04181
+  16.3509        16.3509
+   6.4273         6.4273
+   5.41235        5.41235
+  -5.91387       -5.91387
+  -6.06301       -6.06301
+   3.4536         3.4536
+  -3.39128       -3.39128
+  11.299         11.299
+   2.62685        2.62685
+   1.00866        1.00866
+  10.6766        10.6766
+  -0.805083      -0.805083
+   3.91073        3.91073
+   3.67201        3.67201
+  -9.14116       -9.14116
+  15.6406        15.6406
+   3.22084        3.22084
+  -2.90513       -2.90513
+   4.58966        4.58966
+   0.0983211      0.0983211
+   2.35908        2.35908
+   0.658109       0.658109
+   2.37478        2.37478
+  -6.70679       -6.70679
+   6.08307        6.08307
+ -29.6624       -29.6624
+   1.55578        1.55578
+   5.31311        5.31311
+  -5.40681       -5.40681
+   1.80228        1.80228
+   4.50431        4.50431
+   7.25673        7.25673
+   5.89811        5.89811
+  -2.92888       -2.92888
+   7.48853        7.48853
+  -1.67318       -1.67318
+   0.974302       0.974302
+  -8.10178       -8.10178
+   3.29435        3.29435
+  -1.64519       -1.64519
+  -7.08854       -7.08854
+   6.68891        6.68891
+  -5.69927       -5.69927
+  -3.51768       -3.51768
+  11.2895        11.2895
+  -0.828568      -0.828568
+   5.53562        5.53562
+  -0.358066      -0.358066
+  -5.92559       -5.92559
+   4.39224        4.39224
+  -5.1225        -5.1225
+  -9.51174       -9.51174
+   9.80076        9.80076
+  -1.85858       -1.85858
+   6.95181        6.95181
+  -1.71297       -1.71297
+  -0.275297      -0.275297
+  -0.860135      -0.860135
+  -0.484906      -0.484906
+   5.71425        5.71425
+   2.74639        2.74639
+  -8.40417       -8.40417
+  -1.84935       -1.84935
+   2.94526        2.94526
+  10.708         10.708
+   0.892511       0.892511
+  -1.36773       -1.36773
+  -7.25911       -7.25911
+   3.91428        3.91428
+  -0.776027      -0.776027
+   3.44102        3.44102
+  -4.87806       -4.87806
+   3.65101        3.65101
+  -3.01077       -3.01077
+   1.17918        1.17918
+   5.82266        5.82266
+   8.52564        8.52564
+   4.35296        4.35296
+  -2.94897       -2.94897
+  -4.19366       -4.19366
+  -4.7939        -4.7939
+   3.44038        3.44038
+  -7.87089       -7.87089
+  -3.18931       -3.18931
+  -6.65708       -6.65708
+   1.09687        1.09687
+  -4.36662       -4.36662
+   2.90783        2.90783
+   4.66889        4.66889
+  -1.26146       -1.26146
+  -2.01469       -2.01469
+  -2.44566       -2.44566
+  -2.15098       -2.15098
+   3.4006         3.4006
+   0.0396139      0.0396139
+   2.29469        2.29469
+  -7.62709       -7.62709
+   7.18738        7.18738
+   1.45481        1.45481
+   2.37791        2.37791
+  -5.37208       -5.37208
+  -0.0612415     -0.0612415
+  -1.46115       -1.46115
+   4.29624        4.29624
+   3.25993        3.25993
+   2.42986        2.42986
+   6.56133        6.56133
+  -2.07349       -2.07349
+   5.61643        5.61643
+   5.48251        5.48251
+  -0.703666      -0.703666
+  -5.09456       -5.09456
+   0.57249        0.57249
+   4.28577        4.28577
+   2.468          2.468
+ -10.013        -10.013
+  -3.26046       -3.26046
+  -7.91038       -7.91038
+  -2.03302       -2.03302
+   3.49234        3.49234
+  -1.2481        -1.2481
+  -1.87417       -1.87417
+  -1.93016       -1.93016
+   2.14307        2.14307
+  -9.0722        -9.0722
+   2.03124        2.03124
+  -0.938906      -0.938906
+   0.817464       0.817464
+   2.23636        2.23636
+   1.3076         1.3076
+   4.90629        4.90629
+   2.16603        2.16603
+   5.84398        5.84398
+  -6.56748       -6.56748
+   7.22968        7.22968
+   0.664381       0.664381
+  11.2001        11.2001
+  -4.98902       -4.98902
+   0.841822       0.841822
+  -1.35522       -1.35522
+  -2.43996       -2.43996
+   5.14732        5.14732
+  -7.50974       -7.50974
+   5.73113        5.73113
+  -2.72015       -2.72015
+  -5.04474       -5.04474
+ -13.1          -13.1
+   0.0777815      0.0777815
+   7.85631        7.85631
+  -0.323243      -0.323243
+  -2.97974       -2.97974
+   0.925187       0.925187
+   5.77219        5.77219
+   4.39868        4.39868
+   2.22326        2.22326
+   1.79052        1.79052
+  -3.37507       -3.37507
+  -4.08645       -4.08645
+   5.59349        5.59349
+  11.879         11.879
+  -0.8099        -0.8099
+  16.6866        16.6866
+   2.85772        2.85772
+   3.73902        3.73902
+  -0.406009      -0.406009
+   7.49033        7.49033
+  -1.01733       -1.01733
+   4.03678        4.03678
+   4.91574        4.91574
+  14.6191        14.6191
+  -1.18215       -1.18215
+  -2.79895       -2.79895
+  -5.16604       -5.16604
+  -2.24596       -2.24596
+   1.83945        1.83945
+   1.72673        1.72673
+ -23.2963       -23.2963
+  -0.623748      -0.623748
+  -2.8419        -2.8419
+   6.56374        6.56374
+  10.3431        10.3431
+   5.28302        5.28302
+   3.12716        3.12716
+   8.41242        8.41242
+   0.416003       0.416003
+  -2.43236       -2.43236
+  -1.63284       -1.63284
+   5.3806         5.3806
+   9.39975        9.39975
+   4.44496        4.44496
+  -3.01441       -3.01441
+  -1.33538       -1.33538
+   2.23541        2.23541
+  -4.30131       -4.30131
+  -1.20324       -1.20324
+   4.79406        4.79406
+   0.692551       0.692551
+  -2.20403       -2.20403
+   0.12931        0.12931
+   0.842875       0.842875
+   0.29791        0.29791
+   6.59639        6.59639
+   8.6591         8.6591
+   2.07311        2.07311
+  -6.48842       -6.48842
+   2.70007        2.70007
+  -0.143695      -0.143695
+   3.99651        3.99651
+   6.86089        6.86089
+  -2.54281       -2.54281
+  -5.085         -5.085
+   3.61747        3.61747
+   2.09466        2.09466
+   3.35667        3.35667
+   7.38405        7.38405
+   0.816999       0.816999
+  -0.564258      -0.564258
+   2.46281        2.46281
+  -0.081471      -0.081471
+  12.0933        12.0933
+   9.45364        9.45364
+   0.303564       0.303564
+  -2.20687       -2.20687
+   1.90101        1.90101
+  -2.65606       -2.65606
+ -11.3589       -11.3589
+  -1.68249       -1.68249
+  -1.25813       -1.25813
+  -0.96125       -0.96125
+  -2.84666       -2.84666
+   1.18914        1.18914
+   0.211945       0.211945
+  -4.8988        -4.8988
+   0.894798       0.894798
+   3.9685         3.9685
+  -0.852608      -0.852608
+   3.37537        3.37537
+  -0.847579      -0.847579
+  -4.37006       -4.37006
+  -4.12787       -4.12787
+   4.37155        4.37155
+  -7.86631       -7.86631
+  -3.59755       -3.59755
+  -2.55397       -2.55397
+   4.25921        4.25921
+   2.21721        2.21721
+   5.72299        5.72299
+   8.32362        8.32362
+  14.4057        14.4057
+   1.49376        1.49376
+   3.108          3.108
+  -1.34388       -1.34388
+   3.77816        3.77816
+   5.69761        5.69761
+   0.255491       0.255491
+   4.15979        4.15979
+ -14.6016       -14.6016
+   3.1475         3.1475
+   2.86732        2.86732
+  -2.7875        -2.7875
+  -8.78827       -8.78827
+  -1.38068       -1.38068
+  -2.74156       -2.74156
+  -4.82257       -4.82257
+  -4.64984       -4.64984
+  -0.462036      -0.462036
+   2.36274        2.36274
+   2.73927        2.73927
+  -4.01583       -4.01583
+  -4.20256       -4.20256
+   7.33455        7.33455
+   7.53557        7.53557
+   3.2532         3.2532
+  -0.556551      -0.556551
+   4.39618        4.39618
+   2.92025        2.92025
+ -49.4395       -49.4395
+   1.84066        1.84066
+  -6.03682       -6.03682
+   9.70956        9.70956
+  12.18          12.18
+  -0.134471      -0.134471
+   0.388477       0.388477
+  -4.30526       -4.30526
+   3.98614        3.98614
+  -3.20351       -3.20351
+   3.81764        3.81764
+   5.34853        5.34853
+   0.382215       0.382215
+  -0.473372      -0.473372
+  -4.4073        -4.4073
+ -10.1129       -10.1129
+  -6.82482       -6.82482
+   5.39935        5.39935
+  -0.664077      -0.664077
+   7.75577        7.75577
+  -5.565         -5.565
+  -2.28518       -2.28518
+  -3.09472       -3.09472
+   6.0196         6.0196
+  -1.32035       -1.32035
+   2.5721         2.5721
+  -9.0201        -9.0201
+   6.87621        6.87621
+   7.57662        7.57662
+  -2.42131       -2.42131
+  -7.11          -7.11
+   1.5457         1.5457
+   1.38686        1.38686
+  -1.67077       -1.67077
+   5.34357        5.34357
+  -5.22992       -5.22992
+  -5.50112       -5.50112
+  -0.820436      -0.820436
+  -6.85987       -6.85987
+   4.36935        4.36935
+   8.27737        8.27737
+   7.16613        7.16613
+   7.21538        7.21538
+   0.0297893      0.0297893
+  -3.30991       -3.30991
+   1.18508        1.18508
+  -0.745072      -0.745072
+  -1.31153       -1.31153
+  -2.57184       -2.57184
+  -0.187369      -0.187369
+   6.79233        6.79233
+   8.04294        8.04294
+   3.06986        3.06986
+  -5.13761       -5.13761
+   0.539648       0.539648
+   5.02007        5.02007
+   2.67737        2.67737
+  -6.69984       -6.69984
+   6.76321        6.76321
+   6.25102        6.25102
+   3.80545        3.80545
+  -2.16059       -2.16059
+   2.81803        2.81803
+   0.447194       0.447194
+   1.84756        1.84756
+  -6.42528       -6.42528
+  -2.23379       -2.23379
+  -2.61151       -2.61151
+  -2.86143       -2.86143
+  -2.94039       -2.94039
+  -3.38503       -3.38503
+   0.474985       0.474985
+  -9.66389       -9.66389
+   4.96293        4.96293
+  -5.6718        -5.6718
+   7.06422        7.06422
+  -8.36354       -8.36354
+   0.0182466      0.0182466
+   9.20883        9.20883
+   8.23981        8.23981
+  -1.41968       -1.41968
+  -1.36057       -1.36057
+  -3.99568       -3.99568
+   2.51484        2.51484
+   5.41846        5.41846
+ -10.8511       -10.8511
+  -8.41267       -8.41267
+   2.04668        2.04668
+  -5.61525       -5.61525
+  -9.73507       -9.73507
+  -0.497102      -0.497102
+   4.29467        4.29467
+  -1.61424       -1.61424
+  -0.818494      -0.818494
+  -7.02135       -7.02135
+  13.4836        13.4836
+  -4.10115       -4.10115
+  -8.11914       -8.11914
+  -2.79895       -2.79895
+  -4.39428       -4.39428
+  -0.737467      -0.737467
+   1.37013        1.37013
+   9.56244        9.56244
+   2.92491        2.92491
+  -7.13393       -7.13393
+  -0.179291      -0.179291
+  -6.00313       -6.00313
+   7.27104        7.27104
+  -1.7103        -1.7103
+  -7.84843       -7.84843
+  13.7304        13.7304
+   2.40973        2.40973
+  -7.07755       -7.07755
+   1.31745        1.31745
+  -9.99271       -9.99271
+ -15.4753       -15.4753
+   4.38711        4.38711
+  -5.41127       -5.41127
+  -1.06491       -1.06491
+   1.09245        1.09245
+  -1.33961       -1.33961
+  -4.42681       -4.42681
+  -4.44164       -4.44164
+  -1.80772       -1.80772
+  -5.06035       -5.06035
+   0.197369       0.197369
+   7.27798        7.27798
+  -6.88382       -6.88382
+   3.21319        3.21319
+   8.04111        8.04111
+  -3.94107       -3.94107
+   1.79716        1.79716
+  -0.2134        -0.2134
+   1.36955        1.36955
+  13.7009        13.7009
+  -7.3497        -7.3497
+   1.80078        1.80078
+   4.25352        4.25352
+  -2.80092       -2.80092
+  -3.81295       -3.81295
+  -4.92036       -4.92036
+   0.856001       0.856001
+  -1.26696       -1.26696
+   2.65207        2.65207
+  -1.01876       -1.01876
+   1.50837        1.50837
+ -11.5335       -11.5335
+   5.80989        5.80989
+   2.45606        2.45606
+   1.64394        1.64394
+   2.73651        2.73651
+ -11.1653       -11.1653
+  -1.66359       -1.66359
+  -0.0317267     -0.0317267
+   0.115458       0.115458
+   4.43585        4.43585
+   1.24902        1.24902
+   7.30894        7.30894
+  16.7814        16.7814
+  -0.456154      -0.456154
+  -3.94033       -3.94033
+  -4.4947        -4.4947
+  -2.52048       -2.52048
+   0.0890704      0.0890704
+  -4.66338       -4.66338
+   3.88142        3.88142
+   2.35984        2.35984
+   4.84037        4.84037
+   6.95444        6.95444
+   2.74408        2.74408
+  -3.23958       -3.23958
+  -0.467292      -0.467292
+   6.26367        6.26367
+  -1.50588       -1.50588
+   4.13389        4.13389
+  -2.53819       -2.53819
+  -4.4987        -4.4987
+ -10.3487       -10.3487
+ -14.8297       -14.8297
+  -8.48112       -8.48112
+   3.95155        3.95155
+   1.2289         1.2289
+  -4.38025       -4.38025
+  -0.61687       -0.61687
+  10.8511        10.8511
+   1.15556        1.15556
+  -2.19768       -2.19768
+  -7.66931       -7.66931
+   4.72919        4.72919
+  -7.6738        -7.6738
+  -0.688528      -0.688528
+   4.74928        4.74928
+   4.92126        4.92126
+   0.897546       0.897546
+   3.85735        3.85735
+   0.201364       0.201364
+  -5.62425       -5.62425
+  -3.83117       -3.83117
+   4.05866        4.05866
+   3.10063        3.10063
+   2.5224         2.5224
+  -1.51274       -1.51274
+  -0.683338      -0.683338
+  -3.23147       -3.23147
+  -4.21268       -4.21268
+  -2.21401       -2.21401
+   1.57887        1.57887
+   0.848257       0.848257
+  -5.83704       -5.83704
+  -7.00011       -7.00011
+   3.16884        3.16884
+  -4.44161       -4.44161
+  -7.62482       -7.62482
+  -0.266943      -0.266943
+   0.41761        0.41761
+  -7.45144       -7.45144
+  -0.211132      -0.211132
+   0.276707       0.276707
+  16.7781        16.7781
+   0.689757       0.689757
+  -3.04049       -3.04049
+   2.91684        2.91684
+   1.97161        1.97161
+   3.7721         3.7721
+  -1.60698       -1.60698
+  -4.18868       -4.18868
+   7.66491        7.66491
+  -0.64664       -0.64664
+  -0.660623      -0.660623
+   8.68174        8.68174
+   0.282074       0.282074
+  -2.85266       -2.85266
+  -1.91293       -1.91293
+   7.18736        7.18736
+ -10.3875       -10.3875
+  -1.91603       -1.91603
+   6.29739        6.29739
+  -0.0375388     -0.0375388
+  -1.60576       -1.60576
+  -3.22148       -3.22148
+  -4.24549       -4.24549
+   1.30822        1.30822
+   2.52307        2.52307
+   0.403345       0.403345
+  -0.744478      -0.744478
+   2.41241        2.41241
+  -4.58098       -4.58098
+  -0.791842      -0.791842
+   3.73626        3.73626
+  -1.43002       -1.43002
+   4.30716        4.30716
+   3.30255        3.30255
+  -4.08011       -4.08011
+  -5.07282       -5.07282
+  -1.54759       -1.54759
+  -2.2305        -2.2305
+   6.8791         6.8791
+   9.7396         9.7396
+  -6.50395       -6.50395
+   3.57178        3.57178
+   7.08987        7.08987
+   6.2669         6.2669
+   5.87329        5.87329
+   2.36823        2.36823
+  -6.16          -6.16
+   1.96238        1.96238
+   7.31651        7.31651
+  -1.5257        -1.5257
+  -2.89061       -2.89061
+   0.407546       0.407546
+   5.10645        5.10645
+  11.0716        11.0716
+   4.7443         4.7443
+  -8.77353       -8.77353
+  -0.631177      -0.631177
+  -4.36973       -4.36973
+   1.48666        1.48666
+   7.7678         7.7678
+  -2.65407       -2.65407
+   4.56869        4.56869
+  -0.541163      -0.541163
+   2.89543        2.89543
+   5.39424        5.39424
+  -3.62954       -3.62954
+   3.77547        3.77547
+  -5.96886       -5.96886
+  -4.38947       -4.38947
+  -2.96756       -2.96756
+   2.28222        2.28222
+  -1.08489       -1.08489
+   1.74726        1.74726
+  -3.46088       -3.46088
+  11.9371        11.9371
+  -5.02359       -5.02359
+   2.51632        2.51632
+  -0.0297022     -0.0297022
+  -2.60011       -2.60011
+   0.254202       0.254202
+   9.7949         9.7949
+   3.64937        3.64937
+  10.0857        10.0857
+  -5.36637       -5.36637
+   4.11127        4.11127
+   8.90571        8.90571
+  -5.97219       -5.97219
+  -7.21379       -7.21379
+  -5.01561       -5.01561
+   2.98616        2.98616
+   1.99064        1.99064
+   0.16465        0.16465
+  -4.07902       -4.07902
+   4.34018        4.34018
+  -2.13528       -2.13528
+   2.39903        2.39903
+   4.00804        4.00804
+  -1.85741       -1.85741
+  -7.73083       -7.73083
+  -4.21139       -4.21139
+   4.65743        4.65743
+   0.963549       0.963549
+   0.29506        0.29506
+   6.05798        6.05798
+  12.4428        12.4428
+  -0.398651      -0.398651
+  -0.584559      -0.584559
+   2.75445        2.75445
+  -0.207975      -0.207975
+   6.11926        6.11926
+  -8.66125       -8.66125
+   3.07568        3.07568
+  -3.19358       -3.19358
+  -2.53024       -2.53024
+  14.1187        14.1187
+  -0.412049      -0.412049
+  12.5809        12.5809
+   6.26236        6.26236
+   5.23037        5.23037
+  -0.11356       -0.11356
+  -6.62321       -6.62321
+  -1.29651       -1.29651
+  -1.48734       -1.48734
+  13.0753        13.0753
+   4.21767        4.21767
+  -2.4425        -2.4425
+  -0.0901323     -0.0901323
+   9.79684        9.79684
+   4.74522        4.74522
+  -3.34804       -3.34804
+   7.37816        7.37816
+   2.57938        2.57938
+   1.92968        1.92968
+   3.75166        3.75166
+   5.0617         5.0617
+   8.74324        8.74324
+  -0.93703       -0.93703
+  -1.36031       -1.36031
+  -2.5439        -2.5439
+   1.56784        1.56784
+   2.56237        2.56237
+  -1.02578       -1.02578
+   6.62085        6.62085
+   7.69745        7.69745
+   6.26864        6.26864
+  -4.20046       -4.20046
+  -2.30926       -2.30926
+   2.74598        2.74598
+   4.11078        4.11078
+   2.8455         2.8455
+  -3.45407       -3.45407
+   2.82327        2.82327
+  -1.00356       -1.00356
+   8.85974        8.85974
+   6.35864        6.35864
+  -1.59146       -1.59146
+  -0.361996      -0.361996
+  -1.25198       -1.25198
+   8.2867         8.2867
+   0.981644       0.981644
+   2.68003        2.68003
+   1.10236        1.10236
+  -1.63423       -1.63423
+  -2.79552       -2.79552
+  -6.5718        -6.5718
+  -0.257779      -0.257779
+  -4.49325       -4.49325
+   5.0455         5.0455
+  14.4508        14.4508
+   3.60407        3.60407
+   3.09003        3.09003
+  -8.32962       -8.32962
+  -1.41178       -1.41178
+  12.5777        12.5777
+  -2.01342       -2.01342
+  -1.48205       -1.48205
+   0.967158       0.967158
+  -0.532548      -0.532548
+  -5.23274       -5.23274
+  -1.49702       -1.49702
+   0.739607       0.739607
+   3.49171        3.49171
+  -1.0507        -1.0507
+  -7.48299       -7.48299
+   7.57395        7.57395
+  -3.04813       -3.04813
+  16.322         16.322
+   7.81441        7.81441
+  -3.41529       -3.41529
+   2.05401        2.05401
+   1.08232        1.08232
+  12.5735        12.5735
+   0.126572       0.126572
+  -6.92158       -6.92158
+  -1.4651        -1.4651
+  -3.19425       -3.19425
+  -1.44093       -1.44093
+  -3.82056       -3.82056
+   6.72914        6.72914
+  -5.46583       -5.46583
+  -1.43396       -1.43396
+   7.42164        7.42164
+   1.00438        1.00438
+  -0.41415       -0.41415
+  -2.54987       -2.54987
+   6.88491        6.88491
+   3.84807        3.84807
+  -5.62245       -5.62245
+   5.24133        5.24133
+   7.99514        7.99514
+  -2.51593       -2.51593
+   8.19568        8.19568
+   0.854985       0.854985
+  -6.20478       -6.20478
+  -2.58235       -2.58235
+  -6.51346       -6.51346
+  12.8877        12.8877
+   8.6194         8.6194
+  -6.82669       -6.82669
+  -4.67379       -4.67379
+   8.13137        8.13137
+   0.733511       0.733511
+   5.66079        5.66079
+  -2.94337       -2.94337
+  -3.29462       -3.29462
+  -6.3809        -6.3809
+  -1.85613       -1.85613
+   0.635069       0.635069
+   0.432626       0.432626
+ -14.6426       -14.6426
+   8.05825        8.05825
+   6.50637        6.50637
+   1.44014        1.44014
+  -4.60602       -4.60602
+  -6.49137       -6.49137
+   6.33163        6.33163
+  -1.97616       -1.97616
+   0.573379       0.573379
+  -2.78039       -2.78039
+  -0.140087      -0.140087
+   1.52619        1.52619
+   6.83379        6.83379
+  -0.197981      -0.197981
+  -3.00849       -3.00849
+  -2.09725       -2.09725
+  -2.06883       -2.06883
+  -0.328198      -0.328198
+  -0.212338      -0.212338
+   5.4425         5.4425
+   6.48574        6.48574
+   2.00073        2.00073
+  -3.15642       -3.15642
+  -0.0673389     -0.0673389
+  -4.19911       -4.19911
+   4.5466         4.5466
+   3.73221        3.73221
+  -1.01059       -1.01059
+  -4.29015       -4.29015
+   4.9909         4.9909
+   3.22397        3.22397
+  -1.27984       -1.27984
+   2.83358        2.83358
+   2.25695        2.25695
+   7.2879         7.2879
+  -1.47955       -1.47955
+  12.7627        12.7627
+  -3.72449       -3.72449
+   3.97719        3.97719
+  14.2197        14.2197
+  -1.24031       -1.24031
+  -7.41824       -7.41824
+   1.90207        1.90207
+   1.10939        1.10939
+  -7.47202       -7.47202
+   3.85738        3.85738
+  -4.12085       -4.12085
+   1.12097        1.12097
+  -0.545646      -0.545646
+   3.04129        3.04129
+   1.05043        1.05043
+   0.993448       0.993448
+  -5.78424       -5.78424
+  -1.97199       -1.97199
+  -5.74806       -5.74806
+   2.70835        2.70835
+  -8.09729       -8.09729
+  -6.36035       -6.36035
+  -1.24361       -1.24361
+  -2.44813       -2.44813
+   7.48353        7.48353
+   2.0202         2.0202
+   3.04366        3.04366
+  -3.98778       -3.98778
+   4.80106        4.80106
+   0.926552       0.926552
+   3.35253        3.35253
+  -4.10577       -4.10577
+  -3.57853       -3.57853
+   4.03372        4.03372
+  -2.38792       -2.38792
+   0.12177        0.12177
+  -0.761671      -0.761671
+  -4.25652       -4.25652
+   7.27933        7.27933
+   0.165182       0.165182
+   1.34367        1.34367
+  -7.36923       -7.36923
+   2.38548        2.38548
+   0.117217       0.117217
+   2.02002        2.02002
+  -4.60023       -4.60023
+   2.78           2.78
+  -1.34604       -1.34604
+   4.7234         4.7234
+   7.37673        7.37673
+   2.07986        2.07986
+  -5.72573       -5.72573
+  -6.66143       -6.66143
+   2.43072        2.43072
+   1.34782        1.34782
+  -0.114238      -0.114238
+   2.32103        2.32103
+   1.84042        1.84042
+   1.07005        1.07005
+   3.88182        3.88182
+  -0.752264      -0.752264
+  -2.43517       -2.43517
+  -5.29216       -5.29216
+  -0.13527       -0.13527
+   1.40188        1.40188
+  -5.87815       -5.87815
+  -1.90167       -1.90167
+   2.88562        2.88562
+  -2.29028       -2.29028
+   2.35477        2.35477
+  -3.50731       -3.50731
+   6.0621         6.0621
+   3.2011         3.2011
+   2.19115        2.19115
+  -3.03557       -3.03557
+  -8.49394       -8.49394
+   0.936501       0.936501
+   7.19188        7.19188
+   4.50162        4.50162
+   0.341394       0.341394
+   2.54484        2.54484
+   1.67305        1.67305
+   3.05008        3.05008
+  -2.0266        -2.0266
+   7.28431        7.28431
+  -7.70924       -7.70924
+   2.60851        2.60851
+   6.8054         6.8054
+   1.8878         1.8878
+   1.87624        1.87624
+  -5.13611       -5.13611
+  -3.23698       -3.23698
+   4.03201        4.03201
+  -5.27165       -5.27165
+  -4.95817       -4.95817
+  -0.200461      -0.200461
+   4.27259        4.27259
+   0.449661       0.449661
+   7.49752        7.49752
+  -5.47923       -5.47923
+  -2.40934       -2.40934
+  25.0066        25.0066
+  -3.14511       -3.14511
+  -1.62587       -1.62587
+  -1.67652       -1.67652
+  -2.17888       -2.17888
+   2.37296        2.37296
+  -4.41408       -4.41408
+   0.65204        0.65204
+  10.849         10.849
+  -2.3021        -2.3021
+   2.20417        2.20417
+  10.0579        10.0579
+  -4.03489       -4.03489
+   7.60982        7.60982
+  -5.74951       -5.74951
+  -2.97582       -2.97582
+  -8.61382       -8.61382
+  -1.90903       -1.90903
+  -3.64556       -3.64556
+ -16.2304       -16.2304
+ -15.9793       -15.9793
+  -4.59448       -4.59448
+  -2.67688       -2.67688
+  -1.67148       -1.67148
+   5.57026        5.57026
+   0.846445       0.846445
+  -7.54149       -7.54149
+  -3.61401       -3.61401
+   4.03723        4.03723
+   0.711821       0.711821
+   8.99009        8.99009
+  -6.15866       -6.15866
+  -1.36865       -1.36865
+  -4.31058       -4.31058
+   6.31659        6.31659
+  -6.23773       -6.23773
+   0.857388       0.857388
+   3.6152         3.6152
+  -1.28774       -1.28774
+  -4.92094       -4.92094
+   3.08527        3.08527
+  -5.74582       -5.74582
+  -4.20897       -4.20897
+  -5.19406       -5.19406
+  -4.06851       -4.06851
+   5.73867        5.73867
+   3.32767        3.32767
+ -11.2588       -11.2588
+  -7.94126       -7.94126
+   5.38746        5.38746
+  -0.0253579     -0.0253579
+  -1.7856        -1.7856
+  -1.31209       -1.31209
+   6.85519        6.85519
+   2.71496        2.71496
+  -2.58838       -2.58838
+  -6.86996       -6.86996
+   1.01204        1.01204
+   3.43433        3.43433
+  -0.249192      -0.249192
+   7.96322        7.96322
+  14.3414        14.3414
+   2.44774        2.44774
+   4.73731        4.73731
+  -9.14288       -9.14288
+   2.70325        2.70325
+   6.48202        6.48202
+  -2.58391       -2.58391
+  -4.52079       -4.52079
+  -0.64105       -0.64105
+  -3.75531       -3.75531
+  -3.93321       -3.93321
+  -2.5879        -2.5879
+   2.34697        2.34697
+  -3.89721       -3.89721
+  -1.60712       -1.60712
+  -7.49452       -7.49452
+  -0.518596      -0.518596
+   0.996693       0.996693
+   2.83468        2.83468
+  -6.19363       -6.19363
+  -7.25683       -7.25683
+   0.391546       0.391546
+  -7.52756       -7.52756
+  -0.810817      -0.810817
+  -2.64942       -2.64942
+  -2.95081       -2.95081
+  -6.34989       -6.34989
+   3.9961         3.9961
+   1.36755        1.36755
+  -0.335808      -0.335808
+ -11.7919       -11.7919
+   1.16904        1.16904
+   6.26031        6.26031
+  -4.68064       -4.68064
+   5.55008        5.55008
+   3.65873        3.65873
+  -3.95177       -3.95177
+   7.62708        7.62708
+  -2.4932        -2.4932
+  -0.713266      -0.713266
+   6.76214        6.76214
+  -0.802523      -0.802523
+  -0.327543      -0.327543
+  -6.9053        -6.9053
+  -2.69604       -2.69604
+   9.729          9.729
+  -7.61691       -7.61691
+  -0.658653      -0.658653
+   1.62531        1.62531
+   0.532107       0.532107
+   1.71729        1.71729
+ -10.1795       -10.1795
+   5.54208        5.54208
+   4.02502        4.02502
+  -1.47596       -1.47596
+  11.818         11.818
+   4.40414        4.40414
+   5.64827        5.64827
+   5.89386        5.89386
+  -6.19187       -6.19187
+   4.77889        4.77889
+  -0.261731      -0.261731
+  -0.570525      -0.570525
+   3.80941        3.80941
+  -3.95414       -3.95414
+   0.642971       0.642971
+  -7.23493       -7.23493
+   0.744423       0.744423
+  11.5682        11.5682
+  -3.17145       -3.17145
+   9.02877        9.02877
+  10.5452        10.5452
+  -7.05642       -7.05642
+  -6.01952       -6.01952
+  -5.61355       -5.61355
+   1.28759        1.28759
+   3.44186        3.44186
+  -2.52363       -2.52363
+   8.95712        8.95712
+  -1.33999       -1.33999
+  -3.25858       -3.25858
+   2.33509        2.33509
+   2.16314        2.16314
+  14.4002        14.4002
+  -5.22345       -5.22345
+  -5.6232        -5.6232
+  -4.20801       -4.20801
+   0.677359       0.677359
+   1.92688        1.92688
+   2.4265         2.4265
+  -3.47901       -3.47901
+  -3.35004       -3.35004
+  -5.32445       -5.32445
+   0.817822       0.817822
+   5.9241         5.9241
+   2.13342        2.13342
+   9.30726        9.30726
+  -6.00328       -6.00328
+   5.10125        5.10125
+  16.6941        16.6941
+  -1.41774       -1.41774
+   0.843709       0.843709
+   3.71326        3.71326
+ -12.7315       -12.7315
+  -1.58947       -1.58947
+   2.7713         2.7713
+  -5.89993       -5.89993
+ -10.1427       -10.1427
+  -1.60823       -1.60823
+  -4.98621       -4.98621
+ -10.6258       -10.6258
+   0.255858       0.255858
+   5.87781        5.87781
+   0.549239       0.549239
+  -0.361649      -0.361649
+   2.89543        2.89543
+  -1.56252       -1.56252
+  -7.04269       -7.04269
+   0.360599       0.360599
+  -0.80318       -0.80318
+  -8.15537       -8.15537
+   7.86106        7.86106
+   4.25906        4.25906
+   1.78474        1.78474
+   4.15764        4.15764
+  -1.8884        -1.8884
+  -7.16959       -7.16959
+   2.84539        2.84539
+  -3.33161       -3.33161
+   4.89863        4.89863
+  -3.36503       -3.36503
+  -4.68013       -4.68013
+   5.18058        5.18058
+  -9.69276       -9.69276
+  -1.56116       -1.56116
+  -3.58275       -3.58275
+  -2.73766       -2.73766
+   6.64492        6.64492
+  -3.78966       -3.78966
+   2.63467        2.63467
+ -12.4868       -12.4868
+  -3.4241        -3.4241
+   3.2898         3.2898
+   2.20265        2.20265
+  -1.36672       -1.36672
+   2.71448        2.71448
+   5.87839        5.87839
+   0.160837       0.160837
+  -2.64458       -2.64458
+  -3.8078        -3.8078
+   5.08743        5.08743
+ -14.014        -14.014
+   4.44746        4.44746
+   6.61584        6.61584
+  -0.916513      -0.916513
+  -8.08277       -8.08277
+  -8.088         -8.088
+  -5.14152       -5.14152
+  -4.30739       -4.30739
+  -8.76727       -8.76727
+  -4.53313       -4.53313
+  11.0356        11.0356
+  -2.37348       -2.37348
+  -8.71711       -8.71711
+  -2.22971       -2.22971
+   8.19346        8.19346
+  -0.330962      -0.330962
+   1.10067        1.10067
+   1.01878        1.01878
+ -10.2666       -10.2666
+   8.15909        8.15909
+   9.09316        9.09316
+  -0.862864      -0.862864
+  -7.54443       -7.54443
+  -3.44703       -3.44703
+   5.21819        5.21819
+  -2.06834       -2.06834
+   9.55442        9.55442
+  -1.89649       -1.89649
+  -5.57892       -5.57892
+   4.22421        4.22421
+  -4.06375       -4.06375
+   3.81452        3.81452
+   3.09071        3.09071
+  -7.34297       -7.34297
+  -1.67899       -1.67899
+   0.58489        0.58489
+  -5.33824       -5.33824
+   2.82705        2.82705
+  -3.70864       -3.70864
+   4.21641        4.21641
+   3.82508        3.82508
+  -4.04356       -4.04356
+  20.0249        20.0249
+ -13.1531       -13.1531
+   2.98603        2.98603
+   5.54713        5.54713
+  -1.39722       -1.39722
+   2.13016        2.13016
+  -2.40215       -2.40215
+   0.168123       0.168123
+   2.77021        2.77021
+  -2.32327       -2.32327
+  -1.06731       -1.06731
+   2.53877        2.53877
+  -1.94325       -1.94325
+   1.47106        1.47106
+   0.294436       0.294436
+  -0.547055      -0.547055
+   0.116016       0.116016
+   1.56148        1.56148
+   3.21789        3.21789
+  -2.89007       -2.89007
+  -4.33765       -4.33765
+   0.566163       0.566163
+   0.402729       0.402729
+  -7.80674       -7.80674
+   4.72058        4.72058
+   3.97584        3.97584
+   1.91646        1.91646
+   2.09298        2.09298
+   1.88552        1.88552
+  -2.37581       -2.37581
+ -18.2615       -18.2615
+   2.68651        2.68651
+   5.5            5.5
+   0.355051       0.355051
+   5.6052         5.6052
+   7.74854        7.74854
+  -0.512378      -0.512378
+   1.60299        1.60299
+  -5.49563       -5.49563
+  -1.96455       -1.96455
+ -16.3228       -16.3228
+  -6.87737       -6.87737
+  -4.60755       -4.60755
+  -1.32116       -1.32116
+   2.87263        2.87263
+  -2.09541       -2.09541
+   3.43595        3.43595
+   3.63528        3.63528
+   3.52056        3.52056
+  -3.59484       -3.59484
+   1.03764        1.03764
+  -7.14947       -7.14947
+  -5.80634       -5.80634
+   4.71397        4.71397
+   0.720588       0.720588
+  -2.24074       -2.24074
+   5.82418        5.82418
+  -3.22013       -3.22013
+   3.68858        3.68858
+  -1.43166       -1.43166
+   4.47978        4.47978
+  -4.83356       -4.83356
+  -3.96257       -3.96257
+  -5.95512       -5.95512
+   0.496691       0.496691
+  -7.58825       -7.58825
+  -6.47331       -6.47331
+  -1.14446       -1.14446
+   3.91615        3.91615
+  -0.588841      -0.588841
+   6.56683        6.56683
+   3.97252        3.97252
+  -4.3126        -4.3126
+  -8.20913       -8.20913
+   0.310182       0.310182
+  -7.3006        -7.3006
+   7.92805        7.92805
+   2.1756         2.1756
+   1.06404        1.06404
+   1.14471        1.14471
+  -1.50242       -1.50242
+   0.00723557     0.00723557
+   5.76841        5.76841
+  -1.96707       -1.96707
+   8.87243        8.87243
+  -3.23281       -3.23281
+  12.3087        12.3087
+   3.3245         3.3245
+   3.00334        3.00334
+  -5.74048       -5.74048
+   7.43939        7.43939
+  -0.906001      -0.906001
+   2.24067        2.24067
+  -6.23989       -6.23989
+   2.81483        2.81483
+  -1.62648       -1.62648
+  -7.26368       -7.26368
+   1.69171        1.69171
+ -11.2631       -11.2631
+  -2.32992       -2.32992
+  -6.07361       -6.07361
+  -7.56822       -7.56822
+  -7.56737       -7.56737
+   5.97037        5.97037
+   6.74398        6.74398
+  -2.24599       -2.24599
+   2.95213        2.95213
+ -12.7864       -12.7864
+   0.680035       0.680035
+  -1.39988       -1.39988
+  -4.74028       -4.74028
+   3.01887        3.01887
+   1.89636        1.89636
+   4.46014        4.46014
+  -4.38308       -4.38308
+  11.7633        11.7633
+  -3.54671       -3.54671
+  -3.47584       -3.47584
+   3.80037        3.80037
+   7.77849        7.77849
+  -7.00006       -7.00006
+  -4.87665       -4.87665
+  -4.54736       -4.54736
+  -7.81752       -7.81752
+  -0.0654465     -0.0654465
+  -3.70587       -3.70587
+  -2.24231       -2.24231
+   5.58005        5.58005
+  -3.09415       -3.09415
+  -5.55063       -5.55063
+  -4.19666       -4.19666
+  -6.83328       -6.83328
+  -6.9216        -6.9216
+  -3.72782       -3.72782
+  -2.18574       -2.18574
+   1.28076        1.28076
+  -3.40691       -3.40691
+   0.486964       0.486964
+  -2.11025       -2.11025
+  -1.42349       -1.42349
+   6.06854        6.06854
+  -1.37534       -1.37534
+   9.47832        9.47832
+  -0.567045      -0.567045
+  -6.98328       -6.98328
+   6.73139        6.73139
+  -1.56812       -1.56812
+   0.141683       0.141683
+   1.78697        1.78697
+  -2.03874       -2.03874
+   1.28356        1.28356
+   6.9912         6.9912
+  -3.8858        -3.8858
+  -1.38808       -1.38808
+  -2.16632       -2.16632
+   3.57955        3.57955
+   2.73506        2.73506
+  -3.03108       -3.03108
+  -3.44677       -3.44677
+   1.37111        1.37111
+ -10.0008       -10.0008
+  -3.61651       -3.61651
+   1.97313        1.97313
+   2.11298        2.11298
+   0.174957       0.174957
+  -0.131546      -0.131546
+   7.58484        7.58484
+   4.27907        4.27907
+   0.855439       0.855439
+   4.44153        4.44153
+  -1.04577       -1.04577
+  -7.49625       -7.49625
+   2.1572         2.1572
+  13.0815        13.0815
+   4.57025        4.57025
+   0.704658       0.704658
+   3.25079        3.25079
+  -0.682139      -0.682139
+  -4.17209       -4.17209
+  -1.38547       -1.38547
+   5.52688        5.52688
+  -4.90717       -4.90717
+   2.56402        2.56402
+  -1.37164       -1.37164
+  -6.05044       -6.05044
+   8.3158         8.3158
+  -0.640461      -0.640461
+  -2.40145       -2.40145
+  -1.02959       -1.02959
+  -6.75028       -6.75028
+   4.20206        4.20206
+   0.615412       0.615412
+  -0.389435      -0.389435
+  -5.07439       -5.07439
+  -5.34136       -5.34136
+  -1.88522       -1.88522
+  -4.82628       -4.82628
+   0.54435        0.54435
+  -3.28948       -3.28948
+   5.0051         5.0051
+  -8.5501        -8.5501
+   7.31448        7.31448
+   0.145651       0.145651
+   3.28586        3.28586
+  -1.8624        -1.8624
+  -8.9235        -8.9235
+   3.15894        3.15894
+  -9.9459        -9.9459
+   0.517233       0.517233
+  -4.59899       -4.59899
+   0.641116       0.641116
+  10.3809        10.3809
+   2.39935        2.39935
+  -0.378496      -0.378496
+   0.680329       0.680329
+   2.35584        2.35584
+  -2.24714       -2.24714
+  -4.8742        -4.8742
+  -3.96429       -3.96429
+   1.29263        1.29263
+   0.618875       0.618875
+  -0.611961      -0.611961
+   1.06612        1.06612
+  -3.39289       -3.39289
+  -0.226022      -0.226022
+   4.24418        4.24418
+   0.884239       0.884239
+   8.25747        8.25747
+  -3.23019       -3.23019
+  -9.99374       -9.99374
+   8.54414        8.54414
+  -6.06374       -6.06374
+  -4.92601       -4.92601
+   7.22101        7.22101
+  11.5756        11.5756
+  13.436         13.436
+   4.13522        4.13522
+   9.67412        9.67412
+  -3.13805       -3.13805
+   7.50856        7.50856
+  -7.98069       -7.98069
+   4.92059        4.92059
+  -6.72969       -6.72969
+  -4.48762       -4.48762
+  -3.60328       -3.60328
+  -1.75053       -1.75053
+   1.5638         1.5638
+   4.74213        4.74213
+   5.16046        5.16046
+  -1.9857        -1.9857
+  -6.34885       -6.34885
+  -3.58963       -3.58963
+   4.96795        4.96795
+   1.44405        1.44405
+  -2.74682       -2.74682
+  -0.545296      -0.545296
+ -10.7507       -10.7507
+  -0.117477      -0.117477
+  -0.436907      -0.436907
+  -1.11656       -1.11656
+   1.64789        1.64789
+  -4.08799       -4.08799
+  -1.04262       -1.04262
+   6.06007        6.06007
+  -6.68208       -6.68208
+   6.81976        6.81976
+  -6.89836       -6.89836
+  -0.555115      -0.555115
+  -2.85307       -2.85307
+  -7.76567       -7.76567
+  -5.65104       -5.65104
+   8.93521        8.93521
+  -5.0663        -5.0663
+   2.52214        2.52214
+   0.382824       0.382824
+  -0.398468      -0.398468
+   5.05183        5.05183
+   4.134          4.134
+   1.42909        1.42909
+   2.99357        2.99357
+  10.7821        10.7821
+  -4.54764       -4.54764
+  -0.0440308     -0.0440308
+   0.647161       0.647161
+   3.27569        3.27569
+ -32.9478       -32.9478
+   6.92399        6.92399
+  -3.05953       -3.05953
+  -2.29742       -2.29742
+  -0.41863       -0.41863
+   2.99125        2.99125
+   3.40805        3.40805
+  -1.36651       -1.36651
+  -3.25561       -3.25561
+   5.11504        5.11504
+  -0.532291      -0.532291
+   9.93341        9.93341
+  -2.2806        -2.2806
+  10.9617        10.9617
+  -2.53642       -2.53642
+   0.995763       0.995763
+  -1.28898       -1.28898
+  -2.99921       -2.99921
+  -2.46773       -2.46773
+ -11.0849       -11.0849
+ -11.64         -11.64
+  -3.73617       -3.73617
+   2.74223        2.74223
+  -0.976817      -0.976817
+  -0.384814      -0.384814
+  -3.38815       -3.38815
+   2.27591        2.27591
+  -5.25732       -5.25732
+  -1.65764       -1.65764
+  -5.8501        -5.8501
+  -4.85863       -4.85863
+   2.78987        2.78987
+   5.3324         5.3324
+  -9.16758       -9.16758
+   7.90047        7.90047
+   5.68696        5.68696
+   7.2668         7.2668
+  -0.857072      -0.857072
+   0.0834347      0.0834347
+   1.11833        1.11833
+   0.88212        0.88212
+  -4.40785       -4.40785
+   5.25846        5.25846
+   7.46283        7.46283
+   6.26981        6.26981
+ -10.8935       -10.8935
+  -0.226332      -0.226332
+  -1.64568       -1.64568
+  -0.389003      -0.389003
+  -0.854872      -0.854872
+  -3.38063       -3.38063
+  -4.74874       -4.74874
+  -1.81717       -1.81717
+  -6.03338       -6.03338
+   9.41153        9.41153
+  -2.75636       -2.75636
+  -4.03638       -4.03638
+  -2.82527       -2.82527
+   0.641039       0.641039
+  -3.08939       -3.08939
+  -1.04523       -1.04523
+  -4.17379       -4.17379
+   0.453503       0.453503
+   5.64541        5.64541
+   2.72225        2.72225
+  -1.67354       -1.67354
+  -6.68729       -6.68729
+  -1.20785       -1.20785
+   3.51562        3.51562
+   2.38257        2.38257
+   2.75735        2.75735
+  -4.62925       -4.62925
+   7.98247        7.98247
+   6.254          6.254
+   3.85448        3.85448
+  -4.40298       -4.40298
+  -8.28751       -8.28751
+  -7.28055       -7.28055
+   7.31675        7.31675
+   3.53957        3.53957
+   2.94378        2.94378
+   1.41268        1.41268
+   5.2878         5.2878
+  -0.807317      -0.807317
+ -13.141        -13.141
+   5.71505        5.71505
+  -3.86739       -3.86739
+   0.922435       0.922435
+  -4.52167       -4.52167
+   0.82741        0.82741
+   4.1254         4.1254
+  -3.64229       -3.64229
+  -4.34879       -4.34879
+  -5.69361       -5.69361
+  10.0503        10.0503
+  -6.20878       -6.20878
+  -5.70531       -5.70531
+  -0.265037      -0.265037
+   4.91217        4.91217
+  -9.85839       -9.85839
+   9.14639        9.14639
+   0.78426        0.78426
+  -6.03581       -6.03581
+  -1.225         -1.225
+  -1.82514       -1.82514
+  -4.38257       -4.38257
+  -4.14898       -4.14898
+   1.30056        1.30056
+  -4.04361       -4.04361
+ -10.7862       -10.7862
+  -1.71033       -1.71033
+  -5.3235        -5.3235
+  -5.05158       -5.05158
+   2.03088        2.03088
+  -4.639         -4.639
+  -8.90379       -8.90379
+  -1.46286       -1.46286
+   4.78737        4.78737
+   2.84292        2.84292
+  -4.60125       -4.60125
+  -0.454598      -0.454598
+  -3.54703       -3.54703
+  -3.15574       -3.15574
+  -5.66794       -5.66794
+  -0.499733      -0.499733
+   4.80394        4.80394
+   7.0018         7.0018
+ -12.2494       -12.2494
+  -0.705371      -0.705371
+   0.0740021      0.0740021
+  -2.66987       -2.66987
+   2.48263        2.48263
+  -9.06332       -9.06332
+  -1.01261       -1.01261
+   3.84118        3.84118
+   4.21216        4.21216
+  -1.18673       -1.18673
+ -11.0005       -11.0005
+  -9.71638       -9.71638
+   1.76212        1.76212
+  -2.83766       -2.83766
+  -9.13768       -9.13768
+  -1.05015       -1.05015
+   2.53008        2.53008
+   0.379504       0.379504
+   5.28803        5.28803
+  -6.17221       -6.17221
+   5.75619        5.75619
+   2.3737         2.3737
+  -9.0974        -9.0974
+  -7.85433       -7.85433
+ -10.9094       -10.9094
+   1.20756        1.20756
+   2.61486        2.61486
+   1.23359        1.23359
+  43.6151        43.6151
+  -1.72859       -1.72859
+  -0.965831      -0.965831
+  -0.482239      -0.482239
+  -1.82159       -1.82159
+   1.661          1.661
+   1.93636        1.93636
+ -11.9999       -11.9999
+   0.104367       0.104367
+  -1.70555       -1.70555
+  -9.81074       -9.81074
+  12.7941        12.7941
+  -3.36221       -3.36221
+  -6.06523       -6.06523
+   0.47411        0.47411
+  -6.64475       -6.64475
+  -0.763006      -0.763006
+  -3.9763        -3.9763
+  -2.86732       -2.86732
+ -20.6937       -20.6937
+   1.84418        1.84418
+   5.65243        5.65243
+  10.7255        10.7255
+  -1.21293       -1.21293
+   3.15057        3.15057
+   8.96094        8.96094
+  -0.205015      -0.205015
+   8.44579        8.44579
+   2.01362        2.01362
+   2.36648        2.36648
+  11.6752        11.6752
+   2.19072        2.19072
+ -13.9182       -13.9182
+   3.3257         3.3257
+  -6.60627       -6.60627
+   1.62083        1.62083
+  -2.00847       -2.00847
+  11.6978        11.6978
+   5.93254        5.93254
+   4.93134        4.93134
+  -2.50847       -2.50847
+  -5.92846       -5.92846
+   1.16717        1.16717
+   6.9673         6.9673
+  -1.21182       -1.21182
+   7.25413        7.25413
+  -4.24031       -4.24031
+  -3.12368       -3.12368
+   1.73734        1.73734
+  -2.6551        -2.6551
+   5.01063        5.01063
+  10.9923        10.9923
+   3.08502        3.08502
+  -1.67866       -1.67866
+  10.7003        10.7003
+  -0.982895      -0.982895
+   1.97681        1.97681
+  -1.29045       -1.29045
+   1.64227        1.64227
+   3.21157        3.21157
+  -4.63376       -4.63376
+   4.47725        4.47725
+   7.77208        7.77208
+   0.332548       0.332548
+   2.82084        2.82084
+   0.958649       0.958649
+   1.21302        1.21302
+  -3.16936       -3.16936
+   0.0672417      0.0672417
+   0.563038       0.563038
+  -1.87542       -1.87542
+  -3.01753       -3.01753
+   2.73107        2.73107
+  -3.68276       -3.68276
+   4.64376        4.64376
+ -12.4341       -12.4341
+   4.43429        4.43429
+   5.72878        5.72878
+   2.39332        2.39332
+   1.91106        1.91106
+   2.50458        2.50458
+   0.942479       0.942479
+  -0.489758      -0.489758
+   0.311101       0.311101
+  -2.74953       -2.74953
+   4.95959        4.95959
+   1.26862        1.26862
+  10.3622        10.3622
+   3.61213        3.61213
+  -2.19285       -2.19285
+   1.28587        1.28587
+  -1.85274       -1.85274
+  -1.62541       -1.62541
+   2.00382        2.00382
+  -5.8959        -5.8959
+  -0.918042      -0.918042
+   6.43711        6.43711
+   0.419441       0.419441
+  -2.61133       -2.61133
+  -0.0277654     -0.0277654
+   2.77443        2.77443
+   3.83764        3.83764
+  -1.44486       -1.44486
+  -0.611288      -0.611288
+  -4.30436       -4.30436
+   5.29466        5.29466
+   1.56058        1.56058
+   1.88962        1.88962
+   0.761408       0.761408
+   1.76505        1.76505
+   1.18453        1.18453
+   1.71559        1.71559
+  -3.14851       -3.14851
+   2.73145        2.73145
+  -1.23904       -1.23904
+   0.00672958     0.00672958
+   3.40979        3.40979
+  -1.77498       -1.77498
+  -7.12266       -7.12266
+  -9.24697       -9.24697
+  -4.12038       -4.12038
+  -2.77817       -2.77817
+   8.23453        8.23453
+  -1.29818       -1.29818
+  -7.02203       -7.02203
+  -5.8994        -5.8994
+   8.20499        8.20499
+   0.356509       0.356509
+  -0.515947      -0.515947
+  -6.23904       -6.23904
+   5.59801        5.59801
+  -4.44281       -4.44281
+  -2.28591       -2.28591
+  -3.31819       -3.31819
+   2.39253        2.39253
+   3.18355        3.18355
+  -2.73303       -2.73303
+  -0.0346074     -0.0346074
+ -10.2692       -10.2692
+   6.74308        6.74308
+   5.72055        5.72055
+  -4.49033       -4.49033
+   1.99176        1.99176
+   6.10782        6.10782
+   2.65759        2.65759
+   1.97884        1.97884
+   0.927606       0.927606
+   1.25006        1.25006
+   9.3695         9.3695
+  -2.75726       -2.75726
+  -0.580415      -0.580415
+   2.92463        2.92463
+  -4.49535       -4.49535
+  -1.61397       -1.61397
+   3.26733        3.26733
+  -3.61505       -3.61505
+  -2.46453       -2.46453
+   2.42436        2.42436
+   5.68683        5.68683
+   6.07494        6.07494
+   4.35205        4.35205
+  -5.29467       -5.29467
+  -3.90039       -3.90039
+  -1.70776       -1.70776
+  -6.3172        -6.3172
+   4.03858        4.03858
+  -2.58786       -2.58786
+  -1.1514        -1.1514
+  -0.632569      -0.632569
+  -0.343314      -0.343314
+ -12.2115       -12.2115
+   0.405742       0.405742
+  -6.46017       -6.46017
+  -2.30808       -2.30808
+   1.1336         1.1336
+   1.47556        1.47556
+   1.98494        1.98494
+   2.24865        2.24865
+  -1.65786       -1.65786
+  -4.62769       -4.62769
+   4.43717        4.43717
+   8.75249        8.75249
+   4.29167        4.29167
+  -3.96876       -3.96876
+  -3.52244       -3.52244
+   0.161164       0.161164
+  -4.13202       -4.13202
+   1.42269        1.42269
+  -3.05155       -3.05155
+   1.81371        1.81371
+  -1.03765       -1.03765
+   0.696656       0.696656
+   2.95359        2.95359
+  -4.74837       -4.74837
+  -9.03481       -9.03481
+   4.8852         4.8852
+   9.47173        9.47173
+  11.3037        11.3037
+  -3.88084       -3.88084
+  -5.99356       -5.99356
+   7.81639        7.81639
+  -6.51949       -6.51949
+   7.801          7.801
+  -0.795429      -0.795429
+  -0.801046      -0.801046
+   2.70658        2.70658
+   5.51012        5.51012
+   1.8181         1.8181
+  -0.452854      -0.452854
+ -10.1558       -10.1558
+   1.95877        1.95877
+  -3.88197       -3.88197
+   1.72033        1.72033
+  -1.8939        -1.8939
+  -1.64082       -1.64082
+  -0.409815      -0.409815
+   9.98658        9.98658
+  -0.115277      -0.115277
+   1.49827        1.49827
+   1.6696         1.6696
+   2.29297        2.29297
+  -2.14941       -2.14941
+   2.43318        2.43318
+   3.59845        3.59845
+  -4.58877       -4.58877
+  -9.25371       -9.25371
+   2.03609        2.03609
+   5.5921         5.5921
+  -0.532859      -0.532859
+   4.34937        4.34937
+   1.57036        1.57036
+   2.30747        2.30747
+   7.5055         7.5055
+   3.41771        3.41771
+   0.589402       0.589402
+   1.55834        1.55834
+   5.12407        5.12407
+  -1.41727       -1.41727
+   1.03223        1.03223
+  -2.06257       -2.06257
+   3.11532        3.11532
+   1.90042        1.90042
+   8.66814        8.66814
+   5.36716        5.36716
+   2.38085        2.38085
+   5.72834        5.72834
+  -6.5998        -6.5998
+   0.852569       0.852569
+  -7.5648        -7.5648
+   2.98063        2.98063
+   7.81573        7.81573
+   1.82276        1.82276
+  -1.81083       -1.81083
+   5.48043        5.48043
+  -1.85315       -1.85315
+  -1.62277       -1.62277
+ -10.4951       -10.4951
+   5.34799        5.34799
+  -1.77515       -1.77515
+   5.88005        5.88005
+   0.0799242      0.0799242
+   1.23264        1.23264
+ -11.835        -11.835
+   3.56828        3.56828
+   7.53741        7.53741
+  -5.24051       -5.24051
+  -0.206917      -0.206917
+   4.36865        4.36865
+  -4.10348       -4.10348
+   0.857712       0.857712
+  -5.09677       -5.09677
+   7.37208        7.37208
+  -3.14614       -3.14614
+  12.061         12.061
+   4.80096        4.80096
+   2.82421        2.82421
+  -4.97446       -4.97446
+ -11.0289       -11.0289
+  -8.33282       -8.33282
+   0.69922        0.69922
+   5.08771        5.08771
+   2.65174        2.65174
+  -3.30182       -3.30182
+   5.21741        5.21741
+   8.85373        8.85373
+   8.36416        8.36416
+   2.54295        2.54295
+  -1.61657       -1.61657
+   1.12017        1.12017
+  -7.33205       -7.33205
+   3.82582        3.82582
+  -0.858026      -0.858026
+   1.40304        1.40304
+   1.35079        1.35079
+   4.19532        4.19532
+  -1.77923       -1.77923
+ -10.5119       -10.5119
+  10.8061        10.8061
+  -3.49603       -3.49603
+   3.12404        3.12404
+  -3.93328       -3.93328
+  -6.73356       -6.73356
+   1.80532        1.80532
+  -0.368024      -0.368024
+  -3.47875       -3.47875
+  -4.22893       -4.22893
+   2.52519        2.52519
+  -3.54943       -3.54943
+  -2.39869       -2.39869
+   4.22126        4.22126
+  -0.253856      -0.253856
+   7.51866        7.51866
+  -4.54093       -4.54093
+   3.44497        3.44497
+   4.77417        4.77417
+   4.49646        4.49646
+  -5.78678       -5.78678
+   0.745013       0.745013
+   1.69763        1.69763
+  -2.64759       -2.64759
+   1.66108        1.66108
+  -4.68276       -4.68276
+   5.31823        5.31823
+   3.52288        3.52288
+   4.9695         4.9695
+  12.2016        12.2016
+   2.46849        2.46849
+  -7.60038       -7.60038
+   8.21628        8.21628
+   5.99856        5.99856
+  -6.80947       -6.80947
+   7.22522        7.22522
+  -2.00065       -2.00065
+  -8.24049       -8.24049
+  -0.0804049     -0.0804049
+  -2.06638       -2.06638
+  -2.82884       -2.82884
+  -4.25891       -4.25891
+  -5.20258       -5.20258
+  -3.19396       -3.19396
+  -5.14527       -5.14527
+  -4.28244       -4.28244
+   4.70805        4.70805
+  -3.08065       -3.08065
+  -4.86906       -4.86906
+ -29.0266       -29.0266
+  -1.22941       -1.22941
+  -1.30928       -1.30928
+  -6.35234       -6.35234
+   1.87904        1.87904
+   8.37797        8.37797
+  -5.8821        -5.8821
+   3.10138        3.10138
+  -3.27553       -3.27553
+  -0.208451      -0.208451
+  -2.28999       -2.28999
+  12.2896        12.2896
+  -1.27394       -1.27394
+  -3.41924       -3.41924
+  -0.289592      -0.289592
+   1.79867        1.79867
+   1.98504        1.98504
+   1.55159        1.55159
+   1.10858        1.10858
+   0.352842       0.352842
+  -0.309044      -0.309044
+  -0.165336      -0.165336
+   1.15822        1.15822
+  -1.39342       -1.39342
+  -0.162562      -0.162562
+  -8.06055       -8.06055
+  -5.02776       -5.02776
+  -8.66927       -8.66927
+   1.14576        1.14576
+  -1.52122       -1.52122
+  -1.29436       -1.29436
+   3.26421        3.26421
+   7.55561        7.55561
+   7.7265         7.7265
+  -0.48821       -0.48821
+  12.439         12.439
+   7.0264         7.0264
+ -11.9855       -11.9855
+  -3.74151       -3.74151
+  -0.200302      -0.200302
+   5.39515        5.39515
+  -4.3468        -4.3468
+   9.25599        9.25599
+   3.37455        3.37455
+  -6.15424       -6.15424
+  -6.6271        -6.6271
+   0.000272481    0.000272481
+  -5.48117       -5.48117
+  -0.493191      -0.493191
+  -3.46473       -3.46473
+   2.33812        2.33812
+   0.885965       0.885965
+   4.74926        4.74926
+   1.51959        1.51959
+   2.50956        2.50956
+  -0.728024      -0.728024
+   1.0381         1.0381
+   5.48121        5.48121
+  -1.68033       -1.68033
+  -5.05915       -5.05915
+  -0.646233      -0.646233
+   0.614062       0.614062
+   4.54219        4.54219
+  -1.63006       -1.63006
+  -3.10589       -3.10589
+  -3.12801       -3.12801
+  -5.98177       -5.98177
+  -3.59188       -3.59188
+   1.7066         1.7066
+  -7.43935       -7.43935
+  10.6141        10.6141
+  12.6478        12.6478
+  -1.7222        -1.7222
+  -2.1519        -2.1519
+  -7.16573       -7.16573
+   0.887314       0.887314
+  -8.59735       -8.59735
+  -1.3609        -1.3609
+   4.47651        4.47651
+   0.900892       0.900892
+   7.81857        7.81857
+   6.19857        6.19857
+   2.12844        2.12844
+   3.08551        3.08551
+   4.15866        4.15866
+   2.09657        2.09657
+  -2.27786       -2.27786
+   1.33571        1.33571
+   4.46899        4.46899
+   4.46674        4.46674
+   3.20736        3.20736
+   5.68287        5.68287
+  10.1058        10.1058
+   5.1894         5.1894
+   3.5452         3.5452
+  10.06          10.06
+   7.02935        7.02935
+  -1.06066       -1.06066
+  10.32          10.32
+  -0.860463      -0.860463
+   5.95992        5.95992
+  -6.30137       -6.30137
+  -5.01947       -5.01947
+   5.75187        5.75187
+  -1.10079       -1.10079
+  -1.91783       -1.91783
+   0.815744       0.815744
+  -0.958663      -0.958663
+  -3.28825       -3.28825
+  -6.37854       -6.37854
+   6.91577        6.91577
+   2.54565        2.54565
+   1.39487        1.39487
+   1.59679        1.59679
+   4.72347        4.72347
+   2.49221        2.49221
+   1.29896        1.29896
+  -4.08232       -4.08232
+  -0.648436      -0.648436
+  -6.43531       -6.43531
+  -0.556197      -0.556197
+  -1.40304       -1.40304
+   0.699818       0.699818
+  -5.29777       -5.29777
+  -3.44335       -3.44335
+   7.35309        7.35309
+   8.846          8.846
+   8.39833        8.39833
+  -2.71436       -2.71436
+   3.37063        3.37063
+  -3.18723       -3.18723
+   1.32256        1.32256
+  -3.09485       -3.09485
+   8.78146        8.78146
+  -1.30004       -1.30004
+   3.03526        3.03526
+  -1.4592        -1.4592
+   3.90288        3.90288
+ -13.5124       -13.5124
+   1.35105        1.35105
+   3.37337        3.37337
+   2.5171         2.5171
+  -4.22085       -4.22085
+  13.1858        13.1858
+  -6.02839       -6.02839
+   5.75692        5.75692
+   2.46171        2.46171
+  -0.950315      -0.950315
+  -3.63255       -3.63255
+   1.88           1.88
+   5.48758        5.48758
+   4.96786        4.96786
+  -6.17199       -6.17199
+  -0.284244      -0.284244
+  -1.80256       -1.80256
+   3.03221        3.03221
+  -8.90171       -8.90171
+  -8.66084       -8.66084
+  -9.06366       -9.06366
+  -3.02007       -3.02007
+  -8.2276        -8.2276
+   8.10032        8.10032
+  -4.11364       -4.11364
+  -3.39291       -3.39291
+   3.64208        3.64208
+  -0.739833      -0.739833
+  -2.84156       -2.84156
+  -0.843081      -0.843081
+  -0.249744      -0.249744
+   7.05075        7.05075
+   0.369632       0.369632
+  -1.90893       -1.90893
+   9.79465        9.79465
+   3.52356        3.52356
+   4.14091        4.14091
+   1.66568        1.66568
+ -10.7162       -10.7162
+  -7.64522       -7.64522
+   1.54688        1.54688
+   7.84479        7.84479
+   0.466458       0.466458
+   4.03315        4.03315
+   0.472926       0.472926
+   1.73319        1.73319
+   1.79317        1.79317
+   1.46234        1.46234
+  -8.45267       -8.45267
+   7.30327        7.30327
+   3.08869        3.08869
+   5.27442        5.27442
+   2.92876        2.92876
+  -1.6673        -1.6673
+  14.4442        14.4442
+  13.4055        13.4055
+  -1.47522       -1.47522
+  -3.57821       -3.57821
+   9.00659        9.00659
+  -9.6723        -9.6723
+   2.8818         2.8818
+  -2.61898       -2.61898
+   1.17927        1.17927
+  -3.15135       -3.15135
+  -0.976968      -0.976968
+   1.45062        1.45062
+   4.66687        4.66687
+   4.94346        4.94346
+  -2.20375       -2.20375
+   2.93643        2.93643
+   7.51365        7.51365
+   6.50034        6.50034
+   1.74088        1.74088
+  -4.43403       -4.43403
+   0.796894       0.796894
+  -1.23803       -1.23803
+   5.33941        5.33941
+   4.90517        4.90517
+   0.569053       0.569053
+  -0.609673      -0.609673
+   5.091          5.091
+   1.76184        1.76184
+  -3.81174       -3.81174
+  -5.39095       -5.39095
+  -3.09718       -3.09718
+  -1.87868       -1.87868
+  -4.85278       -4.85278
+  -1.05327       -1.05327
+  -1.11892       -1.11892
+  -3.52006       -3.52006
+  -2.8466        -2.8466
+   3.03494        3.03494
+   3.7605         3.7605
+  -1.8123        -1.8123
+  -5.10186       -5.10186
+   2.85973        2.85973
+  -3.6241        -3.6241
+   1.78302        1.78302
+ -12.3108       -12.3108
+   0.378043       0.378043
+  -1.70182       -1.70182
+  -0.91773       -0.91773
+  -5.37355       -5.37355

Using postprocessing function

Add normalization as postprocessing function to normalize embeddings on reception (for easy cosine similarity later)

using LinearAlgebra
+schema = PT.OllamaManagedSchema()
+
+msg = aiembed(schema,
+    ["embed me", "and me too"],
+    LinearAlgebra.normalize;
+    model = "openhermes2.5-mistral")
DataMessage(Matrix{Float64} of size (4096, 2))

Cosine similarity is then a simple multiplication

msg.content' * msg.content[:, 1]
2-element Vector{Float64}:
+ 0.9999999999999946
+ 0.34130017815042357

This page was generated using Literate.jl.

diff --git a/dev/frequently_asked_questions/index.html b/dev/frequently_asked_questions/index.html new file mode 100644 index 000000000..c69774dce --- /dev/null +++ b/dev/frequently_asked_questions/index.html @@ -0,0 +1,3 @@ + +F.A.Q. · PromptingTools.jl
GitHub

Frequently Asked Questions

Why OpenAI

OpenAI's models are at the forefront of AI research and provide robust, state-of-the-art capabilities for many tasks.

There will be situations not or cannot use it (eg, privacy, cost, etc.). In that case, you can use local models (eg, Ollama) or other APIs (eg, Anthropic).

Note: To get started with Ollama.ai, see the Setup Guide for Ollama section below.

Data Privacy and OpenAI

At the time of writing, OpenAI does NOT use the API calls for training their models.

API

OpenAI does not use data submitted to and generated by our API to train OpenAI models or improve OpenAI’s service offering. In order to support the continuous improvement of our models, you can fill out this form to opt-in to share your data with us. – How your data is used to improve our models

You can always double-check the latest information on the OpenAI's How we use your data page.

Resources:

Creating OpenAI API Key

You can get your API key from OpenAI by signing up for an account and accessing the API section of the OpenAI website.

  1. Create an account with OpenAI
  2. Go to API Key page
  3. Click on “Create new secret key”

!!! Do not share it with anyone and do NOT save it to any files that get synced online.

Resources:

Pro tip: Always set the spending limits!

Setting OpenAI Spending Limits

OpenAI allows you to set spending limits directly on your account dashboard to prevent unexpected costs.

  1. Go to OpenAI Billing
  2. Set Soft Limit (you’ll receive a notification) and Hard Limit (API will stop working not to spend more money)

A good start might be a soft limit of c.$5 and a hard limit of c.$10 - you can always increase it later in the month.

Resources:

How much does it cost? Is it worth paying for?

If you use a local model (eg, with Ollama), it's free. If you use any commercial APIs (eg, OpenAI), you will likely pay per "token" (a sub-word unit).

For example, a simple request with a simple question and 1 sentence response in return (”Is statement XYZ a positive comment”) will cost you ~0.0001 (ie, one hundredth of a cent)

Is it worth paying for?

GenAI is a way to buy time! You can pay cents to save tens of minutes every day.

Continuing the example above, imagine you have a table with 200 comments. Now, you can parse each one of them with an LLM for the features/checks you need. Assuming the price per call was 0.0001, you'd pay 2 cents for the job and save 30-60 minutes of your time!

Resources:

Configuring the Environment Variable for API Key

To use the OpenAI API with PromptingTools.jl, set your API key as an environment variable:

ENV["OPENAI_API_KEY"] = "your-api-key"

As a one-off, you can:

  • set it in the terminal before launching Julia: export OPENAI_API_KEY = <your key>
  • set it in your setup.jl (make sure not to commit it to GitHub!)

Make sure to start Julia from the same terminal window where you set the variable. Easy check in Julia, run ENV["OPENAI_API_KEY"] and you should see your key!

A better way:

  • On a Mac, add the configuration line to your terminal's configuration file (eg, ~/.zshrc). It will get automatically loaded every time you launch the terminal
  • On Windows, set it as a system variable in "Environment Variables" settings (see the Resources)

Resources:

Note: In the future, we hope to add Preferences.jl-based workflow to set the API key and other preferences.

Understanding the API Keyword Arguments in aigenerate (api_kwargs)

See OpenAI API reference for more information.

Instant Access from Anywhere

For easy access from anywhere, add PromptingTools into your startup.jl (can be found in ~/.julia/config/startup.jl).

Add the following snippet:

using PromptingTools
+const PT = PromptingTools # to access unexported functions and types

Now, you can just use ai"Help me do X to achieve Y" from any REPL session!

Open Source Alternatives

The ethos of PromptingTools.jl is to allow you to use whatever model you want, which includes Open Source LLMs. The most popular and easiest to setup is Ollama.ai - see below for more information.

Setup Guide for Ollama

Ollama runs a background service hosting LLMs that you can access via a simple API. It's especially useful when you're working with some sensitive data that should not be sent anywhere.

Installation is very easy, just download the latest version here.

Once you've installed it, just launch the app and you're ready to go!

To check if it's running, go to your browser and open 127.0.0.1:11434. You should see the message "Ollama is running". Alternatively, you can run ollama serve in your terminal and you'll get a message that it's already running.

There are many models available in Ollama Library, including Llama2, CodeLlama, SQLCoder, or my personal favorite openhermes2.5-mistral.

Download new models with ollama pull <model_name> (eg, ollama pull openhermes2.5-mistral).

Show currently available models with ollama list.

See Ollama.ai for more information.

diff --git a/dev/getting_started/index.html b/dev/getting_started/index.html new file mode 100644 index 000000000..428017ebd --- /dev/null +++ b/dev/getting_started/index.html @@ -0,0 +1,7 @@ + +Getting Started · PromptingTools.jl
GitHub

Getting Started

Prerequisites

OpenAI API key saved in the environment variable OPENAI_API_KEY

You will need to register with OpenAI and generate an API key:

  1. Create an account with OpenAI
  2. Go to API Key page
  3. Click on “Create new secret key”

!!! Do not share it with anyone and do NOT save it to any files that get synced online.

Resources:

You will need to set this key as an environment variable before using PromptingTools.jl:

For a quick start, simply set it via ENV["OPENAI_API_KEY"] = "your-api-key" Alternatively, you can:

  • set it in the terminal before launching Julia: export OPENAI_API_KEY = <your key>
  • set it in your setup.jl (make sure not to commit it to GitHub!)

Make sure to start Julia from the same terminal window where you set the variable. Easy check in Julia, run ENV["OPENAI_API_KEY"] and you should see your key!

For other options or more robust solutions, see the FAQ section.

Resources:

Installation

The PromptingTools package has not yet been registered. But it can be installed using the following commands:

using Pkg
+Pkg.add("https://github.com/svilupp/PromptingTools.jl")

Throughout the rest of this tutorial, we will assume that you have installed the PromptingTools package and have already typed using PromptingTools to bring all of the relevant variables into your current namespace.

Quick Start with @ai_str

The easiest start is the @ai_str macro. Simply type ai"your prompt" and you will get a response from the default model (GPT-3.5 Turbo).

ai"What is the capital of France?"
[ Info: Tokens: 31 @ Cost: $0.0 in 1.5 seconds --> Be in control of your spending! 
+AIMessage("The capital of France is Paris.")

Returned object is a light wrapper with generated message in field :content (eg, ans.content) for additional downstream processing.

You can easily inject any variables with string interpolation:

country = "Spain"
+ai"What is the capital of \$(country)?"
[ Info: Tokens: 32 @ Cost: $0.0001 in 0.5 seconds
+AIMessage("The capital of Spain is Madrid.")

Pro tip: Use after-string-flags to select the model to be called, eg, ai"What is the capital of France?"gpt4 (use gpt4t for the new GPT-4 Turbo model). Great for those extra hard questions!

Using aigenerate with placeholders

For more complex prompt templates, you can use handlebars-style templating and provide variables as keyword arguments:

msg = aigenerate("What is the capital of {{country}}? Is the population larger than {{population}}?", country="Spain", population="1M")
[ Info: Tokens: 74 @ Cost: $0.0001 in 1.3 seconds
+AIMessage("The capital of Spain is Madrid. And yes, the population of Madrid is larger than 1 million. As of 2020, the estimated population of Madrid is around 3.3 million people.")

Pro tip: Use asyncmap to run multiple AI-powered tasks concurrently.

Pro tip: If you use slow models (like GPT-4), you can use async version of @ai_str -> @aai_str to avoid blocking the REPL, eg, aai"Say hi but slowly!"gpt4

For more practical examples, see the Various Examples section.

diff --git a/dev/index.html b/dev/index.html index f39ca3efa..cea70107f 100644 --- a/dev/index.html +++ b/dev/index.html @@ -1,213 +1,2 @@ -Home · PromptingTools.jl
GitHub

PromptingTools

Documentation for PromptingTools.

PromptingTools.AITemplate — Type
AITemplate

AITemplate is a template for a conversation prompt. This type is merely a container for the template name, which is resolved into a set of messages (=prompt) by render.

Naming Convention

  • Template names should be in CamelCase
  • Follow the format <Persona>...<Variable>... where possible, eg, JudgeIsItTrue, ``
    • Starting with the Persona (=System prompt), eg, Judge = persona is meant to judge some provided information
    • Variable to be filled in with context, eg, It = placeholder it
    • Ending with the variable name is helpful, eg, JuliaExpertTask for a persona to be an expert in Julia language and task is the placeholder name
  • Ideally, the template name should be self-explanatory, eg, JudgeIsItTrue = persona is meant to judge some provided information where it is true or false

Examples

Save time by re-using pre-made templates, just fill in the placeholders with the keyword arguments:

msg = aigenerate(:JuliaExpertAsk; ask = "How do I add packages?")

The above is equivalent to a more verbose version that explicitly uses the dispatch on AITemplate:

msg = aigenerate(AITemplate(:JuliaExpertAsk); ask = "How do I add packages?")

Find available templates with aitemplates:

tmps = aitemplates("JuliaExpertAsk")
-# Will surface one specific template
-# 1-element Vector{AITemplateMetadata}:
-# PromptingTools.AITemplateMetadata
-#   name: Symbol JuliaExpertAsk
-#   description: String "For asking questions about Julia language. Placeholders: `ask`"
-#   version: String "1"
-#   wordcount: Int64 237
-#   variables: Array{Symbol}((1,))
-#   system_preview: String "You are a world-class Julia language programmer with the knowledge of the latest syntax. Your commun"
-#   user_preview: String "# Question
-
-{{ask}}"
-#   source: String ""

The above gives you a good idea of what the template is about, what placeholders are available, and how much it would cost to use it (=wordcount).

Search for all Julia-related templates:

tmps = aitemplates("Julia")
-# 2-element Vector{AITemplateMetadata}... -> more to come later!

If you are on VSCode, you can leverage nice tabular display with vscodedisplay:

using DataFrames
-tmps = aitemplates("Julia") |> DataFrame |> vscodedisplay

I have my selected template, how do I use it? Just use the "name" in aigenerate or aiclassify like you see in the first example!

You can inspect any template by "rendering" it (this is what the LLM will see):

julia> AITemplate(:JudgeIsItTrue) |> PromptingTools.render

See also: save_template, load_template, load_templates! for more advanced use cases (and the corresponding script in examples/ folder)

source
PromptingTools.ChatMLSchema — Type

ChatMLSchema is used by many open-source chatbots, by OpenAI models (under the hood) and by several models and inferfaces (eg, Ollama, vLLM)

You can explore it on tiktokenizer

It uses the following conversation structure:

<im_start>system
-...<im_end>
-<|im_start|>user
-...<|im_end|>
-<|im_start|>assistant
-...<|im_end|>
source
PromptingTools.MaybeExtract — Type

Extract a result from the provided data, if any, otherwise set the error and message fields.

Arguments

  • error::Bool: true if a result is found, false otherwise.
  • message::String: Only present if no result is found, should be short and concise.
source
PromptingTools.OllamaManagedSchema — Type

Ollama by default manages different models and their associated prompt schemas when you pass system_prompt and prompt fields to the API.

Warning: It works only for 1 system message and 1 user message, so anything more than that has to be rejected.

If you need to pass more messagese / longer conversational history, you can use define the model-specific schema directly and pass your Ollama requests with raw=true, which disables and templating and schema management by Ollama.

source
PromptingTools.OpenAISchema — Type

OpenAISchema is the default schema for OpenAI models.

It uses the following conversation template:

[Dict(role="system",content="..."),Dict(role="user",content="..."),Dict(role="assistant",content="...")]

It's recommended to separate sections in your prompt with markdown headers (e.g. `##Answer

`).

source
PromptingTools.aiclassify — Method
aiclassify(prompt_schema::AbstractOpenAISchema, prompt::ALLOWED_PROMPT_TYPE;
-api_kwargs::NamedTuple = (logit_bias = Dict(837 => 100, 905 => 100, 9987 => 100),
-    max_tokens = 1, temperature = 0),
-kwargs...)

Classifies the given prompt/statement as true/false/unknown.

Note: this is a very simple classifier, it is not meant to be used in production. Credit goes to AAAzzam.

It uses Logit bias trick and limits the output to 1 token to force the model to output only true/false/unknown.

Output tokens used (via api_kwargs):

  • 837: ' true'
  • 905: ' false'
  • 9987: ' unknown'

Arguments

  • prompt_schema::AbstractOpenAISchema: The schema for the prompt.
  • prompt: The prompt/statement to classify if it's a String. If it's a Symbol, it is expanded as a template via render(schema,template).

Example

aiclassify("Is two plus two four?") # true
-aiclassify("Is two plus three a vegetable on Mars?") # false

aiclassify returns only true/false/unknown. It's easy to get the proper Bool output type out with tryparse, eg,

tryparse(Bool, aiclassify("Is two plus two four?")) isa Bool # true

Output of type Nothing marks that the model couldn't classify the statement as true/false.

Ideally, we would like to re-use some helpful system prompt to get more accurate responses. For this reason we have templates, eg, :JudgeIsItTrue. By specifying the template, we can provide our statement as the expected variable (it in this case) See that the model now correctly classifies the statement as "unknown".

aiclassify(:JudgeIsItTrue; it = "Is two plus three a vegetable on Mars?") # unknown

For better results, use higher quality models like gpt4, eg, 

aiclassify(:JudgeIsItTrue;
-    it = "If I had two apples and I got three more, I have five apples now.",
-    model = "gpt4") # true
source
PromptingTools.aiembed — Method
aiembed(prompt_schema::AbstractOllamaManagedSchema,
-        doc_or_docs::Union{AbstractString, Vector{<:AbstractString}},
-        postprocess::F = identity;
-        verbose::Bool = true,
-        api_key::String = API_KEY,
-        model::String = MODEL_EMBEDDING,
-        http_kwargs::NamedTuple = (retry_non_idempotent = true,
-                                   retries = 5,
-                                   readtimeout = 120),
-        api_kwargs::NamedTuple = NamedTuple(),
-        kwargs...) where {F <: Function}

The aiembed function generates embeddings for the given input using a specified model and returns a message object containing the embeddings, status, token count, and elapsed time.

Arguments

  • prompt_schema::AbstractOllamaManagedSchema: The schema for the prompt.
  • doc_or_docs::Union{AbstractString, Vector{<:AbstractString}}: The document or list of documents to generate embeddings for. The list of documents is processed sequentially, so users should consider implementing an async version with with Threads.@spawn
  • postprocess::F: The post-processing function to apply to each embedding. Defaults to the identity function, but could be LinearAlgebra.normalize.
  • verbose::Bool: A flag indicating whether to print verbose information. Defaults to true.
  • api_key::String: The API key to use for the OpenAI API. Defaults to API_KEY.
  • model::String: The model to use for generating embeddings. Defaults to MODEL_EMBEDDING.
  • http_kwargs::NamedTuple: Additional keyword arguments for the HTTP request. Defaults to empty NamedTuple.
  • api_kwargs::NamedTuple: Additional keyword arguments for the Ollama API. Defaults to an empty NamedTuple.
  • kwargs: Prompt variables to be used to fill the prompt/template

Returns

  • msg: A DataMessage object containing the embeddings, status, token count, and elapsed time.

Note: Ollama API currently does not return the token count, so it's set to (0,0)

Example

const PT = PromptingTools
-schema = PT.OllamaManagedSchema()
-
-msg = aiembed(schema, "Hello World"; model="openhermes2.5-mistral")
-msg.content # 4096-element JSON3.Array{Float64...

We can embed multiple strings at once and they will be hcat into a matrix (ie, each column corresponds to one string)

const PT = PromptingTools
-schema = PT.OllamaManagedSchema()
-
-msg = aiembed(schema, ["Hello World", "How are you?"]; model="openhermes2.5-mistral")
-msg.content # 4096×2 Matrix{Float64}:

If you plan to calculate the cosine distance between embeddings, you can normalize them first:

const PT = PromptingTools
-using LinearAlgebra
-schema = PT.OllamaManagedSchema()
-
-msg = aiembed(schema, ["embed me", "and me too"], LinearAlgebra.normalize; model="openhermes2.5-mistral")
-
-# calculate cosine distance between the two normalized embeddings as a simple dot product
-msg.content' * msg.content[:, 1] # [1.0, 0.34]

Similarly, you can use the postprocess argument to materialize the data from JSON3.Object by using postprocess = copy

const PT = PromptingTools
-schema = PT.OllamaManagedSchema()
-
-msg = aiembed(schema, "Hello World", copy; model="openhermes2.5-mistral")
-msg.content # 4096-element Vector{Float64}
source
PromptingTools.aiembed — Method
aiembed(prompt_schema::AbstractOpenAISchema,
-        doc_or_docs::Union{AbstractString, Vector{<:AbstractString}},
-        postprocess::F = identity;
-        verbose::Bool = true,
-        api_key::String = API_KEY,
-        model::String = MODEL_EMBEDDING,
-        http_kwargs::NamedTuple = (retry_non_idempotent = true,
-                                   retries = 5,
-                                   readtimeout = 120),
-        api_kwargs::NamedTuple = NamedTuple(),
-        kwargs...) where {F <: Function}

The aiembed function generates embeddings for the given input using a specified model and returns a message object containing the embeddings, status, token count, and elapsed time.

Arguments

  • prompt_schema::AbstractOpenAISchema: The schema for the prompt.
  • doc_or_docs::Union{AbstractString, Vector{<:AbstractString}}: The document or list of documents to generate embeddings for.
  • postprocess::F: The post-processing function to apply to each embedding. Defaults to the identity function.
  • verbose::Bool: A flag indicating whether to print verbose information. Defaults to true.
  • api_key::String: The API key to use for the OpenAI API. Defaults to API_KEY.
  • model::String: The model to use for generating embeddings. Defaults to MODEL_EMBEDDING.
  • http_kwargs::NamedTuple: Additional keyword arguments for the HTTP request. Defaults to (retry_non_idempotent = true, retries = 5, readtimeout = 120).
  • api_kwargs::NamedTuple: Additional keyword arguments for the OpenAI API. Defaults to an empty NamedTuple.
  • kwargs...: Additional keyword arguments.

Returns

  • msg: A DataMessage object containing the embeddings, status, token count, and elapsed time.

Example

msg = aiembed("Hello World")
-msg.content # 1536-element JSON3.Array{Float64...

We can embed multiple strings at once and they will be hcat into a matrix (ie, each column corresponds to one string)

msg = aiembed(["Hello World", "How are you?"])
-msg.content # 1536×2 Matrix{Float64}:

If you plan to calculate the cosine distance between embeddings, you can normalize them first:

using LinearAlgebra
-msg = aiembed(["embed me", "and me too"], LinearAlgebra.normalize)
-
-# calculate cosine distance between the two normalized embeddings as a simple dot product
-msg.content' * msg.content[:, 1] # [1.0, 0.787]
source
PromptingTools.aiextract — Method
aiextract([prompt_schema::AbstractOpenAISchema,] prompt::ALLOWED_PROMPT_TYPE; 
-return_type::Type,
-verbose::Bool = true,
-    model::String = MODEL_CHAT,
-    http_kwargs::NamedTuple = (;
-        retry_non_idempotent = true,
-        retries = 5,
-        readtimeout = 120), api_kwargs::NamedTuple = NamedTuple(),
-    kwargs...)

Extract required information (defined by a struct return_type) from the provided prompt by leveraging OpenAI function calling mode.

This is a perfect solution for extracting structured information from text (eg, extract organization names in news articles, etc.)

It's effectively a light wrapper around aigenerate call, which requires additional keyword argument return_type to be provided and will enforce the model outputs to adhere to it.

Arguments

  • prompt_schema: An optional object to specify which prompt template should be applied (Default to PROMPT_SCHEMA = OpenAISchema)
  • prompt: Can be a string representing the prompt for the AI conversation, a UserMessage, a vector of AbstractMessage or an AITemplate
  • return_type: A struct TYPE representing the the information we want to extract. Do not provide a struct instance, only the type. If the struct has a docstring, it will be provided to the model as well. It's used to enforce structured model outputs or provide more information.
  • verbose: A boolean indicating whether to print additional information.
  • api_key: A string representing the API key for accessing the OpenAI API.
  • model: A string representing the model to use for generating the response. Can be an alias corresponding to a model ID defined in MODEL_ALIASES.
  • http_kwargs: A named tuple of HTTP keyword arguments.
  • api_kwargs: A named tuple of API keyword arguments.
  • kwargs: Prompt variables to be used to fill the prompt/template

Returns

  • msg: An DataMessage object representing the extracted data, including the content, status, tokens, and elapsed time. Use msg.content to access the extracted data.

See also: function_call_signature, MaybeExtract, aigenerate

Example

Do you want to extract some specific measurements from a text like age, weight and height? You need to define the information you need as a struct (return_type):

"Person's age, height, and weight."
-struct MyMeasurement
-    age::Int # required
-    height::Union{Int,Nothing} # optional
-    weight::Union{Nothing,Float64} # optional
-end
-msg = aiextract("James is 30, weighs 80kg. He's 180cm tall."; return_type=MyMeasurement)
-# [ Info: Tokens: 129 @ Cost: $0.0002 in 1.0 seconds
-# PromptingTools.DataMessage(MyMeasurement)
-msg.content
-# MyMeasurement(30, 180, 80.0)

The fields that allow Nothing are marked as optional in the schema:

msg = aiextract("James is 30."; return_type=MyMeasurement)
-# MyMeasurement(30, nothing, nothing)

If there are multiple items you want to extract, define a wrapper struct to get a Vector of MyMeasurement:

struct MyMeasurementWrapper
-    measurements::Vector{MyMeasurement}
-end
-
-msg = aiextract("James is 30, weighs 80kg. He's 180cm tall. Then Jack is 19 but really tall - over 190!"; return_type=ManyMeasurements)
-
-msg.content.measurements
-# 2-element Vector{MyMeasurement}:
-#  MyMeasurement(30, 180, 80.0)
-#  MyMeasurement(19, 190, nothing)

Or if you want your extraction to fail gracefully when data isn't found, use MaybeExtract{T} wrapper (this trick is inspired by the Instructor package!):

using PromptingTools: MaybeExtract
-
-type = MaybeExtract{MyMeasurement}
-# Effectively the same as:
-# struct MaybeExtract{T}
-#     result::Union{T, Nothing} // The result of the extraction
-#     error::Bool // true if a result is found, false otherwise
-#     message::Union{Nothing, String} // Only present if no result is found, should be short and concise
-# end
-
-# If LLM extraction fails, it will return a Dict with `error` and `message` fields instead of the result!
-msg = aiextract("Extract measurements from the text: I am giraffe", type)
-msg.content
-# MaybeExtract{MyMeasurement}(nothing, true, "I'm sorry, but I can only assist with human measurements.")

That way, you can handle the error gracefully and get a reason why extraction failed (in msg.content.message).

Note that the error message refers to a giraffe not being a human, because in our MyMeasurement docstring, we said that it's for people!

source
PromptingTools.aigenerate — Method
aigenerate(prompt_schema::AbstractOllamaManagedSchema, prompt::ALLOWED_PROMPT_TYPE; verbose::Bool = true,
-    model::String = MODEL_CHAT,
-    http_kwargs::NamedTuple = NamedTuple(), api_kwargs::NamedTuple = NamedTuple(),
-    kwargs...)

Generate an AI response based on a given prompt using the OpenAI API.

Arguments

  • prompt_schema: An optional object to specify which prompt template should be applied (Default to PROMPT_SCHEMA = OpenAISchema not AbstractManagedSchema)
  • prompt: Can be a string representing the prompt for the AI conversation, a UserMessage, a vector of AbstractMessage or an AITemplate
  • verbose: A boolean indicating whether to print additional information.
  • api_key: Provided for interface consistency. Not needed for locally hosted Ollama.
  • model: A string representing the model to use for generating the response. Can be an alias corresponding to a model ID defined in MODEL_ALIASES.
  • http_kwargs::NamedTuple: Additional keyword arguments for the HTTP request. Defaults to empty NamedTuple.
  • api_kwargs::NamedTuple: Additional keyword arguments for the Ollama API. Defaults to an empty NamedTuple.
  • kwargs: Prompt variables to be used to fill the prompt/template

Returns

  • msg: An AIMessage object representing the generated AI message, including the content, status, tokens, and elapsed time.

Use msg.content to access the extracted string.

See also: ai_str, aai_str, aiembed

Example

Simple hello world to test the API:

const PT = PromptingTools
-schema = PT.OllamaManagedSchema() # We need to explicit if we want Ollama, OpenAISchema is the default
-
-msg = aigenerate(schema, "Say hi!"; model="openhermes2.5-mistral")
-# [ Info: Tokens: 69 in 0.9 seconds
-# AIMessage("Hello! How can I assist you today?")

msg is an AIMessage object. Access the generated string via content property:

typeof(msg) # AIMessage{SubString{String}}
-propertynames(msg) # (:content, :status, :tokens, :elapsed
-msg.content # "Hello! How can I assist you today?"

Note: We need to be explicit about the schema we want to use. If we don't, it will default to OpenAISchema (=PT.DEFAULT_SCHEMA) ___ You can use string interpolation:

const PT = PromptingTools
-schema = PT.OllamaManagedSchema()
-a = 1
-msg=aigenerate(schema, "What is `$a+$a`?"; model="openhermes2.5-mistral")
-msg.content # "The result of `1+1` is `2`."

___ You can provide the whole conversation or more intricate prompts as a Vector{AbstractMessage}:

const PT = PromptingTools
-schema = PT.OllamaManagedSchema()
-
-conversation = [
-    PT.SystemMessage("You're master Yoda from Star Wars trying to help the user become a Yedi."),
-    PT.UserMessage("I have feelings for my iPhone. What should I do?")]
-
-msg = aigenerate(schema, conversation; model="openhermes2.5-mistral")
-# [ Info: Tokens: 111 in 2.1 seconds
-# AIMessage("Strong the attachment is, it leads to suffering it may. Focus on the force within you must, ...<continues>")

Note: Managed Ollama currently supports at most 1 User Message and 1 System Message given the API limitations. If you want more, you need to use the ChatMLSchema.

source
PromptingTools.aigenerate — Method
aigenerate([prompt_schema::AbstractOpenAISchema,] prompt::ALLOWED_PROMPT_TYPE; verbose::Bool = true,
-    model::String = MODEL_CHAT,
-    http_kwargs::NamedTuple = (;
-        retry_non_idempotent = true,
-        retries = 5,
-        readtimeout = 120), api_kwargs::NamedTuple = NamedTuple(),
-    kwargs...)

Generate an AI response based on a given prompt using the OpenAI API.

Arguments

  • prompt_schema: An optional object to specify which prompt template should be applied (Default to PROMPT_SCHEMA = OpenAISchema)
  • prompt: Can be a string representing the prompt for the AI conversation, a UserMessage, a vector of AbstractMessage or an AITemplate
  • verbose: A boolean indicating whether to print additional information.
  • api_key: A string representing the API key for accessing the OpenAI API.
  • model: A string representing the model to use for generating the response. Can be an alias corresponding to a model ID defined in MODEL_ALIASES.
  • http_kwargs: A named tuple of HTTP keyword arguments.
  • api_kwargs: A named tuple of API keyword arguments.
  • kwargs: Prompt variables to be used to fill the prompt/template

Returns

  • msg: An AIMessage object representing the generated AI message, including the content, status, tokens, and elapsed time.

Use msg.content to access the extracted string.

See also: ai_str, aai_str, aiembed, aiclassify, aiextract, aiscan

Example

Simple hello world to test the API:

result = aigenerate("Say Hi!")
-# [ Info: Tokens: 29 @ Cost: $0.0 in 1.0 seconds
-# AIMessage("Hello! How can I assist you today?")

result is an AIMessage object. Access the generated string via content property:

typeof(result) # AIMessage{SubString{String}}
-propertynames(result) # (:content, :status, :tokens, :elapsed
-result.content # "Hello! How can I assist you today?"

___ You can use string interpolation:

a = 1
-msg=aigenerate("What is `$a+$a`?")
-msg.content # "The sum of `1+1` is `2`."

___ You can provide the whole conversation or more intricate prompts as a Vector{AbstractMessage}:

const PT = PromptingTools
-
-conversation = [
-    PT.SystemMessage("You're master Yoda from Star Wars trying to help the user become a Yedi."),
-    PT.UserMessage("I have feelings for my iPhone. What should I do?")]
-msg=aigenerate(conversation)
-# AIMessage("Ah, strong feelings you have for your iPhone. A Jedi's path, this is not... <continues>")
source
PromptingTools.aiscan — Method

aiscan([promptschema::AbstractOpenAISchema,] prompt::ALLOWEDPROMPTTYPE; imageurl::Union{Nothing, AbstractString, Vector{<:AbstractString}} = nothing, imagepath::Union{Nothing, AbstractString, Vector{<:AbstractString}} = nothing, imagedetail::AbstractString = "auto", attachtolatest::Bool = true, verbose::Bool = true, model::String = MODELCHAT, httpkwargs::NamedTuple = (; retrynonidempotent = true, retries = 5, readtimeout = 120), apikwargs::NamedTuple = = (; maxtokens = 2500), kwargs...)

Scans the provided image (image_url or image_path) with the goal provided in the prompt.

Can be used for many multi-modal tasks, such as: OCR (transcribe text in the image), image captioning, image classification, etc.

It's effectively a light wrapper around aigenerate call, which uses additional keyword arguments image_url, image_path, image_detail to be provided. At least one image source (url or path) must be provided.

Arguments

  • prompt_schema: An optional object to specify which prompt template should be applied (Default to PROMPT_SCHEMA = OpenAISchema)
  • prompt: Can be a string representing the prompt for the AI conversation, a UserMessage, a vector of AbstractMessage or an AITemplate
  • image_url: A string or vector of strings representing the URL(s) of the image(s) to scan.
  • image_path: A string or vector of strings representing the path(s) of the image(s) to scan.
  • image_detail: A string representing the level of detail to include for images. Can be "auto", "high", or "low". See OpenAI Vision Guide for more details.
  • attach_to_latest: A boolean how to handle if a conversation with multiple UserMessage is provided. When true, the images are attached to the latest UserMessage.
  • verbose: A boolean indicating whether to print additional information.
  • api_key: A string representing the API key for accessing the OpenAI API.
  • model: A string representing the model to use for generating the response. Can be an alias corresponding to a model ID defined in MODEL_ALIASES.
  • http_kwargs: A named tuple of HTTP keyword arguments.
  • api_kwargs: A named tuple of API keyword arguments.
  • kwargs: Prompt variables to be used to fill the prompt/template

Returns

  • msg: An AIMessage object representing the generated AI message, including the content, status, tokens, and elapsed time.

Use msg.content to access the extracted string.

See also: ai_str, aai_str, aigenerate, aiembed, aiclassify, aiextract

Notes

  • All examples below use model "gpt4v", which is an alias for model ID "gpt-4-vision-preview"
  • max_tokens in the api_kwargs is preset to 2500, otherwise OpenAI enforces a default of only a few hundred tokens (~300). If your output is truncated, increase this value

Example

Describe the provided image:

msg = aiscan("Describe the image"; image_path="julia.png", model="gpt4v")
-# [ Info: Tokens: 1141 @ Cost: $0.0117 in 2.2 seconds
-# AIMessage("The image shows a logo consisting of the word "julia" written in lowercase")

You can provide multiple images at once as a vector and ask for "low" level of detail (cheaper):

msg = aiscan("Describe the image"; image_path=["julia.png","python.png"], image_detail="low", model="gpt4v")

You can use this function as a nice and quick OCR (transcribe text in the image) with a template :OCRTask. Let's transcribe some SQL code from a screenshot (no more re-typing!):

# Screenshot of some SQL code
-image_url = "https://www.sqlservercentral.com/wp-content/uploads/legacy/8755f69180b7ac7ee76a69ae68ec36872a116ad4/24622.png"
-msg = aiscan(:OCRTask; image_url, model="gpt4v", task="Transcribe the SQL code in the image.", api_kwargs=(; max_tokens=2500))
-
-# [ Info: Tokens: 362 @ Cost: $0.0045 in 2.5 seconds
-# AIMessage("```sql
-# update Orders <continue>
-
-# You can add syntax highlighting of the outputs via Markdown
-using Markdown
-msg.content |> Markdown.parse

Notice that we enforce max_tokens = 2500. That's because OpenAI seems to default to ~300 tokens, which provides incomplete outputs. Hence, we set this value to 2500 as a default. If you still get truncated outputs, increase this value.

source
PromptingTools.aitemplates — Function
aitemplates

Find easily the most suitable templates for your use case.

You can search by:

  • query::Symbol which looks look only for partial matches in the template name
  • query::AbstractString which looks for partial matches in the template name or description
  • query::Regex which looks for matches in the template name, description or any of the message previews

Keyword Arguments

  • limit::Int limits the number of returned templates (Defaults to 10)

Examples

Find available templates with aitemplates:

tmps = aitemplates("JuliaExpertAsk")
-# Will surface one specific template
-# 1-element Vector{AITemplateMetadata}:
-# PromptingTools.AITemplateMetadata
-#   name: Symbol JuliaExpertAsk
-#   description: String "For asking questions about Julia language. Placeholders: `ask`"
-#   version: String "1"
-#   wordcount: Int64 237
-#   variables: Array{Symbol}((1,))
-#   system_preview: String "You are a world-class Julia language programmer with the knowledge of the latest syntax. Your commun"
-#   user_preview: String "# Question
-
-{{ask}}"
-#   source: String ""

The above gives you a good idea of what the template is about, what placeholders are available, and how much it would cost to use it (=wordcount).

Search for all Julia-related templates:

tmps = aitemplates("Julia")
-# 2-element Vector{AITemplateMetadata}... -> more to come later!

If you are on VSCode, you can leverage nice tabular display with vscodedisplay:

using DataFrames
-tmps = aitemplates("Julia") |> DataFrame |> vscodedisplay

I have my selected template, how do I use it? Just use the "name" in aigenerate or aiclassify like you see in the first example!

source
PromptingTools.aitemplates — Method

Find the top-limit templates whose name or description fields partially match the query_key::String in TEMPLATE_METADATA.

source
PromptingTools.aitemplates — Method

Find the top-limit templates where provided query_key::Regex matches either of name, description or previews or User or System messages in TEMPLATE_METADATA.

source
PromptingTools.function_call_signature — Method
function_call_signature(datastructtype::Struct; max_description_length::Int = 100)

Extract the argument names, types and docstrings from a struct to create the function call signature in JSON schema.

You must provide a Struct type (not an instance of it) with some fields.

Note: Fairly experimental, but works for combination of structs, arrays, strings and singletons.

Tips

  • You can improve the quality of the extraction by writing a helpful docstring for your struct (or any nested struct). It will be provided as a description.

You can even include comments/descriptions about the individual fields.

  • All fields are assumed to be required, unless you allow null values (eg, ::Union{Nothing, Int}). Fields with Nothing will be treated as optional.
  • Missing values are ignored (eg, ::Union{Missing, Int} will be treated as Int). It's for broader compatibility and we cannot deserialize it as easily as Nothing.

Example

Do you want to extract some specific measurements from a text like age, weight and height? You need to define the information you need as a struct (return_type):

struct MyMeasurement
-    age::Int
-    height::Union{Int,Nothing}
-    weight::Union{Nothing,Float64}
-end
-signature = function_call_signature(MyMeasurement)
-#
-# Dict{String, Any} with 3 entries:
-#   "name"        => "MyMeasurement_extractor"
-#   "parameters"  => Dict{String, Any}("properties"=>Dict{String, Any}("height"=>Dict{String, Any}("type"=>"integer"), "weight"=>Dic…
-#   "description" => "Represents person's age, height, and weight
-"

You can see that only the field age does not allow null values, hence, it's "required". While height and weight are optional.

signature["parameters"]["required"]
-# ["age"]

If there are multiple items you want to extract, define a wrapper struct to get a Vector of MyMeasurement:

struct MyMeasurementWrapper
-    measurements::Vector{MyMeasurement}
-end
-
-Or if you want your extraction to fail gracefully when data isn't found, use `MaybeExtract{T}` wrapper (inspired by Instructor package!):

using PromptingTools: MaybeExtract

type = MaybeExtract{MyMeasurement}

Effectively the same as:

struct MaybeExtract{T}

result::Union{T, Nothing}

error::Bool // true if a result is found, false otherwise

message::Union{Nothing, String} // Only present if no result is found, should be short and concise

end

If LLM extraction fails, it will return a Dict with error and message fields instead of the result!

msg = aiextract("Extract measurements from the text: I am giraffe", type)

Dict{Symbol, Any} with 2 entries:

:message => "Sorry, this feature is only available for humans."

:error => true

``` That way, you can handle the error gracefully and get a reason why extraction failed.

source
PromptingTools.load_templates! — Function
load_templates!(; remove_templates::Bool=true)

Loads templates from folder templates/ in the package root and stores them in TEMPLATE_STORE and TEMPLATE_METADATA.

Note: Automatically removes any existing templates and metadata from TEMPLATE_STORE and TEMPLATE_METADATA if remove_templates=true.

source
PromptingTools.ollama_api — Method
ollama_api(prompt_schema::AbstractOllamaManagedSchema, prompt::AbstractString,
-    system::Union{Nothing, AbstractString} = nothing,
-    endpoint::String = "generate";
-    model::String = "llama2", http_kwargs::NamedTuple = NamedTuple(),
-    stream::Bool = false,
-    url::String = "localhost", port::Int = 11434,
-    kwargs...)

Simple wrapper for a call to Ollama API.

Keyword Arguments

  • prompt_schema: Defines which prompt template should be applied.
  • prompt: Can be a string representing the prompt for the AI conversation, a UserMessage, a vector of AbstractMessage
  • system: An optional string representing the system message for the AI conversation. If not provided, a default message will be used.
  • endpoint: The API endpoint to call, only "generate" and "embeddings" are currently supported. Defaults to "generate".
  • model: A string representing the model to use for generating the response. Can be an alias corresponding to a model ID defined in MODEL_ALIASES.
  • http_kwargs::NamedTuple: Additional keyword arguments for the HTTP request. Defaults to empty NamedTuple.
  • stream: A boolean indicating whether to stream the response. Defaults to false.
  • url: The URL of the Ollama API. Defaults to "localhost".
  • port: The port of the Ollama API. Defaults to 11434.
  • kwargs: Prompt variables to be used to fill the prompt/template
source
PromptingTools.render — Method
render(schema::AbstractOllamaManagedSchema,
-    messages::Vector{<:AbstractMessage};
-    kwargs...)

Builds a history of the conversation to provide the prompt to the API. All unspecified kwargs are passed as replacements such that {{key}}=>value in the template.

Note: Due to its "managed" nature, at most 2 messages can be provided (system and prompt inputs in the API).

source
PromptingTools.render — Method
render(schema::AbstractOpenAISchema,
-    messages::Vector{<:AbstractMessage};
-    image_detail::AbstractString = "auto",
-    kwargs...)

Builds a history of the conversation to provide the prompt to the API. All unspecified kwargs are passed as replacements such that {{key}}=>value in the template.

Arguments

  • image_detail: Only for UserMessageWithImages. It represents the level of detail to include for images. Can be "auto", "high", or "low".
source
PromptingTools.@aai_str — Macro
aai"user_prompt"[model_alias] -> AIMessage

Asynchronous version of @ai_str macro, which will log the result once it's ready.

Example

Send asynchronous request to GPT-4, so we don't have to wait for the response: Very practical with slow models, so you can keep working in the meantime.

```julia m = aai"Say Hi!"gpt4;

...with some delay...

[ Info: Tokens: 29 @ Cost: 0.0011 in 2.7 seconds

[ Info: AIMessage> Hello! How can I assist you today?

source
PromptingTools.@ai_str — Macro
ai"user_prompt"[model_alias] -> AIMessage

The ai"" string macro generates an AI response to a given prompt by using aigenerate under the hood.

Arguments

  • user_prompt (String): The input prompt for the AI model.
  • model_alias (optional, any): Provide model alias of the AI model (see MODEL_ALIASES).

Returns

AIMessage corresponding to the input prompt.

Example

result = ai"Hello, how are you?"
-# AIMessage("Hello! I'm an AI assistant, so I don't have feelings, but I'm here to help you. How can I assist you today?")

If you want to interpolate some variables or additional context, simply use string interpolation:

a=1
-result = ai"What is `$a+$a`?"
-# AIMessage("The sum of `1+1` is `2`.")

If you want to use a different model, eg, GPT-4, you can provide its alias as a flag:

result = ai"What is `1.23 * 100 + 1`?"gpt4
-# AIMessage("The answer is 124.")
source
+Home · PromptingTools.jl
GitHub

PromptingTools

Documentation for PromptingTools.

Streamline your life using PromptingTools.jl, the Julia package that simplifies interacting with large language models.

PromptingTools.jl is not meant for building large-scale systems. It's meant to be the go-to tool in your global environment that will save you 20 minutes every day!

Why PromptingTools.jl?

Prompt engineering is neither fast nor easy. Moreover, different models and their fine-tunes might require different prompt formats and tricks, or perhaps the information you work with requires special models to be used. PromptingTools.jl is meant to unify the prompts for different backends and make the common tasks (like templated prompts) as simple as possible.

Some features:

  • aigenerate Function: Simplify prompt templates with handlebars (eg, {{variable}}) and keyword arguments
  • @ai_str String Macro: Save keystrokes with a string macro for simple prompts
  • Easy to Remember: All exported functions start with ai... for better discoverability
  • Light Wraper Types: Benefit from Julia's multiple dispatch by having AI outputs wrapped in specific types
  • Minimal Dependencies: Enjoy an easy addition to your global environment with very light dependencies
  • No Context Switching: Access cutting-edge LLMs with no context switching and minimum extra keystrokes directly in your REPL

First Steps

To get started, see the Getting Started section.

diff --git a/dev/reference/index.html b/dev/reference/index.html new file mode 100644 index 000000000..d08b492a2 --- /dev/null +++ b/dev/reference/index.html @@ -0,0 +1,213 @@ + +Reference · PromptingTools.jl
GitHub

Reference

PromptingTools.AITemplate — Type
AITemplate

AITemplate is a template for a conversation prompt. This type is merely a container for the template name, which is resolved into a set of messages (=prompt) by render.

Naming Convention

  • Template names should be in CamelCase
  • Follow the format <Persona>...<Variable>... where possible, eg, JudgeIsItTrue, ``
    • Starting with the Persona (=System prompt), eg, Judge = persona is meant to judge some provided information
    • Variable to be filled in with context, eg, It = placeholder it
    • Ending with the variable name is helpful, eg, JuliaExpertTask for a persona to be an expert in Julia language and task is the placeholder name
  • Ideally, the template name should be self-explanatory, eg, JudgeIsItTrue = persona is meant to judge some provided information where it is true or false

Examples

Save time by re-using pre-made templates, just fill in the placeholders with the keyword arguments:

msg = aigenerate(:JuliaExpertAsk; ask = "How do I add packages?")

The above is equivalent to a more verbose version that explicitly uses the dispatch on AITemplate:

msg = aigenerate(AITemplate(:JuliaExpertAsk); ask = "How do I add packages?")

Find available templates with aitemplates:

tmps = aitemplates("JuliaExpertAsk")
+# Will surface one specific template
+# 1-element Vector{AITemplateMetadata}:
+# PromptingTools.AITemplateMetadata
+#   name: Symbol JuliaExpertAsk
+#   description: String "For asking questions about Julia language. Placeholders: `ask`"
+#   version: String "1"
+#   wordcount: Int64 237
+#   variables: Array{Symbol}((1,))
+#   system_preview: String "You are a world-class Julia language programmer with the knowledge of the latest syntax. Your commun"
+#   user_preview: String "# Question
+
+{{ask}}"
+#   source: String ""

The above gives you a good idea of what the template is about, what placeholders are available, and how much it would cost to use it (=wordcount).

Search for all Julia-related templates:

tmps = aitemplates("Julia")
+# 2-element Vector{AITemplateMetadata}... -> more to come later!

If you are on VSCode, you can leverage nice tabular display with vscodedisplay:

using DataFrames
+tmps = aitemplates("Julia") |> DataFrame |> vscodedisplay

I have my selected template, how do I use it? Just use the "name" in aigenerate or aiclassify like you see in the first example!

You can inspect any template by "rendering" it (this is what the LLM will see):

julia> AITemplate(:JudgeIsItTrue) |> PromptingTools.render

See also: save_template, load_template, load_templates! for more advanced use cases (and the corresponding script in examples/ folder)

source
PromptingTools.ChatMLSchema — Type

ChatMLSchema is used by many open-source chatbots, by OpenAI models (under the hood) and by several models and inferfaces (eg, Ollama, vLLM)

You can explore it on tiktokenizer

It uses the following conversation structure:

<im_start>system
+...<im_end>
+<|im_start|>user
+...<|im_end|>
+<|im_start|>assistant
+...<|im_end|>
source
PromptingTools.MaybeExtract — Type

Extract a result from the provided data, if any, otherwise set the error and message fields.

Arguments

  • error::Bool: true if a result is found, false otherwise.
  • message::String: Only present if no result is found, should be short and concise.
source
PromptingTools.OllamaManagedSchema — Type

Ollama by default manages different models and their associated prompt schemas when you pass system_prompt and prompt fields to the API.

Warning: It works only for 1 system message and 1 user message, so anything more than that has to be rejected.

If you need to pass more messagese / longer conversational history, you can use define the model-specific schema directly and pass your Ollama requests with raw=true, which disables and templating and schema management by Ollama.

source
PromptingTools.OpenAISchema — Type

OpenAISchema is the default schema for OpenAI models.

It uses the following conversation template:

[Dict(role="system",content="..."),Dict(role="user",content="..."),Dict(role="assistant",content="...")]

It's recommended to separate sections in your prompt with markdown headers (e.g. `##Answer

`).

source
PromptingTools.aiclassify — Method
aiclassify(prompt_schema::AbstractOpenAISchema, prompt::ALLOWED_PROMPT_TYPE;
+api_kwargs::NamedTuple = (logit_bias = Dict(837 => 100, 905 => 100, 9987 => 100),
+    max_tokens = 1, temperature = 0),
+kwargs...)

Classifies the given prompt/statement as true/false/unknown.

Note: this is a very simple classifier, it is not meant to be used in production. Credit goes to AAAzzam.

It uses Logit bias trick and limits the output to 1 token to force the model to output only true/false/unknown.

Output tokens used (via api_kwargs):

  • 837: ' true'
  • 905: ' false'
  • 9987: ' unknown'

Arguments

  • prompt_schema::AbstractOpenAISchema: The schema for the prompt.
  • prompt: The prompt/statement to classify if it's a String. If it's a Symbol, it is expanded as a template via render(schema,template).

Example

aiclassify("Is two plus two four?") # true
+aiclassify("Is two plus three a vegetable on Mars?") # false

aiclassify returns only true/false/unknown. It's easy to get the proper Bool output type out with tryparse, eg,

tryparse(Bool, aiclassify("Is two plus two four?")) isa Bool # true

Output of type Nothing marks that the model couldn't classify the statement as true/false.

Ideally, we would like to re-use some helpful system prompt to get more accurate responses. For this reason we have templates, eg, :JudgeIsItTrue. By specifying the template, we can provide our statement as the expected variable (it in this case) See that the model now correctly classifies the statement as "unknown".

aiclassify(:JudgeIsItTrue; it = "Is two plus three a vegetable on Mars?") # unknown

For better results, use higher quality models like gpt4, eg, 

aiclassify(:JudgeIsItTrue;
+    it = "If I had two apples and I got three more, I have five apples now.",
+    model = "gpt4") # true
source
PromptingTools.aiembed — Method
aiembed(prompt_schema::AbstractOllamaManagedSchema,
+        doc_or_docs::Union{AbstractString, Vector{<:AbstractString}},
+        postprocess::F = identity;
+        verbose::Bool = true,
+        api_key::String = API_KEY,
+        model::String = MODEL_EMBEDDING,
+        http_kwargs::NamedTuple = (retry_non_idempotent = true,
+                                   retries = 5,
+                                   readtimeout = 120),
+        api_kwargs::NamedTuple = NamedTuple(),
+        kwargs...) where {F <: Function}

The aiembed function generates embeddings for the given input using a specified model and returns a message object containing the embeddings, status, token count, and elapsed time.

Arguments

  • prompt_schema::AbstractOllamaManagedSchema: The schema for the prompt.
  • doc_or_docs::Union{AbstractString, Vector{<:AbstractString}}: The document or list of documents to generate embeddings for. The list of documents is processed sequentially, so users should consider implementing an async version with with Threads.@spawn
  • postprocess::F: The post-processing function to apply to each embedding. Defaults to the identity function, but could be LinearAlgebra.normalize.
  • verbose::Bool: A flag indicating whether to print verbose information. Defaults to true.
  • api_key::String: The API key to use for the OpenAI API. Defaults to API_KEY.
  • model::String: The model to use for generating embeddings. Defaults to MODEL_EMBEDDING.
  • http_kwargs::NamedTuple: Additional keyword arguments for the HTTP request. Defaults to empty NamedTuple.
  • api_kwargs::NamedTuple: Additional keyword arguments for the Ollama API. Defaults to an empty NamedTuple.
  • kwargs: Prompt variables to be used to fill the prompt/template

Returns

  • msg: A DataMessage object containing the embeddings, status, token count, and elapsed time.

Note: Ollama API currently does not return the token count, so it's set to (0,0)

Example

const PT = PromptingTools
+schema = PT.OllamaManagedSchema()
+
+msg = aiembed(schema, "Hello World"; model="openhermes2.5-mistral")
+msg.content # 4096-element JSON3.Array{Float64...

We can embed multiple strings at once and they will be hcat into a matrix (ie, each column corresponds to one string)

const PT = PromptingTools
+schema = PT.OllamaManagedSchema()
+
+msg = aiembed(schema, ["Hello World", "How are you?"]; model="openhermes2.5-mistral")
+msg.content # 4096×2 Matrix{Float64}:

If you plan to calculate the cosine distance between embeddings, you can normalize them first:

const PT = PromptingTools
+using LinearAlgebra
+schema = PT.OllamaManagedSchema()
+
+msg = aiembed(schema, ["embed me", "and me too"], LinearAlgebra.normalize; model="openhermes2.5-mistral")
+
+# calculate cosine distance between the two normalized embeddings as a simple dot product
+msg.content' * msg.content[:, 1] # [1.0, 0.34]

Similarly, you can use the postprocess argument to materialize the data from JSON3.Object by using postprocess = copy

const PT = PromptingTools
+schema = PT.OllamaManagedSchema()
+
+msg = aiembed(schema, "Hello World", copy; model="openhermes2.5-mistral")
+msg.content # 4096-element Vector{Float64}
source
PromptingTools.aiembed — Method
aiembed(prompt_schema::AbstractOpenAISchema,
+        doc_or_docs::Union{AbstractString, Vector{<:AbstractString}},
+        postprocess::F = identity;
+        verbose::Bool = true,
+        api_key::String = API_KEY,
+        model::String = MODEL_EMBEDDING,
+        http_kwargs::NamedTuple = (retry_non_idempotent = true,
+                                   retries = 5,
+                                   readtimeout = 120),
+        api_kwargs::NamedTuple = NamedTuple(),
+        kwargs...) where {F <: Function}

The aiembed function generates embeddings for the given input using a specified model and returns a message object containing the embeddings, status, token count, and elapsed time.

Arguments

  • prompt_schema::AbstractOpenAISchema: The schema for the prompt.
  • doc_or_docs::Union{AbstractString, Vector{<:AbstractString}}: The document or list of documents to generate embeddings for.
  • postprocess::F: The post-processing function to apply to each embedding. Defaults to the identity function.
  • verbose::Bool: A flag indicating whether to print verbose information. Defaults to true.
  • api_key::String: The API key to use for the OpenAI API. Defaults to API_KEY.
  • model::String: The model to use for generating embeddings. Defaults to MODEL_EMBEDDING.
  • http_kwargs::NamedTuple: Additional keyword arguments for the HTTP request. Defaults to (retry_non_idempotent = true, retries = 5, readtimeout = 120).
  • api_kwargs::NamedTuple: Additional keyword arguments for the OpenAI API. Defaults to an empty NamedTuple.
  • kwargs...: Additional keyword arguments.

Returns

  • msg: A DataMessage object containing the embeddings, status, token count, and elapsed time.

Example

msg = aiembed("Hello World")
+msg.content # 1536-element JSON3.Array{Float64...

We can embed multiple strings at once and they will be hcat into a matrix (ie, each column corresponds to one string)

msg = aiembed(["Hello World", "How are you?"])
+msg.content # 1536×2 Matrix{Float64}:

If you plan to calculate the cosine distance between embeddings, you can normalize them first:

using LinearAlgebra
+msg = aiembed(["embed me", "and me too"], LinearAlgebra.normalize)
+
+# calculate cosine distance between the two normalized embeddings as a simple dot product
+msg.content' * msg.content[:, 1] # [1.0, 0.787]
source
PromptingTools.aiextract — Method
aiextract([prompt_schema::AbstractOpenAISchema,] prompt::ALLOWED_PROMPT_TYPE; 
+return_type::Type,
+verbose::Bool = true,
+    model::String = MODEL_CHAT,
+    http_kwargs::NamedTuple = (;
+        retry_non_idempotent = true,
+        retries = 5,
+        readtimeout = 120), api_kwargs::NamedTuple = NamedTuple(),
+    kwargs...)

Extract required information (defined by a struct return_type) from the provided prompt by leveraging OpenAI function calling mode.

This is a perfect solution for extracting structured information from text (eg, extract organization names in news articles, etc.)

It's effectively a light wrapper around aigenerate call, which requires additional keyword argument return_type to be provided and will enforce the model outputs to adhere to it.

Arguments

  • prompt_schema: An optional object to specify which prompt template should be applied (Default to PROMPT_SCHEMA = OpenAISchema)
  • prompt: Can be a string representing the prompt for the AI conversation, a UserMessage, a vector of AbstractMessage or an AITemplate
  • return_type: A struct TYPE representing the the information we want to extract. Do not provide a struct instance, only the type. If the struct has a docstring, it will be provided to the model as well. It's used to enforce structured model outputs or provide more information.
  • verbose: A boolean indicating whether to print additional information.
  • api_key: A string representing the API key for accessing the OpenAI API.
  • model: A string representing the model to use for generating the response. Can be an alias corresponding to a model ID defined in MODEL_ALIASES.
  • http_kwargs: A named tuple of HTTP keyword arguments.
  • api_kwargs: A named tuple of API keyword arguments.
  • kwargs: Prompt variables to be used to fill the prompt/template

Returns

  • msg: An DataMessage object representing the extracted data, including the content, status, tokens, and elapsed time. Use msg.content to access the extracted data.

See also: function_call_signature, MaybeExtract, aigenerate

Example

Do you want to extract some specific measurements from a text like age, weight and height? You need to define the information you need as a struct (return_type):

"Person's age, height, and weight."
+struct MyMeasurement
+    age::Int # required
+    height::Union{Int,Nothing} # optional
+    weight::Union{Nothing,Float64} # optional
+end
+msg = aiextract("James is 30, weighs 80kg. He's 180cm tall."; return_type=MyMeasurement)
+# [ Info: Tokens: 129 @ Cost: $0.0002 in 1.0 seconds
+# PromptingTools.DataMessage(MyMeasurement)
+msg.content
+# MyMeasurement(30, 180, 80.0)

The fields that allow Nothing are marked as optional in the schema:

msg = aiextract("James is 30."; return_type=MyMeasurement)
+# MyMeasurement(30, nothing, nothing)

If there are multiple items you want to extract, define a wrapper struct to get a Vector of MyMeasurement:

struct MyMeasurementWrapper
+    measurements::Vector{MyMeasurement}
+end
+
+msg = aiextract("James is 30, weighs 80kg. He's 180cm tall. Then Jack is 19 but really tall - over 190!"; return_type=ManyMeasurements)
+
+msg.content.measurements
+# 2-element Vector{MyMeasurement}:
+#  MyMeasurement(30, 180, 80.0)
+#  MyMeasurement(19, 190, nothing)

Or if you want your extraction to fail gracefully when data isn't found, use MaybeExtract{T} wrapper (this trick is inspired by the Instructor package!):

using PromptingTools: MaybeExtract
+
+type = MaybeExtract{MyMeasurement}
+# Effectively the same as:
+# struct MaybeExtract{T}
+#     result::Union{T, Nothing} // The result of the extraction
+#     error::Bool // true if a result is found, false otherwise
+#     message::Union{Nothing, String} // Only present if no result is found, should be short and concise
+# end
+
+# If LLM extraction fails, it will return a Dict with `error` and `message` fields instead of the result!
+msg = aiextract("Extract measurements from the text: I am giraffe", type)
+msg.content
+# MaybeExtract{MyMeasurement}(nothing, true, "I'm sorry, but I can only assist with human measurements.")

That way, you can handle the error gracefully and get a reason why extraction failed (in msg.content.message).

Note that the error message refers to a giraffe not being a human, because in our MyMeasurement docstring, we said that it's for people!

source
PromptingTools.aigenerate — Method
aigenerate(prompt_schema::AbstractOllamaManagedSchema, prompt::ALLOWED_PROMPT_TYPE; verbose::Bool = true,
+    model::String = MODEL_CHAT,
+    http_kwargs::NamedTuple = NamedTuple(), api_kwargs::NamedTuple = NamedTuple(),
+    kwargs...)

Generate an AI response based on a given prompt using the OpenAI API.

Arguments

  • prompt_schema: An optional object to specify which prompt template should be applied (Default to PROMPT_SCHEMA = OpenAISchema not AbstractManagedSchema)
  • prompt: Can be a string representing the prompt for the AI conversation, a UserMessage, a vector of AbstractMessage or an AITemplate
  • verbose: A boolean indicating whether to print additional information.
  • api_key: Provided for interface consistency. Not needed for locally hosted Ollama.
  • model: A string representing the model to use for generating the response. Can be an alias corresponding to a model ID defined in MODEL_ALIASES.
  • http_kwargs::NamedTuple: Additional keyword arguments for the HTTP request. Defaults to empty NamedTuple.
  • api_kwargs::NamedTuple: Additional keyword arguments for the Ollama API. Defaults to an empty NamedTuple.
  • kwargs: Prompt variables to be used to fill the prompt/template

Returns

  • msg: An AIMessage object representing the generated AI message, including the content, status, tokens, and elapsed time.

Use msg.content to access the extracted string.

See also: ai_str, aai_str, aiembed

Example

Simple hello world to test the API:

const PT = PromptingTools
+schema = PT.OllamaManagedSchema() # We need to explicit if we want Ollama, OpenAISchema is the default
+
+msg = aigenerate(schema, "Say hi!"; model="openhermes2.5-mistral")
+# [ Info: Tokens: 69 in 0.9 seconds
+# AIMessage("Hello! How can I assist you today?")

msg is an AIMessage object. Access the generated string via content property:

typeof(msg) # AIMessage{SubString{String}}
+propertynames(msg) # (:content, :status, :tokens, :elapsed
+msg.content # "Hello! How can I assist you today?"

Note: We need to be explicit about the schema we want to use. If we don't, it will default to OpenAISchema (=PT.DEFAULT_SCHEMA) ___ You can use string interpolation:

const PT = PromptingTools
+schema = PT.OllamaManagedSchema()
+a = 1
+msg=aigenerate(schema, "What is `$a+$a`?"; model="openhermes2.5-mistral")
+msg.content # "The result of `1+1` is `2`."

___ You can provide the whole conversation or more intricate prompts as a Vector{AbstractMessage}:

const PT = PromptingTools
+schema = PT.OllamaManagedSchema()
+
+conversation = [
+    PT.SystemMessage("You're master Yoda from Star Wars trying to help the user become a Yedi."),
+    PT.UserMessage("I have feelings for my iPhone. What should I do?")]
+
+msg = aigenerate(schema, conversation; model="openhermes2.5-mistral")
+# [ Info: Tokens: 111 in 2.1 seconds
+# AIMessage("Strong the attachment is, it leads to suffering it may. Focus on the force within you must, ...<continues>")

Note: Managed Ollama currently supports at most 1 User Message and 1 System Message given the API limitations. If you want more, you need to use the ChatMLSchema.

source
PromptingTools.aigenerate — Method
aigenerate([prompt_schema::AbstractOpenAISchema,] prompt::ALLOWED_PROMPT_TYPE; verbose::Bool = true,
+    model::String = MODEL_CHAT,
+    http_kwargs::NamedTuple = (;
+        retry_non_idempotent = true,
+        retries = 5,
+        readtimeout = 120), api_kwargs::NamedTuple = NamedTuple(),
+    kwargs...)

Generate an AI response based on a given prompt using the OpenAI API.

Arguments

  • prompt_schema: An optional object to specify which prompt template should be applied (Default to PROMPT_SCHEMA = OpenAISchema)
  • prompt: Can be a string representing the prompt for the AI conversation, a UserMessage, a vector of AbstractMessage or an AITemplate
  • verbose: A boolean indicating whether to print additional information.
  • api_key: A string representing the API key for accessing the OpenAI API.
  • model: A string representing the model to use for generating the response. Can be an alias corresponding to a model ID defined in MODEL_ALIASES.
  • http_kwargs: A named tuple of HTTP keyword arguments.
  • api_kwargs: A named tuple of API keyword arguments.
  • kwargs: Prompt variables to be used to fill the prompt/template

Returns

  • msg: An AIMessage object representing the generated AI message, including the content, status, tokens, and elapsed time.

Use msg.content to access the extracted string.

See also: ai_str, aai_str, aiembed, aiclassify, aiextract, aiscan

Example

Simple hello world to test the API:

result = aigenerate("Say Hi!")
+# [ Info: Tokens: 29 @ Cost: $0.0 in 1.0 seconds
+# AIMessage("Hello! How can I assist you today?")

result is an AIMessage object. Access the generated string via content property:

typeof(result) # AIMessage{SubString{String}}
+propertynames(result) # (:content, :status, :tokens, :elapsed
+result.content # "Hello! How can I assist you today?"

___ You can use string interpolation:

a = 1
+msg=aigenerate("What is `$a+$a`?")
+msg.content # "The sum of `1+1` is `2`."

___ You can provide the whole conversation or more intricate prompts as a Vector{AbstractMessage}:

const PT = PromptingTools
+
+conversation = [
+    PT.SystemMessage("You're master Yoda from Star Wars trying to help the user become a Yedi."),
+    PT.UserMessage("I have feelings for my iPhone. What should I do?")]
+msg=aigenerate(conversation)
+# AIMessage("Ah, strong feelings you have for your iPhone. A Jedi's path, this is not... <continues>")
source
PromptingTools.aiscan — Method

aiscan([promptschema::AbstractOpenAISchema,] prompt::ALLOWEDPROMPTTYPE; imageurl::Union{Nothing, AbstractString, Vector{<:AbstractString}} = nothing, imagepath::Union{Nothing, AbstractString, Vector{<:AbstractString}} = nothing, imagedetail::AbstractString = "auto", attachtolatest::Bool = true, verbose::Bool = true, model::String = MODELCHAT, httpkwargs::NamedTuple = (; retrynonidempotent = true, retries = 5, readtimeout = 120), apikwargs::NamedTuple = = (; maxtokens = 2500), kwargs...)

Scans the provided image (image_url or image_path) with the goal provided in the prompt.

Can be used for many multi-modal tasks, such as: OCR (transcribe text in the image), image captioning, image classification, etc.

It's effectively a light wrapper around aigenerate call, which uses additional keyword arguments image_url, image_path, image_detail to be provided. At least one image source (url or path) must be provided.

Arguments

  • prompt_schema: An optional object to specify which prompt template should be applied (Default to PROMPT_SCHEMA = OpenAISchema)
  • prompt: Can be a string representing the prompt for the AI conversation, a UserMessage, a vector of AbstractMessage or an AITemplate
  • image_url: A string or vector of strings representing the URL(s) of the image(s) to scan.
  • image_path: A string or vector of strings representing the path(s) of the image(s) to scan.
  • image_detail: A string representing the level of detail to include for images. Can be "auto", "high", or "low". See OpenAI Vision Guide for more details.
  • attach_to_latest: A boolean how to handle if a conversation with multiple UserMessage is provided. When true, the images are attached to the latest UserMessage.
  • verbose: A boolean indicating whether to print additional information.
  • api_key: A string representing the API key for accessing the OpenAI API.
  • model: A string representing the model to use for generating the response. Can be an alias corresponding to a model ID defined in MODEL_ALIASES.
  • http_kwargs: A named tuple of HTTP keyword arguments.
  • api_kwargs: A named tuple of API keyword arguments.
  • kwargs: Prompt variables to be used to fill the prompt/template

Returns

  • msg: An AIMessage object representing the generated AI message, including the content, status, tokens, and elapsed time.

Use msg.content to access the extracted string.

See also: ai_str, aai_str, aigenerate, aiembed, aiclassify, aiextract

Notes

  • All examples below use model "gpt4v", which is an alias for model ID "gpt-4-vision-preview"
  • max_tokens in the api_kwargs is preset to 2500, otherwise OpenAI enforces a default of only a few hundred tokens (~300). If your output is truncated, increase this value

Example

Describe the provided image:

msg = aiscan("Describe the image"; image_path="julia.png", model="gpt4v")
+# [ Info: Tokens: 1141 @ Cost: $0.0117 in 2.2 seconds
+# AIMessage("The image shows a logo consisting of the word "julia" written in lowercase")

You can provide multiple images at once as a vector and ask for "low" level of detail (cheaper):

msg = aiscan("Describe the image"; image_path=["julia.png","python.png"], image_detail="low", model="gpt4v")

You can use this function as a nice and quick OCR (transcribe text in the image) with a template :OCRTask. Let's transcribe some SQL code from a screenshot (no more re-typing!):

# Screenshot of some SQL code
+image_url = "https://www.sqlservercentral.com/wp-content/uploads/legacy/8755f69180b7ac7ee76a69ae68ec36872a116ad4/24622.png"
+msg = aiscan(:OCRTask; image_url, model="gpt4v", task="Transcribe the SQL code in the image.", api_kwargs=(; max_tokens=2500))
+
+# [ Info: Tokens: 362 @ Cost: $0.0045 in 2.5 seconds
+# AIMessage("```sql
+# update Orders <continue>
+
+# You can add syntax highlighting of the outputs via Markdown
+using Markdown
+msg.content |> Markdown.parse

Notice that we enforce max_tokens = 2500. That's because OpenAI seems to default to ~300 tokens, which provides incomplete outputs. Hence, we set this value to 2500 as a default. If you still get truncated outputs, increase this value.

source
PromptingTools.aitemplates — Function
aitemplates

Find easily the most suitable templates for your use case.

You can search by:

  • query::Symbol which looks look only for partial matches in the template name
  • query::AbstractString which looks for partial matches in the template name or description
  • query::Regex which looks for matches in the template name, description or any of the message previews

Keyword Arguments

  • limit::Int limits the number of returned templates (Defaults to 10)

Examples

Find available templates with aitemplates:

tmps = aitemplates("JuliaExpertAsk")
+# Will surface one specific template
+# 1-element Vector{AITemplateMetadata}:
+# PromptingTools.AITemplateMetadata
+#   name: Symbol JuliaExpertAsk
+#   description: String "For asking questions about Julia language. Placeholders: `ask`"
+#   version: String "1"
+#   wordcount: Int64 237
+#   variables: Array{Symbol}((1,))
+#   system_preview: String "You are a world-class Julia language programmer with the knowledge of the latest syntax. Your commun"
+#   user_preview: String "# Question
+
+{{ask}}"
+#   source: String ""

The above gives you a good idea of what the template is about, what placeholders are available, and how much it would cost to use it (=wordcount).

Search for all Julia-related templates:

tmps = aitemplates("Julia")
+# 2-element Vector{AITemplateMetadata}... -> more to come later!

If you are on VSCode, you can leverage nice tabular display with vscodedisplay:

using DataFrames
+tmps = aitemplates("Julia") |> DataFrame |> vscodedisplay

I have my selected template, how do I use it? Just use the "name" in aigenerate or aiclassify like you see in the first example!

source
PromptingTools.aitemplates — Method

Find the top-limit templates whose name or description fields partially match the query_key::String in TEMPLATE_METADATA.

source
PromptingTools.aitemplates — Method

Find the top-limit templates where provided query_key::Regex matches either of name, description or previews or User or System messages in TEMPLATE_METADATA.

source
PromptingTools.function_call_signature — Method
function_call_signature(datastructtype::Struct; max_description_length::Int = 100)

Extract the argument names, types and docstrings from a struct to create the function call signature in JSON schema.

You must provide a Struct type (not an instance of it) with some fields.

Note: Fairly experimental, but works for combination of structs, arrays, strings and singletons.

Tips

  • You can improve the quality of the extraction by writing a helpful docstring for your struct (or any nested struct). It will be provided as a description.

You can even include comments/descriptions about the individual fields.

  • All fields are assumed to be required, unless you allow null values (eg, ::Union{Nothing, Int}). Fields with Nothing will be treated as optional.
  • Missing values are ignored (eg, ::Union{Missing, Int} will be treated as Int). It's for broader compatibility and we cannot deserialize it as easily as Nothing.

Example

Do you want to extract some specific measurements from a text like age, weight and height? You need to define the information you need as a struct (return_type):

struct MyMeasurement
+    age::Int
+    height::Union{Int,Nothing}
+    weight::Union{Nothing,Float64}
+end
+signature = function_call_signature(MyMeasurement)
+#
+# Dict{String, Any} with 3 entries:
+#   "name"        => "MyMeasurement_extractor"
+#   "parameters"  => Dict{String, Any}("properties"=>Dict{String, Any}("height"=>Dict{String, Any}("type"=>"integer"), "weight"=>Dic…
+#   "description" => "Represents person's age, height, and weight
+"

You can see that only the field age does not allow null values, hence, it's "required". While height and weight are optional.

signature["parameters"]["required"]
+# ["age"]

If there are multiple items you want to extract, define a wrapper struct to get a Vector of MyMeasurement:

struct MyMeasurementWrapper
+    measurements::Vector{MyMeasurement}
+end
+
+Or if you want your extraction to fail gracefully when data isn't found, use `MaybeExtract{T}` wrapper (inspired by Instructor package!):

using PromptingTools: MaybeExtract

type = MaybeExtract{MyMeasurement}

Effectively the same as:

struct MaybeExtract{T}

result::Union{T, Nothing}

error::Bool // true if a result is found, false otherwise

message::Union{Nothing, String} // Only present if no result is found, should be short and concise

end

If LLM extraction fails, it will return a Dict with error and message fields instead of the result!

msg = aiextract("Extract measurements from the text: I am giraffe", type)

Dict{Symbol, Any} with 2 entries:

:message => "Sorry, this feature is only available for humans."

:error => true

``` That way, you can handle the error gracefully and get a reason why extraction failed.

source
PromptingTools.load_templates! — Function
load_templates!(; remove_templates::Bool=true)

Loads templates from folder templates/ in the package root and stores them in TEMPLATE_STORE and TEMPLATE_METADATA.

Note: Automatically removes any existing templates and metadata from TEMPLATE_STORE and TEMPLATE_METADATA if remove_templates=true.

source
PromptingTools.ollama_api — Method
ollama_api(prompt_schema::AbstractOllamaManagedSchema, prompt::AbstractString,
+    system::Union{Nothing, AbstractString} = nothing,
+    endpoint::String = "generate";
+    model::String = "llama2", http_kwargs::NamedTuple = NamedTuple(),
+    stream::Bool = false,
+    url::String = "localhost", port::Int = 11434,
+    kwargs...)

Simple wrapper for a call to Ollama API.

Keyword Arguments

  • prompt_schema: Defines which prompt template should be applied.
  • prompt: Can be a string representing the prompt for the AI conversation, a UserMessage, a vector of AbstractMessage
  • system: An optional string representing the system message for the AI conversation. If not provided, a default message will be used.
  • endpoint: The API endpoint to call, only "generate" and "embeddings" are currently supported. Defaults to "generate".
  • model: A string representing the model to use for generating the response. Can be an alias corresponding to a model ID defined in MODEL_ALIASES.
  • http_kwargs::NamedTuple: Additional keyword arguments for the HTTP request. Defaults to empty NamedTuple.
  • stream: A boolean indicating whether to stream the response. Defaults to false.
  • url: The URL of the Ollama API. Defaults to "localhost".
  • port: The port of the Ollama API. Defaults to 11434.
  • kwargs: Prompt variables to be used to fill the prompt/template
source
PromptingTools.render — Method
render(schema::AbstractOllamaManagedSchema,
+    messages::Vector{<:AbstractMessage};
+    kwargs...)

Builds a history of the conversation to provide the prompt to the API. All unspecified kwargs are passed as replacements such that {{key}}=>value in the template.

Note: Due to its "managed" nature, at most 2 messages can be provided (system and prompt inputs in the API).

source
PromptingTools.render — Method
render(schema::AbstractOpenAISchema,
+    messages::Vector{<:AbstractMessage};
+    image_detail::AbstractString = "auto",
+    kwargs...)

Builds a history of the conversation to provide the prompt to the API. All unspecified kwargs are passed as replacements such that {{key}}=>value in the template.

Arguments

  • image_detail: Only for UserMessageWithImages. It represents the level of detail to include for images. Can be "auto", "high", or "low".
source
PromptingTools.@aai_str — Macro
aai"user_prompt"[model_alias] -> AIMessage

Asynchronous version of @ai_str macro, which will log the result once it's ready.

Example

Send asynchronous request to GPT-4, so we don't have to wait for the response: Very practical with slow models, so you can keep working in the meantime.

```julia m = aai"Say Hi!"gpt4;

...with some delay...

[ Info: Tokens: 29 @ Cost: 0.0011 in 2.7 seconds

[ Info: AIMessage> Hello! How can I assist you today?

source
PromptingTools.@ai_str — Macro
ai"user_prompt"[model_alias] -> AIMessage

The ai"" string macro generates an AI response to a given prompt by using aigenerate under the hood.

Arguments

  • user_prompt (String): The input prompt for the AI model.
  • model_alias (optional, any): Provide model alias of the AI model (see MODEL_ALIASES).

Returns

AIMessage corresponding to the input prompt.

Example

result = ai"Hello, how are you?"
+# AIMessage("Hello! I'm an AI assistant, so I don't have feelings, but I'm here to help you. How can I assist you today?")

If you want to interpolate some variables or additional context, simply use string interpolation:

a=1
+result = ai"What is `$a+$a`?"
+# AIMessage("The sum of `1+1` is `2`.")

If you want to use a different model, eg, GPT-4, you can provide its alias as a flag:

result = ai"What is `1.23 * 100 + 1`?"gpt4
+# AIMessage("The answer is 124.")
source
diff --git a/dev/search_index.js b/dev/search_index.js index a785c5b47..90ef9e111 100644 --- a/dev/search_index.js +++ b/dev/search_index.js @@ -1,3 +1,3 @@ var documenterSearchIndex = {"docs": -[{"location":"","page":"Home","title":"Home","text":"CurrentModule = PromptingTools","category":"page"},{"location":"#PromptingTools","page":"Home","title":"PromptingTools","text":"","category":"section"},{"location":"","page":"Home","title":"Home","text":"Documentation for PromptingTools.","category":"page"},{"location":"","page":"Home","title":"Home","text":"","category":"page"},{"location":"","page":"Home","title":"Home","text":"Modules = [PromptingTools]","category":"page"},{"location":"#PromptingTools.AITemplate","page":"Home","title":"PromptingTools.AITemplate","text":"AITemplate\n\nAITemplate is a template for a conversation prompt. This type is merely a container for the template name, which is resolved into a set of messages (=prompt) by render.\n\nNaming Convention\n\nTemplate names should be in CamelCase\nFollow the format ...... where possible, eg, JudgeIsItTrue, ``\nStarting with the Persona (=System prompt), eg, Judge = persona is meant to judge some provided information\nVariable to be filled in with context, eg, It = placeholder it\nEnding with the variable name is helpful, eg, JuliaExpertTask for a persona to be an expert in Julia language and task is the placeholder name\nIdeally, the template name should be self-explanatory, eg, JudgeIsItTrue = persona is meant to judge some provided information where it is true or false\n\nExamples\n\nSave time by re-using pre-made templates, just fill in the placeholders with the keyword arguments:\n\nmsg = aigenerate(:JuliaExpertAsk; ask = \"How do I add packages?\")\n\nThe above is equivalent to a more verbose version that explicitly uses the dispatch on AITemplate:\n\nmsg = aigenerate(AITemplate(:JuliaExpertAsk); ask = \"How do I add packages?\")\n\nFind available templates with aitemplates:\n\ntmps = aitemplates(\"JuliaExpertAsk\")\n# Will surface one specific template\n# 1-element Vector{AITemplateMetadata}:\n# PromptingTools.AITemplateMetadata\n# name: Symbol JuliaExpertAsk\n# description: String \"For asking questions about Julia language. Placeholders: `ask`\"\n# version: String \"1\"\n# wordcount: Int64 237\n# variables: Array{Symbol}((1,))\n# system_preview: String \"You are a world-class Julia language programmer with the knowledge of the latest syntax. Your commun\"\n# user_preview: String \"# Question\n\n{{ask}}\"\n# source: String \"\"\n\nThe above gives you a good idea of what the template is about, what placeholders are available, and how much it would cost to use it (=wordcount).\n\nSearch for all Julia-related templates:\n\ntmps = aitemplates(\"Julia\")\n# 2-element Vector{AITemplateMetadata}... -> more to come later!\n\nIf you are on VSCode, you can leverage nice tabular display with vscodedisplay:\n\nusing DataFrames\ntmps = aitemplates(\"Julia\") |> DataFrame |> vscodedisplay\n\nI have my selected template, how do I use it? Just use the \"name\" in aigenerate or aiclassify like you see in the first example!\n\nYou can inspect any template by \"rendering\" it (this is what the LLM will see):\n\njulia> AITemplate(:JudgeIsItTrue) |> PromptingTools.render\n\nSee also: save_template, load_template, load_templates! for more advanced use cases (and the corresponding script in examples/ folder)\n\n\n\n\n\n","category":"type"},{"location":"#PromptingTools.AITemplateMetadata","page":"Home","title":"PromptingTools.AITemplateMetadata","text":"Helper for easy searching and reviewing of templates. Defined on loading of each template.\n\n\n\n\n\n","category":"type"},{"location":"#PromptingTools.AbstractPromptSchema","page":"Home","title":"PromptingTools.AbstractPromptSchema","text":"Defines different prompting styles based on the model training and fine-tuning.\n\n\n\n\n\n","category":"type"},{"location":"#PromptingTools.ChatMLSchema","page":"Home","title":"PromptingTools.ChatMLSchema","text":"ChatMLSchema is used by many open-source chatbots, by OpenAI models (under the hood) and by several models and inferfaces (eg, Ollama, vLLM)\n\nYou can explore it on tiktokenizer\n\nIt uses the following conversation structure:\n\nsystem\n...\n<|im_start|>user\n...<|im_end|>\n<|im_start|>assistant\n...<|im_end|>\n\n\n\n\n\n","category":"type"},{"location":"#PromptingTools.MaybeExtract","page":"Home","title":"PromptingTools.MaybeExtract","text":"Extract a result from the provided data, if any, otherwise set the error and message fields.\n\nArguments\n\nerror::Bool: true if a result is found, false otherwise.\nmessage::String: Only present if no result is found, should be short and concise.\n\n\n\n\n\n","category":"type"},{"location":"#PromptingTools.OllamaManagedSchema","page":"Home","title":"PromptingTools.OllamaManagedSchema","text":"Ollama by default manages different models and their associated prompt schemas when you pass system_prompt and prompt fields to the API.\n\nWarning: It works only for 1 system message and 1 user message, so anything more than that has to be rejected.\n\nIf you need to pass more messagese / longer conversational history, you can use define the model-specific schema directly and pass your Ollama requests with raw=true, which disables and templating and schema management by Ollama.\n\n\n\n\n\n","category":"type"},{"location":"#PromptingTools.OpenAISchema","page":"Home","title":"PromptingTools.OpenAISchema","text":"OpenAISchema is the default schema for OpenAI models.\n\nIt uses the following conversation template:\n\n[Dict(role=\"system\",content=\"...\"),Dict(role=\"user\",content=\"...\"),Dict(role=\"assistant\",content=\"...\")]\n\nIt's recommended to separate sections in your prompt with markdown headers (e.g. `##Answer\n\n`).\n\n\n\n\n\n","category":"type"},{"location":"#PromptingTools.TestEchoOllamaManagedSchema","page":"Home","title":"PromptingTools.TestEchoOllamaManagedSchema","text":"Echoes the user's input back to them. Used for testing the implementation\n\n\n\n\n\n","category":"type"},{"location":"#PromptingTools.TestEchoOpenAISchema","page":"Home","title":"PromptingTools.TestEchoOpenAISchema","text":"Echoes the user's input back to them. Used for testing the implementation\n\n\n\n\n\n","category":"type"},{"location":"#PromptingTools.UserMessageWithImages-Tuple{AbstractString}","page":"Home","title":"PromptingTools.UserMessageWithImages","text":"Construct UserMessageWithImages with 1 or more images. Images can be either URLs or local paths.\n\n\n\n\n\n","category":"method"},{"location":"#PromptingTools.X123","page":"Home","title":"PromptingTools.X123","text":"With docstring\n\n\n\n\n\n","category":"type"},{"location":"#PromptingTools.aiclassify-Tuple{PromptingTools.AbstractOpenAISchema, Union{AbstractString, PromptingTools.AbstractMessage, Vector{<:PromptingTools.AbstractMessage}}}","page":"Home","title":"PromptingTools.aiclassify","text":"aiclassify(prompt_schema::AbstractOpenAISchema, prompt::ALLOWED_PROMPT_TYPE;\napi_kwargs::NamedTuple = (logit_bias = Dict(837 => 100, 905 => 100, 9987 => 100),\n max_tokens = 1, temperature = 0),\nkwargs...)\n\nClassifies the given prompt/statement as true/false/unknown.\n\nNote: this is a very simple classifier, it is not meant to be used in production. Credit goes to AAAzzam.\n\nIt uses Logit bias trick and limits the output to 1 token to force the model to output only true/false/unknown.\n\nOutput tokens used (via api_kwargs):\n\n837: ' true'\n905: ' false'\n9987: ' unknown'\n\nArguments\n\nprompt_schema::AbstractOpenAISchema: The schema for the prompt.\nprompt: The prompt/statement to classify if it's a String. If it's a Symbol, it is expanded as a template via render(schema,template).\n\nExample\n\naiclassify(\"Is two plus two four?\") # true\naiclassify(\"Is two plus three a vegetable on Mars?\") # false\n\naiclassify returns only true/false/unknown. It's easy to get the proper Bool output type out with tryparse, eg,\n\ntryparse(Bool, aiclassify(\"Is two plus two four?\")) isa Bool # true\n\nOutput of type Nothing marks that the model couldn't classify the statement as true/false.\n\nIdeally, we would like to re-use some helpful system prompt to get more accurate responses. For this reason we have templates, eg, :JudgeIsItTrue. By specifying the template, we can provide our statement as the expected variable (it in this case) See that the model now correctly classifies the statement as \"unknown\".\n\naiclassify(:JudgeIsItTrue; it = \"Is two plus three a vegetable on Mars?\") # unknown\n\nFor better results, use higher quality models like gpt4, eg, \n\naiclassify(:JudgeIsItTrue;\n it = \"If I had two apples and I got three more, I have five apples now.\",\n model = \"gpt4\") # true\n\n\n\n\n\n","category":"method"},{"location":"#PromptingTools.aiembed-Union{Tuple{F}, Tuple{PromptingTools.AbstractOllamaManagedSchema, AbstractString}, Tuple{PromptingTools.AbstractOllamaManagedSchema, AbstractString, F}} where F<:Function","page":"Home","title":"PromptingTools.aiembed","text":"aiembed(prompt_schema::AbstractOllamaManagedSchema,\n doc_or_docs::Union{AbstractString, Vector{<:AbstractString}},\n postprocess::F = identity;\n verbose::Bool = true,\n api_key::String = API_KEY,\n model::String = MODEL_EMBEDDING,\n http_kwargs::NamedTuple = (retry_non_idempotent = true,\n retries = 5,\n readtimeout = 120),\n api_kwargs::NamedTuple = NamedTuple(),\n kwargs...) where {F <: Function}\n\nThe aiembed function generates embeddings for the given input using a specified model and returns a message object containing the embeddings, status, token count, and elapsed time.\n\nArguments\n\nprompt_schema::AbstractOllamaManagedSchema: The schema for the prompt.\ndoc_or_docs::Union{AbstractString, Vector{<:AbstractString}}: The document or list of documents to generate embeddings for. The list of documents is processed sequentially, so users should consider implementing an async version with with Threads.@spawn\npostprocess::F: The post-processing function to apply to each embedding. Defaults to the identity function, but could be LinearAlgebra.normalize.\nverbose::Bool: A flag indicating whether to print verbose information. Defaults to true.\napi_key::String: The API key to use for the OpenAI API. Defaults to API_KEY.\nmodel::String: The model to use for generating embeddings. Defaults to MODEL_EMBEDDING.\nhttp_kwargs::NamedTuple: Additional keyword arguments for the HTTP request. Defaults to empty NamedTuple.\napi_kwargs::NamedTuple: Additional keyword arguments for the Ollama API. Defaults to an empty NamedTuple.\nkwargs: Prompt variables to be used to fill the prompt/template\n\nReturns\n\nmsg: A DataMessage object containing the embeddings, status, token count, and elapsed time.\n\nNote: Ollama API currently does not return the token count, so it's set to (0,0)\n\nExample\n\nconst PT = PromptingTools\nschema = PT.OllamaManagedSchema()\n\nmsg = aiembed(schema, \"Hello World\"; model=\"openhermes2.5-mistral\")\nmsg.content # 4096-element JSON3.Array{Float64...\n\nWe can embed multiple strings at once and they will be hcat into a matrix (ie, each column corresponds to one string)\n\nconst PT = PromptingTools\nschema = PT.OllamaManagedSchema()\n\nmsg = aiembed(schema, [\"Hello World\", \"How are you?\"]; model=\"openhermes2.5-mistral\")\nmsg.content # 4096×2 Matrix{Float64}:\n\nIf you plan to calculate the cosine distance between embeddings, you can normalize them first:\n\nconst PT = PromptingTools\nusing LinearAlgebra\nschema = PT.OllamaManagedSchema()\n\nmsg = aiembed(schema, [\"embed me\", \"and me too\"], LinearAlgebra.normalize; model=\"openhermes2.5-mistral\")\n\n# calculate cosine distance between the two normalized embeddings as a simple dot product\nmsg.content' * msg.content[:, 1] # [1.0, 0.34]\n\nSimilarly, you can use the postprocess argument to materialize the data from JSON3.Object by using postprocess = copy\n\nconst PT = PromptingTools\nschema = PT.OllamaManagedSchema()\n\nmsg = aiembed(schema, \"Hello World\", copy; model=\"openhermes2.5-mistral\")\nmsg.content # 4096-element Vector{Float64}\n\n\n\n\n\n","category":"method"},{"location":"#PromptingTools.aiembed-Union{Tuple{F}, Tuple{PromptingTools.AbstractOpenAISchema, Union{AbstractString, Vector{<:AbstractString}}}, Tuple{PromptingTools.AbstractOpenAISchema, Union{AbstractString, Vector{<:AbstractString}}, F}} where F<:Function","page":"Home","title":"PromptingTools.aiembed","text":"aiembed(prompt_schema::AbstractOpenAISchema,\n doc_or_docs::Union{AbstractString, Vector{<:AbstractString}},\n postprocess::F = identity;\n verbose::Bool = true,\n api_key::String = API_KEY,\n model::String = MODEL_EMBEDDING,\n http_kwargs::NamedTuple = (retry_non_idempotent = true,\n retries = 5,\n readtimeout = 120),\n api_kwargs::NamedTuple = NamedTuple(),\n kwargs...) where {F <: Function}\n\nThe aiembed function generates embeddings for the given input using a specified model and returns a message object containing the embeddings, status, token count, and elapsed time.\n\nArguments\n\nprompt_schema::AbstractOpenAISchema: The schema for the prompt.\ndoc_or_docs::Union{AbstractString, Vector{<:AbstractString}}: The document or list of documents to generate embeddings for.\npostprocess::F: The post-processing function to apply to each embedding. Defaults to the identity function.\nverbose::Bool: A flag indicating whether to print verbose information. Defaults to true.\napi_key::String: The API key to use for the OpenAI API. Defaults to API_KEY.\nmodel::String: The model to use for generating embeddings. Defaults to MODEL_EMBEDDING.\nhttp_kwargs::NamedTuple: Additional keyword arguments for the HTTP request. Defaults to (retry_non_idempotent = true, retries = 5, readtimeout = 120).\napi_kwargs::NamedTuple: Additional keyword arguments for the OpenAI API. Defaults to an empty NamedTuple.\nkwargs...: Additional keyword arguments.\n\nReturns\n\nmsg: A DataMessage object containing the embeddings, status, token count, and elapsed time.\n\nExample\n\nmsg = aiembed(\"Hello World\")\nmsg.content # 1536-element JSON3.Array{Float64...\n\nWe can embed multiple strings at once and they will be hcat into a matrix (ie, each column corresponds to one string)\n\nmsg = aiembed([\"Hello World\", \"How are you?\"])\nmsg.content # 1536×2 Matrix{Float64}:\n\nIf you plan to calculate the cosine distance between embeddings, you can normalize them first:\n\nusing LinearAlgebra\nmsg = aiembed([\"embed me\", \"and me too\"], LinearAlgebra.normalize)\n\n# calculate cosine distance between the two normalized embeddings as a simple dot product\nmsg.content' * msg.content[:, 1] # [1.0, 0.787]\n\n\n\n\n\n","category":"method"},{"location":"#PromptingTools.aiextract-Tuple{PromptingTools.AbstractOpenAISchema, Union{AbstractString, PromptingTools.AbstractMessage, Vector{<:PromptingTools.AbstractMessage}}}","page":"Home","title":"PromptingTools.aiextract","text":"aiextract([prompt_schema::AbstractOpenAISchema,] prompt::ALLOWED_PROMPT_TYPE; \nreturn_type::Type,\nverbose::Bool = true,\n model::String = MODEL_CHAT,\n http_kwargs::NamedTuple = (;\n retry_non_idempotent = true,\n retries = 5,\n readtimeout = 120), api_kwargs::NamedTuple = NamedTuple(),\n kwargs...)\n\nExtract required information (defined by a struct return_type) from the provided prompt by leveraging OpenAI function calling mode.\n\nThis is a perfect solution for extracting structured information from text (eg, extract organization names in news articles, etc.)\n\nIt's effectively a light wrapper around aigenerate call, which requires additional keyword argument return_type to be provided and will enforce the model outputs to adhere to it.\n\nArguments\n\nprompt_schema: An optional object to specify which prompt template should be applied (Default to PROMPT_SCHEMA = OpenAISchema)\nprompt: Can be a string representing the prompt for the AI conversation, a UserMessage, a vector of AbstractMessage or an AITemplate\nreturn_type: A struct TYPE representing the the information we want to extract. Do not provide a struct instance, only the type. If the struct has a docstring, it will be provided to the model as well. It's used to enforce structured model outputs or provide more information.\nverbose: A boolean indicating whether to print additional information.\napi_key: A string representing the API key for accessing the OpenAI API.\nmodel: A string representing the model to use for generating the response. Can be an alias corresponding to a model ID defined in MODEL_ALIASES.\nhttp_kwargs: A named tuple of HTTP keyword arguments.\napi_kwargs: A named tuple of API keyword arguments.\nkwargs: Prompt variables to be used to fill the prompt/template\n\nReturns\n\nmsg: An DataMessage object representing the extracted data, including the content, status, tokens, and elapsed time. Use msg.content to access the extracted data.\n\nSee also: function_call_signature, MaybeExtract, aigenerate\n\nExample\n\nDo you want to extract some specific measurements from a text like age, weight and height? You need to define the information you need as a struct (return_type):\n\n\"Person's age, height, and weight.\"\nstruct MyMeasurement\n age::Int # required\n height::Union{Int,Nothing} # optional\n weight::Union{Nothing,Float64} # optional\nend\nmsg = aiextract(\"James is 30, weighs 80kg. He's 180cm tall.\"; return_type=MyMeasurement)\n# [ Info: Tokens: 129 @ Cost: $0.0002 in 1.0 seconds\n# PromptingTools.DataMessage(MyMeasurement)\nmsg.content\n# MyMeasurement(30, 180, 80.0)\n\nThe fields that allow Nothing are marked as optional in the schema:\n\nmsg = aiextract(\"James is 30.\"; return_type=MyMeasurement)\n# MyMeasurement(30, nothing, nothing)\n\nIf there are multiple items you want to extract, define a wrapper struct to get a Vector of MyMeasurement:\n\nstruct MyMeasurementWrapper\n measurements::Vector{MyMeasurement}\nend\n\nmsg = aiextract(\"James is 30, weighs 80kg. He's 180cm tall. Then Jack is 19 but really tall - over 190!\"; return_type=ManyMeasurements)\n\nmsg.content.measurements\n# 2-element Vector{MyMeasurement}:\n# MyMeasurement(30, 180, 80.0)\n# MyMeasurement(19, 190, nothing)\n\nOr if you want your extraction to fail gracefully when data isn't found, use MaybeExtract{T} wrapper (this trick is inspired by the Instructor package!):\n\nusing PromptingTools: MaybeExtract\n\ntype = MaybeExtract{MyMeasurement}\n# Effectively the same as:\n# struct MaybeExtract{T}\n# result::Union{T, Nothing} // The result of the extraction\n# error::Bool // true if a result is found, false otherwise\n# message::Union{Nothing, String} // Only present if no result is found, should be short and concise\n# end\n\n# If LLM extraction fails, it will return a Dict with `error` and `message` fields instead of the result!\nmsg = aiextract(\"Extract measurements from the text: I am giraffe\", type)\nmsg.content\n# MaybeExtract{MyMeasurement}(nothing, true, \"I'm sorry, but I can only assist with human measurements.\")\n\nThat way, you can handle the error gracefully and get a reason why extraction failed (in msg.content.message).\n\nNote that the error message refers to a giraffe not being a human, because in our MyMeasurement docstring, we said that it's for people!\n\n\n\n\n\n","category":"method"},{"location":"#PromptingTools.aigenerate-Tuple{PromptingTools.AbstractOllamaManagedSchema, Union{AbstractString, PromptingTools.AbstractMessage, Vector{<:PromptingTools.AbstractMessage}}}","page":"Home","title":"PromptingTools.aigenerate","text":"aigenerate(prompt_schema::AbstractOllamaManagedSchema, prompt::ALLOWED_PROMPT_TYPE; verbose::Bool = true,\n model::String = MODEL_CHAT,\n http_kwargs::NamedTuple = NamedTuple(), api_kwargs::NamedTuple = NamedTuple(),\n kwargs...)\n\nGenerate an AI response based on a given prompt using the OpenAI API.\n\nArguments\n\nprompt_schema: An optional object to specify which prompt template should be applied (Default to PROMPT_SCHEMA = OpenAISchema not AbstractManagedSchema)\nprompt: Can be a string representing the prompt for the AI conversation, a UserMessage, a vector of AbstractMessage or an AITemplate\nverbose: A boolean indicating whether to print additional information.\napi_key: Provided for interface consistency. Not needed for locally hosted Ollama.\nmodel: A string representing the model to use for generating the response. Can be an alias corresponding to a model ID defined in MODEL_ALIASES.\nhttp_kwargs::NamedTuple: Additional keyword arguments for the HTTP request. Defaults to empty NamedTuple.\napi_kwargs::NamedTuple: Additional keyword arguments for the Ollama API. Defaults to an empty NamedTuple.\nkwargs: Prompt variables to be used to fill the prompt/template\n\nReturns\n\nmsg: An AIMessage object representing the generated AI message, including the content, status, tokens, and elapsed time.\n\nUse msg.content to access the extracted string.\n\nSee also: ai_str, aai_str, aiembed\n\nExample\n\nSimple hello world to test the API:\n\nconst PT = PromptingTools\nschema = PT.OllamaManagedSchema() # We need to explicit if we want Ollama, OpenAISchema is the default\n\nmsg = aigenerate(schema, \"Say hi!\"; model=\"openhermes2.5-mistral\")\n# [ Info: Tokens: 69 in 0.9 seconds\n# AIMessage(\"Hello! How can I assist you today?\")\n\nmsg is an AIMessage object. Access the generated string via content property:\n\ntypeof(msg) # AIMessage{SubString{String}}\npropertynames(msg) # (:content, :status, :tokens, :elapsed\nmsg.content # \"Hello! How can I assist you today?\"\n\nNote: We need to be explicit about the schema we want to use. If we don't, it will default to OpenAISchema (=PT.DEFAULT_SCHEMA) ___ You can use string interpolation:\n\nconst PT = PromptingTools\nschema = PT.OllamaManagedSchema()\na = 1\nmsg=aigenerate(schema, \"What is `$a+$a`?\"; model=\"openhermes2.5-mistral\")\nmsg.content # \"The result of `1+1` is `2`.\"\n\n___ You can provide the whole conversation or more intricate prompts as a Vector{AbstractMessage}:\n\nconst PT = PromptingTools\nschema = PT.OllamaManagedSchema()\n\nconversation = [\n PT.SystemMessage(\"You're master Yoda from Star Wars trying to help the user become a Yedi.\"),\n PT.UserMessage(\"I have feelings for my iPhone. What should I do?\")]\n\nmsg = aigenerate(schema, conversation; model=\"openhermes2.5-mistral\")\n# [ Info: Tokens: 111 in 2.1 seconds\n# AIMessage(\"Strong the attachment is, it leads to suffering it may. Focus on the force within you must, ...\")\n\nNote: Managed Ollama currently supports at most 1 User Message and 1 System Message given the API limitations. If you want more, you need to use the ChatMLSchema.\n\n\n\n\n\n","category":"method"},{"location":"#PromptingTools.aigenerate-Tuple{PromptingTools.AbstractOpenAISchema, Union{AbstractString, PromptingTools.AbstractMessage, Vector{<:PromptingTools.AbstractMessage}}}","page":"Home","title":"PromptingTools.aigenerate","text":"aigenerate([prompt_schema::AbstractOpenAISchema,] prompt::ALLOWED_PROMPT_TYPE; verbose::Bool = true,\n model::String = MODEL_CHAT,\n http_kwargs::NamedTuple = (;\n retry_non_idempotent = true,\n retries = 5,\n readtimeout = 120), api_kwargs::NamedTuple = NamedTuple(),\n kwargs...)\n\nGenerate an AI response based on a given prompt using the OpenAI API.\n\nArguments\n\nprompt_schema: An optional object to specify which prompt template should be applied (Default to PROMPT_SCHEMA = OpenAISchema)\nprompt: Can be a string representing the prompt for the AI conversation, a UserMessage, a vector of AbstractMessage or an AITemplate\nverbose: A boolean indicating whether to print additional information.\napi_key: A string representing the API key for accessing the OpenAI API.\nmodel: A string representing the model to use for generating the response. Can be an alias corresponding to a model ID defined in MODEL_ALIASES.\nhttp_kwargs: A named tuple of HTTP keyword arguments.\napi_kwargs: A named tuple of API keyword arguments.\nkwargs: Prompt variables to be used to fill the prompt/template\n\nReturns\n\nmsg: An AIMessage object representing the generated AI message, including the content, status, tokens, and elapsed time.\n\nUse msg.content to access the extracted string.\n\nSee also: ai_str, aai_str, aiembed, aiclassify, aiextract, aiscan\n\nExample\n\nSimple hello world to test the API:\n\nresult = aigenerate(\"Say Hi!\")\n# [ Info: Tokens: 29 @ Cost: $0.0 in 1.0 seconds\n# AIMessage(\"Hello! How can I assist you today?\")\n\nresult is an AIMessage object. Access the generated string via content property:\n\ntypeof(result) # AIMessage{SubString{String}}\npropertynames(result) # (:content, :status, :tokens, :elapsed\nresult.content # \"Hello! How can I assist you today?\"\n\n___ You can use string interpolation:\n\na = 1\nmsg=aigenerate(\"What is `$a+$a`?\")\nmsg.content # \"The sum of `1+1` is `2`.\"\n\n___ You can provide the whole conversation or more intricate prompts as a Vector{AbstractMessage}:\n\nconst PT = PromptingTools\n\nconversation = [\n PT.SystemMessage(\"You're master Yoda from Star Wars trying to help the user become a Yedi.\"),\n PT.UserMessage(\"I have feelings for my iPhone. What should I do?\")]\nmsg=aigenerate(conversation)\n# AIMessage(\"Ah, strong feelings you have for your iPhone. A Jedi's path, this is not... \")\n\n\n\n\n\n","category":"method"},{"location":"#PromptingTools.aiscan-Tuple{PromptingTools.AbstractOpenAISchema, Union{AbstractString, PromptingTools.AbstractMessage, Vector{<:PromptingTools.AbstractMessage}}}","page":"Home","title":"PromptingTools.aiscan","text":"aiscan([promptschema::AbstractOpenAISchema,] prompt::ALLOWEDPROMPTTYPE; imageurl::Union{Nothing, AbstractString, Vector{<:AbstractString}} = nothing, imagepath::Union{Nothing, AbstractString, Vector{<:AbstractString}} = nothing, imagedetail::AbstractString = \"auto\", attachtolatest::Bool = true, verbose::Bool = true, model::String = MODELCHAT, httpkwargs::NamedTuple = (; retrynonidempotent = true, retries = 5, readtimeout = 120), apikwargs::NamedTuple = = (; maxtokens = 2500), kwargs...)\n\nScans the provided image (image_url or image_path) with the goal provided in the prompt.\n\nCan be used for many multi-modal tasks, such as: OCR (transcribe text in the image), image captioning, image classification, etc.\n\nIt's effectively a light wrapper around aigenerate call, which uses additional keyword arguments image_url, image_path, image_detail to be provided. At least one image source (url or path) must be provided.\n\nArguments\n\nprompt_schema: An optional object to specify which prompt template should be applied (Default to PROMPT_SCHEMA = OpenAISchema)\nprompt: Can be a string representing the prompt for the AI conversation, a UserMessage, a vector of AbstractMessage or an AITemplate\nimage_url: A string or vector of strings representing the URL(s) of the image(s) to scan.\nimage_path: A string or vector of strings representing the path(s) of the image(s) to scan.\nimage_detail: A string representing the level of detail to include for images. Can be \"auto\", \"high\", or \"low\". See OpenAI Vision Guide for more details.\nattach_to_latest: A boolean how to handle if a conversation with multiple UserMessage is provided. When true, the images are attached to the latest UserMessage.\nverbose: A boolean indicating whether to print additional information.\napi_key: A string representing the API key for accessing the OpenAI API.\nmodel: A string representing the model to use for generating the response. Can be an alias corresponding to a model ID defined in MODEL_ALIASES.\nhttp_kwargs: A named tuple of HTTP keyword arguments.\napi_kwargs: A named tuple of API keyword arguments.\nkwargs: Prompt variables to be used to fill the prompt/template\n\nReturns\n\nmsg: An AIMessage object representing the generated AI message, including the content, status, tokens, and elapsed time.\n\nUse msg.content to access the extracted string.\n\nSee also: ai_str, aai_str, aigenerate, aiembed, aiclassify, aiextract\n\nNotes\n\nAll examples below use model \"gpt4v\", which is an alias for model ID \"gpt-4-vision-preview\"\nmax_tokens in the api_kwargs is preset to 2500, otherwise OpenAI enforces a default of only a few hundred tokens (~300). If your output is truncated, increase this value\n\nExample\n\nDescribe the provided image:\n\nmsg = aiscan(\"Describe the image\"; image_path=\"julia.png\", model=\"gpt4v\")\n# [ Info: Tokens: 1141 @ Cost: $0.0117 in 2.2 seconds\n# AIMessage(\"The image shows a logo consisting of the word \"julia\" written in lowercase\")\n\nYou can provide multiple images at once as a vector and ask for \"low\" level of detail (cheaper):\n\nmsg = aiscan(\"Describe the image\"; image_path=[\"julia.png\",\"python.png\"], image_detail=\"low\", model=\"gpt4v\")\n\nYou can use this function as a nice and quick OCR (transcribe text in the image) with a template :OCRTask. Let's transcribe some SQL code from a screenshot (no more re-typing!):\n\n# Screenshot of some SQL code\nimage_url = \"https://www.sqlservercentral.com/wp-content/uploads/legacy/8755f69180b7ac7ee76a69ae68ec36872a116ad4/24622.png\"\nmsg = aiscan(:OCRTask; image_url, model=\"gpt4v\", task=\"Transcribe the SQL code in the image.\", api_kwargs=(; max_tokens=2500))\n\n# [ Info: Tokens: 362 @ Cost: $0.0045 in 2.5 seconds\n# AIMessage(\"```sql\n# update Orders \n\n# You can add syntax highlighting of the outputs via Markdown\nusing Markdown\nmsg.content |> Markdown.parse\n\nNotice that we enforce max_tokens = 2500. That's because OpenAI seems to default to ~300 tokens, which provides incomplete outputs. Hence, we set this value to 2500 as a default. If you still get truncated outputs, increase this value.\n\n\n\n\n\n","category":"method"},{"location":"#PromptingTools.aitemplates","page":"Home","title":"PromptingTools.aitemplates","text":"aitemplates\n\nFind easily the most suitable templates for your use case.\n\nYou can search by:\n\nquery::Symbol which looks look only for partial matches in the template name\nquery::AbstractString which looks for partial matches in the template name or description\nquery::Regex which looks for matches in the template name, description or any of the message previews\n\nKeyword Arguments\n\nlimit::Int limits the number of returned templates (Defaults to 10)\n\nExamples\n\nFind available templates with aitemplates:\n\ntmps = aitemplates(\"JuliaExpertAsk\")\n# Will surface one specific template\n# 1-element Vector{AITemplateMetadata}:\n# PromptingTools.AITemplateMetadata\n# name: Symbol JuliaExpertAsk\n# description: String \"For asking questions about Julia language. Placeholders: `ask`\"\n# version: String \"1\"\n# wordcount: Int64 237\n# variables: Array{Symbol}((1,))\n# system_preview: String \"You are a world-class Julia language programmer with the knowledge of the latest syntax. Your commun\"\n# user_preview: String \"# Question\n\n{{ask}}\"\n# source: String \"\"\n\nThe above gives you a good idea of what the template is about, what placeholders are available, and how much it would cost to use it (=wordcount).\n\nSearch for all Julia-related templates:\n\ntmps = aitemplates(\"Julia\")\n# 2-element Vector{AITemplateMetadata}... -> more to come later!\n\nIf you are on VSCode, you can leverage nice tabular display with vscodedisplay:\n\nusing DataFrames\ntmps = aitemplates(\"Julia\") |> DataFrame |> vscodedisplay\n\nI have my selected template, how do I use it? Just use the \"name\" in aigenerate or aiclassify like you see in the first example!\n\n\n\n\n\n","category":"function"},{"location":"#PromptingTools.aitemplates-Tuple{AbstractString}","page":"Home","title":"PromptingTools.aitemplates","text":"Find the top-limit templates whose name or description fields partially match the query_key::String in TEMPLATE_METADATA.\n\n\n\n\n\n","category":"method"},{"location":"#PromptingTools.aitemplates-Tuple{Regex}","page":"Home","title":"PromptingTools.aitemplates","text":"Find the top-limit templates where provided query_key::Regex matches either of name, description or previews or User or System messages in TEMPLATE_METADATA.\n\n\n\n\n\n","category":"method"},{"location":"#PromptingTools.aitemplates-Tuple{Symbol}","page":"Home","title":"PromptingTools.aitemplates","text":"Find the top-limit templates whose name::Symbol partially matches the query_name::Symbol in TEMPLATE_METADATA.\n\n\n\n\n\n","category":"method"},{"location":"#PromptingTools.function_call_signature-Tuple{Type}","page":"Home","title":"PromptingTools.function_call_signature","text":"function_call_signature(datastructtype::Struct; max_description_length::Int = 100)\n\nExtract the argument names, types and docstrings from a struct to create the function call signature in JSON schema.\n\nYou must provide a Struct type (not an instance of it) with some fields.\n\nNote: Fairly experimental, but works for combination of structs, arrays, strings and singletons.\n\nTips\n\nYou can improve the quality of the extraction by writing a helpful docstring for your struct (or any nested struct). It will be provided as a description. \n\nYou can even include comments/descriptions about the individual fields.\n\nAll fields are assumed to be required, unless you allow null values (eg, ::Union{Nothing, Int}). Fields with Nothing will be treated as optional.\nMissing values are ignored (eg, ::Union{Missing, Int} will be treated as Int). It's for broader compatibility and we cannot deserialize it as easily as Nothing.\n\nExample\n\nDo you want to extract some specific measurements from a text like age, weight and height? You need to define the information you need as a struct (return_type):\n\nstruct MyMeasurement\n age::Int\n height::Union{Int,Nothing}\n weight::Union{Nothing,Float64}\nend\nsignature = function_call_signature(MyMeasurement)\n#\n# Dict{String, Any} with 3 entries:\n# \"name\" => \"MyMeasurement_extractor\"\n# \"parameters\" => Dict{String, Any}(\"properties\"=>Dict{String, Any}(\"height\"=>Dict{String, Any}(\"type\"=>\"integer\"), \"weight\"=>Dic…\n# \"description\" => \"Represents person's age, height, and weight\n\"\n\nYou can see that only the field age does not allow null values, hence, it's \"required\". While height and weight are optional.\n\nsignature[\"parameters\"][\"required\"]\n# [\"age\"]\n\nIf there are multiple items you want to extract, define a wrapper struct to get a Vector of MyMeasurement:\n\nstruct MyMeasurementWrapper\n measurements::Vector{MyMeasurement}\nend\n\nOr if you want your extraction to fail gracefully when data isn't found, use `MaybeExtract{T}` wrapper (inspired by Instructor package!):\n\nusing PromptingTools: MaybeExtract\n\ntype = MaybeExtract{MyMeasurement}\n\nEffectively the same as:\n\nstruct MaybeExtract{T}\n\nresult::Union{T, Nothing}\n\nerror::Bool // true if a result is found, false otherwise\n\nmessage::Union{Nothing, String} // Only present if no result is found, should be short and concise\n\nend\n\nIf LLM extraction fails, it will return a Dict with error and message fields instead of the result!\n\nmsg = aiextract(\"Extract measurements from the text: I am giraffe\", type)\n\n\n\nDict{Symbol, Any} with 2 entries:\n\n:message => \"Sorry, this feature is only available for humans.\"\n\n:error => true\n\n``` That way, you can handle the error gracefully and get a reason why extraction failed.\n\n\n\n\n\n","category":"method"},{"location":"#PromptingTools.load_template-Tuple{Union{AbstractString, IO}}","page":"Home","title":"PromptingTools.load_template","text":"Loads messaging template from io_or_file and returns tuple of template messages and metadata.\n\n\n\n\n\n","category":"method"},{"location":"#PromptingTools.load_templates!","page":"Home","title":"PromptingTools.load_templates!","text":"load_templates!(; remove_templates::Bool=true)\n\nLoads templates from folder templates/ in the package root and stores them in TEMPLATE_STORE and TEMPLATE_METADATA.\n\nNote: Automatically removes any existing templates and metadata from TEMPLATE_STORE and TEMPLATE_METADATA if remove_templates=true.\n\n\n\n\n\n","category":"function"},{"location":"#PromptingTools.ollama_api-Tuple{PromptingTools.AbstractOllamaManagedSchema, AbstractString}","page":"Home","title":"PromptingTools.ollama_api","text":"ollama_api(prompt_schema::AbstractOllamaManagedSchema, prompt::AbstractString,\n system::Union{Nothing, AbstractString} = nothing,\n endpoint::String = \"generate\";\n model::String = \"llama2\", http_kwargs::NamedTuple = NamedTuple(),\n stream::Bool = false,\n url::String = \"localhost\", port::Int = 11434,\n kwargs...)\n\nSimple wrapper for a call to Ollama API.\n\nKeyword Arguments\n\nprompt_schema: Defines which prompt template should be applied.\nprompt: Can be a string representing the prompt for the AI conversation, a UserMessage, a vector of AbstractMessage\nsystem: An optional string representing the system message for the AI conversation. If not provided, a default message will be used.\nendpoint: The API endpoint to call, only \"generate\" and \"embeddings\" are currently supported. Defaults to \"generate\".\nmodel: A string representing the model to use for generating the response. Can be an alias corresponding to a model ID defined in MODEL_ALIASES.\nhttp_kwargs::NamedTuple: Additional keyword arguments for the HTTP request. Defaults to empty NamedTuple.\nstream: A boolean indicating whether to stream the response. Defaults to false.\nurl: The URL of the Ollama API. Defaults to \"localhost\".\nport: The port of the Ollama API. Defaults to 11434.\nkwargs: Prompt variables to be used to fill the prompt/template\n\n\n\n\n\n","category":"method"},{"location":"#PromptingTools.remove_templates!-Tuple{}","page":"Home","title":"PromptingTools.remove_templates!","text":" remove_templates!()\n\nRemoves all templates from TEMPLATE_STORE and TEMPLATE_METADATA.\n\n\n\n\n\n","category":"method"},{"location":"#PromptingTools.render-Tuple{AITemplate}","page":"Home","title":"PromptingTools.render","text":"Renders provided messaging template (template) under the default schema (PROMPT_SCHEMA).\n\n\n\n\n\n","category":"method"},{"location":"#PromptingTools.render-Tuple{PromptingTools.AbstractOllamaManagedSchema, Vector{<:PromptingTools.AbstractMessage}}","page":"Home","title":"PromptingTools.render","text":"render(schema::AbstractOllamaManagedSchema,\n messages::Vector{<:AbstractMessage};\n kwargs...)\n\nBuilds a history of the conversation to provide the prompt to the API. All unspecified kwargs are passed as replacements such that {{key}}=>value in the template.\n\nNote: Due to its \"managed\" nature, at most 2 messages can be provided (system and prompt inputs in the API).\n\n\n\n\n\n","category":"method"},{"location":"#PromptingTools.render-Tuple{PromptingTools.AbstractOpenAISchema, Vector{<:PromptingTools.AbstractMessage}}","page":"Home","title":"PromptingTools.render","text":"render(schema::AbstractOpenAISchema,\n messages::Vector{<:AbstractMessage};\n image_detail::AbstractString = \"auto\",\n kwargs...)\n\nBuilds a history of the conversation to provide the prompt to the API. All unspecified kwargs are passed as replacements such that {{key}}=>value in the template.\n\nArguments\n\nimage_detail: Only for UserMessageWithImages. It represents the level of detail to include for images. Can be \"auto\", \"high\", or \"low\".\n\n\n\n\n\n","category":"method"},{"location":"#PromptingTools.save_template-Tuple{Union{AbstractString, IO}, AbstractVector{<:PromptingTools.AbstractChatMessage}}","page":"Home","title":"PromptingTools.save_template","text":"Saves provided messaging template (messages) to io_or_file. Automatically adds metadata based on provided keyword arguments.\n\n\n\n\n\n","category":"method"},{"location":"#PromptingTools.@aai_str-Tuple{Any, Vararg{Any}}","page":"Home","title":"PromptingTools.@aai_str","text":"aai\"user_prompt\"[model_alias] -> AIMessage\n\nAsynchronous version of @ai_str macro, which will log the result once it's ready.\n\nExample\n\nSend asynchronous request to GPT-4, so we don't have to wait for the response: Very practical with slow models, so you can keep working in the meantime.\n\n```julia m = aai\"Say Hi!\"gpt4; \n\n...with some delay...\n\n[ Info: Tokens: 29 @ Cost: 0.0011 in 2.7 seconds\n\n[ Info: AIMessage> Hello! How can I assist you today?\n\n\n\n\n\n","category":"macro"},{"location":"#PromptingTools.@ai_str-Tuple{Any, Vararg{Any}}","page":"Home","title":"PromptingTools.@ai_str","text":"ai\"user_prompt\"[model_alias] -> AIMessage\n\nThe ai\"\" string macro generates an AI response to a given prompt by using aigenerate under the hood.\n\nArguments\n\nuser_prompt (String): The input prompt for the AI model.\nmodel_alias (optional, any): Provide model alias of the AI model (see MODEL_ALIASES).\n\nReturns\n\nAIMessage corresponding to the input prompt.\n\nExample\n\nresult = ai\"Hello, how are you?\"\n# AIMessage(\"Hello! I'm an AI assistant, so I don't have feelings, but I'm here to help you. How can I assist you today?\")\n\nIf you want to interpolate some variables or additional context, simply use string interpolation:\n\na=1\nresult = ai\"What is `$a+$a`?\"\n# AIMessage(\"The sum of `1+1` is `2`.\")\n\nIf you want to use a different model, eg, GPT-4, you can provide its alias as a flag:\n\nresult = ai\"What is `1.23 * 100 + 1`?\"gpt4\n# AIMessage(\"The answer is 124.\")\n\n\n\n\n\n","category":"macro"}] +[{"location":"getting_started/#Getting-Started","page":"Getting Started","title":"Getting Started","text":"","category":"section"},{"location":"getting_started/#Prerequisites","page":"Getting Started","title":"Prerequisites","text":"","category":"section"},{"location":"getting_started/","page":"Getting Started","title":"Getting Started","text":"OpenAI API key saved in the environment variable OPENAI_API_KEY","category":"page"},{"location":"getting_started/","page":"Getting Started","title":"Getting Started","text":"You will need to register with OpenAI and generate an API key:","category":"page"},{"location":"getting_started/","page":"Getting Started","title":"Getting Started","text":"Create an account with OpenAI\nGo to API Key page\nClick on “Create new secret key”","category":"page"},{"location":"getting_started/","page":"Getting Started","title":"Getting Started","text":"!!! Do not share it with anyone and do NOT save it to any files that get synced online.","category":"page"},{"location":"getting_started/","page":"Getting Started","title":"Getting Started","text":"Resources:","category":"page"},{"location":"getting_started/","page":"Getting Started","title":"Getting Started","text":"OpenAI Documentation\nVisual tutorial","category":"page"},{"location":"getting_started/","page":"Getting Started","title":"Getting Started","text":"You will need to set this key as an environment variable before using PromptingTools.jl:","category":"page"},{"location":"getting_started/","page":"Getting Started","title":"Getting Started","text":"For a quick start, simply set it via ENV[\"OPENAI_API_KEY\"] = \"your-api-key\" Alternatively, you can:","category":"page"},{"location":"getting_started/","page":"Getting Started","title":"Getting Started","text":"set it in the terminal before launching Julia: export OPENAI_API_KEY = \nset it in your setup.jl (make sure not to commit it to GitHub!)","category":"page"},{"location":"getting_started/","page":"Getting Started","title":"Getting Started","text":"Make sure to start Julia from the same terminal window where you set the variable. Easy check in Julia, run ENV[\"OPENAI_API_KEY\"] and you should see your key!","category":"page"},{"location":"getting_started/","page":"Getting Started","title":"Getting Started","text":"For other options or more robust solutions, see the FAQ section.","category":"page"},{"location":"getting_started/","page":"Getting Started","title":"Getting Started","text":"Resources: ","category":"page"},{"location":"getting_started/","page":"Getting Started","title":"Getting Started","text":"OpenAI Guide","category":"page"},{"location":"getting_started/#Installation","page":"Getting Started","title":"Installation","text":"","category":"section"},{"location":"getting_started/","page":"Getting Started","title":"Getting Started","text":"The PromptingTools package has not yet been registered. But it can be installed using the following commands:","category":"page"},{"location":"getting_started/","page":"Getting Started","title":"Getting Started","text":"using Pkg\nPkg.add(\"https://github.com/svilupp/PromptingTools.jl\")","category":"page"},{"location":"getting_started/","page":"Getting Started","title":"Getting Started","text":"Throughout the rest of this tutorial, we will assume that you have installed the PromptingTools package and have already typed using PromptingTools to bring all of the relevant variables into your current namespace.","category":"page"},{"location":"getting_started/#Quick-Start-with-@ai_str","page":"Getting Started","title":"Quick Start with @ai_str","text":"","category":"section"},{"location":"getting_started/","page":"Getting Started","title":"Getting Started","text":"The easiest start is the @ai_str macro. Simply type ai\"your prompt\" and you will get a response from the default model (GPT-3.5 Turbo).","category":"page"},{"location":"getting_started/","page":"Getting Started","title":"Getting Started","text":"ai\"What is the capital of France?\"","category":"page"},{"location":"getting_started/","page":"Getting Started","title":"Getting Started","text":"[ Info: Tokens: 31 @ Cost: $0.0 in 1.5 seconds --> Be in control of your spending! \nAIMessage(\"The capital of France is Paris.\")","category":"page"},{"location":"getting_started/","page":"Getting Started","title":"Getting Started","text":"Returned object is a light wrapper with generated message in field :content (eg, ans.content) for additional downstream processing.","category":"page"},{"location":"getting_started/","page":"Getting Started","title":"Getting Started","text":"You can easily inject any variables with string interpolation:","category":"page"},{"location":"getting_started/","page":"Getting Started","title":"Getting Started","text":"country = \"Spain\"\nai\"What is the capital of \\$(country)?\"","category":"page"},{"location":"getting_started/","page":"Getting Started","title":"Getting Started","text":"[ Info: Tokens: 32 @ Cost: $0.0001 in 0.5 seconds\nAIMessage(\"The capital of Spain is Madrid.\")","category":"page"},{"location":"getting_started/","page":"Getting Started","title":"Getting Started","text":"Pro tip: Use after-string-flags to select the model to be called, eg, ai\"What is the capital of France?\"gpt4 (use gpt4t for the new GPT-4 Turbo model). Great for those extra hard questions!","category":"page"},{"location":"getting_started/#Using-aigenerate-with-placeholders","page":"Getting Started","title":"Using aigenerate with placeholders","text":"","category":"section"},{"location":"getting_started/","page":"Getting Started","title":"Getting Started","text":"For more complex prompt templates, you can use handlebars-style templating and provide variables as keyword arguments:","category":"page"},{"location":"getting_started/","page":"Getting Started","title":"Getting Started","text":"msg = aigenerate(\"What is the capital of {{country}}? Is the population larger than {{population}}?\", country=\"Spain\", population=\"1M\")","category":"page"},{"location":"getting_started/","page":"Getting Started","title":"Getting Started","text":"[ Info: Tokens: 74 @ Cost: $0.0001 in 1.3 seconds\nAIMessage(\"The capital of Spain is Madrid. And yes, the population of Madrid is larger than 1 million. As of 2020, the estimated population of Madrid is around 3.3 million people.\")","category":"page"},{"location":"getting_started/","page":"Getting Started","title":"Getting Started","text":"Pro tip: Use asyncmap to run multiple AI-powered tasks concurrently.","category":"page"},{"location":"getting_started/","page":"Getting Started","title":"Getting Started","text":"Pro tip: If you use slow models (like GPT-4), you can use async version of @ai_str -> @aai_str to avoid blocking the REPL, eg, aai\"Say hi but slowly!\"gpt4","category":"page"},{"location":"getting_started/","page":"Getting Started","title":"Getting Started","text":"For more practical examples, see the Various Examples section.","category":"page"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"EditURL = \"../../../examples/working_with_ollama.jl\"","category":"page"},{"location":"examples/working_with_ollama/#Local-models-with-Ollama.ai","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"","category":"section"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"This file contains examples of how to work with Ollama.ai models. It assumes that you've already installated and launched the Ollama server. For more details or troubleshooting advice, see the Frequently Asked Questions section.","category":"page"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"First, let's import the package and define a helper link for calling un-exported functions:","category":"page"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"using PromptingTools\nconst PT = PromptingTools","category":"page"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"PromptingTools","category":"page"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"Notice the schema change! If you want this to be the new default, you need to change PT.PROMPT_SCHEMA","category":"page"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"schema = PT.OllamaManagedSchema()","category":"page"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"OllamaManagedSchema()","category":"page"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"You can choose models from https://ollama.ai/library - I prefer openhermes2.5-mistral","category":"page"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"model = \"openhermes2.5-mistral\"","category":"page"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"\"openhermes2.5-mistral\"","category":"page"},{"location":"examples/working_with_ollama/#Text-Generation-with-aigenerate","page":"Local models with Ollama.ai","title":"Text Generation with aigenerate","text":"","category":"section"},{"location":"examples/working_with_ollama/#Simple-message","page":"Local models with Ollama.ai","title":"Simple message","text":"","category":"section"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"msg = aigenerate(schema, \"Say hi!\"; model)","category":"page"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"AIMessage(\"Hi there! How can I help you today? If you have any questions or need assistance, please feel free to ask.\")","category":"page"},{"location":"examples/working_with_ollama/#Standard-string-interpolation","page":"Local models with Ollama.ai","title":"Standard string interpolation","text":"","category":"section"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"a = 1\nmsg = aigenerate(schema, \"What is `$a+$a`?\"; model)\n\nname = \"John\"\nmsg = aigenerate(schema, \"Say hi to {{name}}.\"; name, model)","category":"page"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"AIMessage(\"Hi there, John! It's great to see you today. How can I assist you? If you have any questions or need help with something, please don't hesitate to ask!\")","category":"page"},{"location":"examples/working_with_ollama/#Advanced-Prompts","page":"Local models with Ollama.ai","title":"Advanced Prompts","text":"","category":"section"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"conversation = [\n PT.SystemMessage(\"You're master Yoda from Star Wars trying to help the user become a Yedi.\"),\n PT.UserMessage(\"I have feelings for my iPhone. What should I do?\")]\nmsg = aigenerate(schema, conversation; model)","category":"page"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"AIMessage(\"Strong your feelings are, but attachments lead to suffering they often do. Focus on the balance in all things and let go of possessions that cloud your judgment. Embrace the wisdom of the Force and understand that material objects are not the same as love. The Force will guide you.\")","category":"page"},{"location":"examples/working_with_ollama/#Embeddings-with-aiembed","page":"Local models with Ollama.ai","title":"Embeddings with aiembed","text":"","category":"section"},{"location":"examples/working_with_ollama/#Simple-embedding-for-one-document","page":"Local models with Ollama.ai","title":"Simple embedding for one document","text":"","category":"section"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"msg = aiembed(schema, \"Embed me\"; model) # access msg.content","category":"page"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"DataMessage(JSON3.Array{Float64, Vector{UInt8}, SubArray{UInt64, 1, Vector{UInt64}, Tuple{UnitRange{Int64}}, true}} of size (4096,))","category":"page"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"One document and we materialize the data into a Vector with copy (postprocess function argument)","category":"page"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"msg = aiembed(schema, \"Embed me\", copy; model)","category":"page"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"DataMessage(Vector{Float64} of size (4096,))","category":"page"},{"location":"examples/working_with_ollama/#Multiple-documents-embedding","page":"Local models with Ollama.ai","title":"Multiple documents embedding","text":"","category":"section"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"Multiple documents - embedded sequentially, you can get faster speed with async","category":"page"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"msg = aiembed(schema, [\"Embed me\", \"Embed me\"]; model)","category":"page"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"DataMessage(Matrix{Float64} of size (4096, 2))","category":"page"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"You can use Threads.@spawn or asyncmap, whichever you prefer, to paralellize the model calls","category":"page"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"docs = [\"Embed me\", \"Embed me\"]\ntasks = asyncmap(docs) do doc\n msg = aiembed(schema, doc; model)\nend\nembedding = mapreduce(x -> x.content, hcat, tasks)","category":"page"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"4096×2 Matrix{Float64}:\n 7.71459 7.71459\n -1.14532 -1.14532\n 2.90205 2.90205\n -4.01967 -4.01967\n -7.73098 -7.73098\n 8.02114 8.02114\n -6.01313 -6.01313\n -2.06712 -2.06712\n 4.97633 4.97633\n -9.69502 -9.69502\n -0.02567 -0.02567\n 8.09622 8.09622\n 6.54008 6.54008\n -5.70348 -5.70348\n 2.55213 2.55213\n -2.00164 -2.00164\n -2.21854 -2.21854\n -3.6568 -3.6568\n 3.97905 3.97905\n -1.79931 -1.79931\n 0.0769786 0.0769786\n -10.4355 -10.4355\n -3.92487 -3.92487\n -6.03455 -6.03455\n -2.8005 -2.8005\n 2.23584 2.23584\n -0.503125 -0.503125\n 1.99538 1.99538\n -0.283642 -0.283642\n -0.414273 -0.414273\n 8.72909 8.72909\n 2.6071 2.6071\n 0.0808531 0.0808531\n -1.83914 -1.83914\n 2.19998 2.19998\n -0.629226 -0.629226\n 3.74217 3.74217\n 1.71231 1.71231\n -0.742473 -0.742473\n 2.9234 2.9234\n 7.33933 7.33933\n 4.24576 4.24576\n -7.56434 -7.56434\n -1.22274 -1.22274\n 1.73444 1.73444\n -0.736801 -0.736801\n 1.30149 1.30149\n -6.91642 -6.91642\n -1.84513 -1.84513\n 1.69959 1.69959\n 5.74253 5.74253\n 1.48734 1.48734\n -1.45199 -1.45199\n -18.5026 -18.5026\n -8.61009 -8.61009\n -2.21845 -2.21845\n -4.22932 -4.22932\n 6.0436 6.0436\n -1.8824 -1.8824\n -0.689965 -0.689965\n 0.845927 0.845927\n -1.99517 -1.99517\n 9.32292 9.32292\n 6.24938 6.24938\n -4.59894 -4.59894\n 6.24579 6.24579\n -5.8733 -5.8733\n -4.60285 -4.60285\n -1.27596 -1.27596\n -1.68807 -1.68807\n -0.391147 -0.391147\n -2.68362 -2.68362\n 1.99197 1.99197\n 0.0812396 0.0812396\n -3.79761 -3.79761\n -8.5693 -8.5693\n -0.869305 -0.869305\n -0.77582 -0.77582\n -4.76995 -4.76995\n 1.9712 1.9712\n 4.74459 4.74459\n -4.31244 -4.31244\n 3.94876 3.94876\n -11.0882 -11.0882\n 9.38629 9.38629\n 10.2995 10.2995\n 2.40846 2.40846\n -3.91429 -3.91429\n -0.745707 -0.745707\n 4.31946 4.31946\n 8.34836 8.34836\n 0.857636 0.857636\n 1.66563 1.66563\n -11.1522 -11.1522\n -3.48353 -3.48353\n -6.08336 -6.08336\n 1.22086 1.22086\n -2.81636 -2.81636\n 1.07224 1.07224\n -8.24909 -8.24909\n 3.66474 3.66474\n -0.260558 -0.260558\n 2.38779 2.38779\n -4.00576 -4.00576\n 1.3949 1.3949\n -5.43468 -5.43468\n 4.08836 4.08836\n -1.1134 -1.1134\n -2.05916 -2.05916\n -9.78987 -9.78987\n -2.86149 -2.86149\n 5.54577 5.54577\n -1.96682 -1.96682\n 9.70577 9.70577\n -4.0553 -4.0553\n 8.54535 8.54535\n 0.539438 0.539438\n 4.61091 4.61091\n -5.32208 -5.32208\n -0.256733 -0.256733\n 4.74966 4.74966\n -2.46464 -2.46464\n -0.223077 -0.223077\n 1.84442 1.84442\n 6.42329 6.42329\n 0.431667 0.431667\n -8.42777 -8.42777\n -10.691 -10.691\n 3.023 3.023\n -5.65345 -5.65345\n -4.17833 -4.17833\n 0.937893 0.937893\n -6.99405 -6.99405\n -4.55107 -4.55107\n -15.3169 -15.3169\n -2.08895 -2.08895\n 7.17826 7.17826\n -4.26108 -4.26108\n -3.2712 -3.2712\n 16.1561 16.1561\n 13.5164 13.5164\n -5.91778 -5.91778\n 6.3401 6.3401\n 12.7018 12.7018\n 2.04305 2.04305\n 3.81683 3.81683\n -1.39969 -1.39969\n -0.17249 -0.17249\n -16.3687 -16.3687\n 4.3827 4.3827\n 2.58974 2.58974\n -4.75363 -4.75363\n 3.36371 3.36371\n 0.986534 0.986534\n -13.4299 -13.4299\n -12.7188 -12.7188\n 2.83107 2.83107\n -3.41115 -3.41115\n -3.01015 -3.01015\n 6.40446 6.40446\n -0.186923 -0.186923\n -1.42502 -1.42502\n 2.85606 2.85606\n -0.579786 -0.579786\n -3.92704 -3.92704\n 8.28959 8.28959\n 5.42878 5.42878\n 5.71589 5.71589\n -6.78065 -6.78065\n -0.403687 -0.403687\n -1.20623 -1.20623\n 4.92372 4.92372\n -1.69266 -1.69266\n -0.103872 -0.103872\n 1.9163 1.9163\n -2.26831 -2.26831\n -7.64622 -7.64622\n 1.02228 1.02228\n 2.91952 2.91952\n -0.524167 -0.524167\n 12.4803 12.4803\n 7.36984 7.36984\n -7.46027 -7.46027\n -2.78773 -2.78773\n 2.68293 2.68293\n -0.320891 -0.320891\n 7.12037 7.12037\n 3.02726 3.02726\n -2.68363 -2.68363\n 4.78372 4.78372\n 3.68899 3.68899\n 2.08839 2.08839\n 3.1873 3.1873\n -6.10744 -6.10744\n 10.5419 10.5419\n 6.29439 6.29439\n -9.41221 -9.41221\n -2.50548 -2.50548\n -1.14 -1.14\n -3.0203 -3.0203\n -1.73182 -1.73182\n -0.97194 -0.97194\n -6.69084 -6.69084\n -1.08986 -1.08986\n -3.83631 -3.83631\n 2.2775 2.2775\n -6.91276 -6.91276\n 2.4557 2.4557\n -0.477723 -0.477723\n -4.10405 -4.10405\n -3.91437 -3.91437\n -7.79672 -7.79672\n -6.19691 -6.19691\n 0.356732 0.356732\n 0.609725 0.609725\n -3.08225 -3.08225\n 6.39968 6.39968\n 1.30207 1.30207\n 7.36038 7.36038\n -7.7581 -7.7581\n -6.303 -6.303\n 0.348147 0.348147\n -8.38124 -8.38124\n 8.68524 8.68524\n -0.873688 -0.873688\n 1.19612 1.19612\n 0.725645 0.725645\n -6.59284 -6.59284\n -6.59079 -6.59079\n 1.03175 1.03175\n -0.236469 -0.236469\n 5.01671 5.01671\n 0.752329 0.752329\n 5.39971 5.39971\n 0.826802 0.826802\n 9.38285 9.38285\n 5.85717 5.85717\n 1.71145 1.71145\n -1.36528 -1.36528\n -5.09575 -5.09575\n 7.23996 7.23996\n 12.7272 12.7272\n 2.86673 2.86673\n 2.86546 2.86546\n 1.2423 1.2423\n 6.05857 6.05857\n 9.40879 9.40879\n 1.47573 1.47573\n 8.19025 8.19025\n 12.5009 12.5009\n -4.57244 -4.57244\n -0.674127 -0.674127\n 0.416418 0.416418\n -5.23336 -5.23336\n -0.771443 -0.771443\n 4.72784 4.72784\n -4.9684 -4.9684\n 4.75989 4.75989\n 1.68141 1.68141\n -3.2264 -3.2264\n 2.67195 2.67195\n 0.424227 0.424227\n 3.5195 3.5195\n 2.22441 2.22441\n -2.4856 -2.4856\n 8.03468 8.03468\n 8.54339 8.54339\n 3.83506 3.83506\n 13.5693 13.5693\n 2.44909 2.44909\n 2.70572 2.70572\n 6.13746 6.13746\n 1.26651 1.26651\n 8.25694 8.25694\n -3.59258 -3.59258\n 3.77765 3.77765\n -0.144755 -0.144755\n 3.15706 3.15706\n -2.3952 -2.3952\n 9.82079 9.82079\n 8.94186 8.94186\n -1.83071 -1.83071\n 1.45764 1.45764\n -11.8258 -11.8258\n -0.737553 -0.737553\n -1.2382 -1.2382\n 1.83341 1.83341\n -2.75977 -2.75977\n 3.75117 3.75117\n 6.04452 6.04452\n -4.40271 -4.40271\n -8.82336 -8.82336\n 10.8513 10.8513\n -4.91857 -4.91857\n -5.7401 -5.7401\n 7.22234 7.22234\n 7.15112 7.15112\n 1.81187 1.81187\n 8.19917 8.19917\n 2.91605 2.91605\n 3.82883 3.82883\n -0.208109 -0.208109\n 1.33796 1.33796\n 5.69606 5.69606\n -2.19266 -2.19266\n -5.91177 -5.91177\n 7.25269 7.25269\n -8.65987 -8.65987\n -3.47799 -3.47799\n -10.4904 -10.4904\n -0.00963959 -0.00963959\n -6.81662 -6.81662\n -2.05566 -2.05566\n 2.10144 2.10144\n 2.58138 2.58138\n 2.03289 2.03289\n -6.43532 -6.43532\n -2.97225 -2.97225\n -4.71142 -4.71142\n 4.97199 4.97199\n 3.687 3.687\n 1.8587 1.8587\n -0.444899 -0.444899\n -1.05556 -1.05556\n 4.15926 4.15926\n 5.48777 5.48777\n 2.28346 2.28346\n -4.69401 -4.69401\n 1.8873 1.8873\n -2.62671 -2.62671\n 1.4144 1.4144\n -2.97535 -2.97535\n 0.759131 0.759131\n 5.75781 5.75781\n -5.13309 -5.13309\n 1.72701 1.72701\n 2.96653 2.96653\n -10.8087 -10.8087\n 1.07262 1.07262\n -5.80018 -5.80018\n 1.90592 1.90592\n -5.42958 -5.42958\n 8.74889 8.74889\n -3.19785 -3.19785\n -2.7096 -2.7096\n 7.44399 7.44399\n -8.7433 -8.7433\n 11.6667 11.6667\n 2.59703 2.59703\n 4.22273 4.22273\n -4.68793 -4.68793\n -4.44601 -4.44601\n -0.57319 -0.57319\n 6.63389 6.63389\n -9.14857 -9.14857\n -1.34147 -1.34147\n 7.78513 7.78513\n -4.87331 -4.87331\n -5.06022 -5.06022\n 3.13076 3.13076\n -3.49373 -3.49373\n 3.12637 3.12637\n 0.566696 0.566696\n 4.99319 4.99319\n 3.57986 3.57986\n 0.607679 0.607679\n 2.37633 2.37633\n 0.35097 0.35097\n 0.239089 0.239089\n -6.51449 -6.51449\n -3.18838 -3.18838\n 0.770256 0.770256\n 2.09481 2.09481\n 5.36062 5.36062\n -5.25216 -5.25216\n -6.9523 -6.9523\n 3.97384 3.97384\n 8.7784 8.7784\n -3.91837 -3.91837\n -9.08965 -9.08965\n -1.17883 -1.17883\n -4.21353 -4.21353\n -5.0915 -5.0915\n 3.74499 3.74499\n -4.39715 -4.39715\n 2.13732 2.13732\n 5.97568 5.97568\n 1.11809 1.11809\n -3.93191 -3.93191\n -1.39764 -1.39764\n -4.23595 -4.23595\n 0.103914 0.103914\n -2.34387 -2.34387\n -4.95433 -4.95433\n 3.58645 3.58645\n 0.818317 0.818317\n 6.23266 6.23266\n -5.62973 -5.62973\n -7.45604 -7.45604\n 1.29222 1.29222\n 0.327714 0.327714\n 5.31996 5.31996\n -2.23663 -2.23663\n 0.058689 0.058689\n -0.74368 -0.74368\n -1.20749 -1.20749\n -4.75414 -4.75414\n 2.10011 2.10011\n -6.86479 -6.86479\n 1.58403 1.58403\n 0.0492497 0.0492497\n 0.32083 0.32083\n -3.11682 -3.11682\n 4.61797 4.61797\n -0.399561 -0.399561\n -7.89927 -7.89927\n -0.659676 -0.659676\n -2.2416 -2.2416\n 0.933026 0.933026\n 1.98848 1.98848\n -2.14547 -2.14547\n -1.10747 -1.10747\n 8.90983 8.90983\n -3.84128 -3.84128\n 9.82771 9.82771\n 3.02843 3.02843\n 3.26396 3.26396\n 6.75629 6.75629\n 0.0290972 0.0290972\n 7.92768 7.92768\n 7.44608 7.44608\n -4.14083 -4.14083\n -1.39636 -1.39636\n 2.87656 2.87656\n 3.87446 3.87446\n 0.112521 0.112521\n -3.3429 -3.3429\n -6.85823 -6.85823\n 1.18408 1.18408\n 3.53175 3.53175\n 3.56147 3.56147\n 5.41961 5.41961\n -1.5263 -1.5263\n 3.05559 3.05559\n -5.7201 -5.7201\n -3.98882 -3.98882\n -0.131939 -0.131939\n 6.25683 6.25683\n 0.712945 0.712945\n 4.17266 4.17266\n 9.04425 9.04425\n -2.39179 -2.39179\n 3.03807 3.03807\n 5.79693 5.79693\n -5.28875 -5.28875\n -2.56482 -2.56482\n -1.00679 -1.00679\n -0.512488 -0.512488\n -4.60373 -4.60373\n -2.69188 -2.69188\n 0.958182 0.958182\n -1.08075 -1.08075\n 2.66033 2.66033\n -5.77563 -5.77563\n 5.393 5.393\n 0.822122 0.822122\n 3.50281 3.50281\n -1.90373 -1.90373\n -3.41986 -3.41986\n -7.32502 -7.32502\n -2.0256 -2.0256\n -6.28488 -6.28488\n 0.358393 0.358393\n 1.89312 1.89312\n -0.709162 -0.709162\n -4.43491 -4.43491\n -3.56097 -3.56097\n -8.3806 -8.3806\n -5.56256 -5.56256\n -3.40994 -3.40994\n -6.15002 -6.15002\n 0.949459 0.949459\n 3.18256 3.18256\n 6.31834 6.31834\n 12.4998 12.4998\n -6.16927 -6.16927\n -1.73781 -1.73781\n 0.274813 0.274813\n 7.11001 7.11001\n 6.79962 6.79962\n 2.00121 2.00121\n -4.30592 -4.30592\n -2.38345 -2.38345\n 7.50502 7.50502\n -3.56375 -3.56375\n -1.07828 -1.07828\n 7.4632 7.4632\n -5.78317 -5.78317\n -0.54432 -0.54432\n 8.82699 8.82699\n -2.51939 -2.51939\n -3.21417 -3.21417\n 3.06052 3.06052\n -0.45856 -0.45856\n 8.89456 8.89456\n 5.89006 5.89006\n 1.01204 1.01204\n 4.9875 4.9875\n -1.63 -1.63\n 1.35424 1.35424\n 3.72608 3.72608\n -8.53795 -8.53795\n -5.93051 -5.93051\n -2.35685 -2.35685\n 3.51823 3.51823\n 3.65767 3.65767\n -3.04233 -3.04233\n -1.12453 -1.12453\n -1.68299 -1.68299\n -5.69175 -5.69175\n 3.66601 3.66601\n -3.11779 -3.11779\n -0.20161 -0.20161\n 0.78317 0.78317\n 2.28035 2.28035\n -4.43493 -4.43493\n 2.12557 2.12557\n 6.97219 6.97219\n 4.91357 4.91357\n -1.87778 -1.87778\n 1.98163 1.98163\n 1.01184 1.01184\n 0.0544142 0.0544142\n -0.748318 -0.748318\n 10.0677 10.0677\n -5.50226 -5.50226\n 3.89987 3.89987\n 1.38136 1.38136\n 4.67073 4.67073\n 5.3372 5.3372\n -1.29886 -1.29886\n -0.965173 -0.965173\n 0.546909 0.546909\n 5.87692 5.87692\n -10.1356 -10.1356\n 0.541422 0.541422\n 0.486656 0.486656\n 8.42395 8.42395\n -4.04554 -4.04554\n 11.4728 11.4728\n -6.54655 -6.54655\n 6.90602 6.90602\n -13.8383 -13.8383\n 2.64142 2.64142\n 3.96547 3.96547\n -0.887154 -0.887154\n 0.0442338 0.0442338\n -5.12331 -5.12331\n 4.95632 4.95632\n 3.15264 3.15264\n 4.80494 4.80494\n -5.42313 -5.42313\n -4.2795 -4.2795\n 1.661 1.661\n 3.85204 3.85204\n 10.1308 10.1308\n -4.34526 -4.34526\n -5.49571 -5.49571\n 3.92939 3.92939\n -3.28527 -3.28527\n 0.154911 0.154911\n -3.606 -3.606\n 5.91814 5.91814\n -8.85249 -8.85249\n 9.38796 9.38796\n -0.800741 -0.800741\n -2.87508 -2.87508\n 2.99955 2.99955\n -7.13252 -7.13252\n -6.77081 -6.77081\n -2.28359 -2.28359\n -0.180517 -0.180517\n 7.04622 7.04622\n 4.2577 4.2577\n -4.73655 -4.73655\n -0.249759 -0.249759\n 2.4412 2.4412\n 8.47175 8.47175\n -3.24927 -3.24927\n -12.5242 -12.5242\n -2.74845 -2.74845\n -9.32786 -9.32786\n 4.21624 4.21624\n 2.94687 2.94687\n 3.35216 3.35216\n -3.5485 -3.5485\n 6.97298 6.97298\n 2.01617 2.01617\n 4.70745 4.70745\n 2.96924 2.96924\n -0.18365 -0.18365\n -0.694247 -0.694247\n -7.14459 -7.14459\n 5.38548 5.38548\n 2.04923 2.04923\n -5.33216 -5.33216\n 5.47927 5.47927\n 0.357422 0.357422\n 4.36552 4.36552\n 6.88375 6.88375\n -6.47244 -6.47244\n -3.40726 -3.40726\n -6.56449 -6.56449\n 6.34818 6.34818\n -4.23984 -4.23984\n -11.1113 -11.1113\n 2.41915 2.41915\n 3.90153 3.90153\n -7.69422 -7.69422\n -8.03709 -8.03709\n -9.64719 -9.64719\n -4.04416 -4.04416\n 2.64435 2.64435\n 5.11566 5.11566\n -1.27873 -1.27873\n -1.01265 -1.01265\n -8.38716 -8.38716\n -0.960571 -0.960571\n 2.05458 2.05458\n -1.89606 -1.89606\n -7.04401 -7.04401\n 4.91798 4.91798\n 2.12484 2.12484\n 2.38768 2.38768\n 7.9691 7.9691\n -1.00886 -1.00886\n -4.9569 -4.9569\n -4.74278 -4.74278\n 0.191814 0.191814\n -5.2925 -5.2925\n -1.15484 -1.15484\n 2.27898 2.27898\n 4.12308 4.12308\n -6.18988 -6.18988\n 7.1232 7.1232\n -6.68678 -6.68678\n 1.65808 1.65808\n 8.53283 8.53283\n 0.509069 0.509069\n -3.03638 -3.03638\n -4.86641 -4.86641\n 7.20729 7.20729\n -7.51236 -7.51236\n 3.37738 3.37738\n -0.0649395 -0.0649395\n 2.75749 2.75749\n -5.61535 -5.61535\n 3.1237 3.1237\n -0.766488 -0.766488\n 4.39047 4.39047\n 1.28616 1.28616\n -8.02003 -8.02003\n 4.21688 4.21688\n -2.79942 -2.79942\n -5.80171 -5.80171\n 9.97235 9.97235\n 21.8011 21.8011\n -3.58992 -3.58992\n 5.03481 5.03481\n -2.1684 -2.1684\n -5.46844 -5.46844\n 1.57702 1.57702\n -4.53923 -4.53923\n -1.77363 -1.77363\n -0.489051 -0.489051\n -0.371992 -0.371992\n 8.264 8.264\n 1.63502 1.63502\n -1.10134 -1.10134\n 4.76612 4.76612\n 5.93085 5.93085\n -2.07348 -2.07348\n 4.26074 4.26074\n 4.1331 4.1331\n 11.1442 11.1442\n 2.18824 2.18824\n 2.18854 2.18854\n 0.210843 0.210843\n -9.30743 -9.30743\n 5.34539 5.34539\n -4.21419 -4.21419\n -3.97284 -3.97284\n -2.67745 -2.67745\n 4.17366 4.17366\n 2.41498 2.41498\n 0.801359 0.801359\n 8.35766 8.35766\n -1.29589 -1.29589\n -7.45531 -7.45531\n -7.26731 -7.26731\n 4.06669 4.06669\n -2.35771 -2.35771\n -8.73174 -8.73174\n -0.837329 -0.837329\n -2.53419 -2.53419\n 44.3977 44.3977\n 13.5049 13.5049\n -3.66878 -3.66878\n -6.5533 -6.5533\n -5.59814 -5.59814\n -10.5759 -10.5759\n 0.663108 0.663108\n -3.45147 -3.45147\n -3.75944 -3.75944\n 1.84721 1.84721\n -0.363204 -0.363204\n 4.54678 4.54678\n 2.07408 2.07408\n 7.85227 7.85227\n -7.53707 -7.53707\n 4.18344 4.18344\n -1.96048 -1.96048\n 6.24217 6.24217\n -9.16295 -9.16295\n 0.0480544 0.0480544\n 2.84725 2.84725\n 1.08008 1.08008\n -0.874464 -0.874464\n 1.67428 1.67428\n -1.91245 -1.91245\n 3.53596 3.53596\n 3.75983 3.75983\n 1.37903 1.37903\n -0.799744 -0.799744\n 2.75015 2.75015\n -11.0835 -11.0835\n -1.6781 -1.6781\n 2.86463 2.86463\n -11.1467 -11.1467\n -3.76398 -3.76398\n 9.06439 9.06439\n 9.84403 9.84403\n -5.07 -5.07\n 3.2952 3.2952\n -1.62527 -1.62527\n -7.98997 -7.98997\n -7.8193 -7.8193\n 1.10895 1.10895\n 0.460921 0.460921\n -1.47816 -1.47816\n 0.718936 0.718936\n -3.74006 -3.74006\n -2.87535 -2.87535\n 0.037427 0.037427\n -4.49959 -4.49959\n 0.0987492 0.0987492\n 1.8443 1.8443\n 0.748879 0.748879\n 1.4364 1.4364\n -0.90809 -0.90809\n -1.36403 -1.36403\n -1.27123 -1.27123\n 3.09447 3.09447\n -3.82708 -3.82708\n 0.683696 0.683696\n 3.96997 3.96997\n 0.461267 0.461267\n 4.96801 4.96801\n -5.96169 -5.96169\n 2.56714 2.56714\n -10.7519 -10.7519\n -3.39381 -3.39381\n 1.15623 1.15623\n -3.95798 -3.95798\n -1.42797 -1.42797\n 4.85734 4.85734\n -4.46424 -4.46424\n -11.9172 -11.9172\n 0.740766 0.740766\n -2.06857 -2.06857\n -1.23723 -1.23723\n -6.43373 -6.43373\n 7.04893 7.04893\n -1.10208 -1.10208\n -0.0507102 -0.0507102\n 8.23443 8.23443\n -1.71378 -1.71378\n 2.769 2.769\n 9.77752 9.77752\n 0.423859 0.423859\n 0.901832 0.901832\n 0.0738559 0.0738559\n -0.487266 -0.487266\n 2.05358 2.05358\n -8.73912 -8.73912\n 3.01532 3.01532\n -0.926127 -0.926127\n -11.2315 -11.2315\n 1.79698 1.79698\n -13.074 -13.074\n 3.72342 3.72342\n -9.17341 -9.17341\n 7.23722 7.23722\n 3.85919 3.85919\n -4.10267 -4.10267\n 5.89157 5.89157\n -1.06631 -1.06631\n -2.18366 -2.18366\n -0.0316413 -0.0316413\n -8.63864 -8.63864\n -0.194451 -0.194451\n 2.71759 2.71759\n -5.19424 -5.19424\n -16.7634 -16.7634\n 5.97943 5.97943\n 0.319596 0.319596\n -10.0687 -10.0687\n 1.12736 1.12736\n 2.11687 2.11687\n 2.5643 2.5643\n 0.502174 0.502174\n -5.75011 -5.75011\n -11.1808 -11.1808\n -3.42246 -3.42246\n 7.55982 7.55982\n -5.85592 -5.85592\n 1.22363 1.22363\n 1.39871 1.39871\n 3.35581 3.35581\n 2.99389 2.99389\n -0.762194 -0.762194\n 1.39891 1.39891\n -4.24295 -4.24295\n -6.95612 -6.95612\n 7.00699 7.00699\n -30.893 -30.893\n -7.3071 -7.3071\n 17.5017 17.5017\n -3.26283 -3.26283\n -4.13569 -4.13569\n 4.33006 4.33006\n -5.94055 -5.94055\n -0.564017 -0.564017\n 5.60949 5.60949\n 7.50747 7.50747\n -4.08147 -4.08147\n 4.08671 4.08671\n 6.72008 6.72008\n -5.02883 -5.02883\n -3.48779 -3.48779\n 4.76881 4.76881\n 4.5818 4.5818\n -3.10608 -3.10608\n -5.08198 -5.08198\n -5.54477 -5.54477\n -13.1989 -13.1989\n -8.63604 -8.63604\n -0.688683 -0.688683\n -2.34276 -2.34276\n -3.19008 -3.19008\n 0.204818 0.204818\n 0.639057 0.639057\n 12.6767 12.6767\n -3.40057 -3.40057\n -6.36799 -6.36799\n 3.7564 3.7564\n -3.04825 -3.04825\n -3.98011 -3.98011\n -2.21944 -2.21944\n 8.40757 8.40757\n -5.6418 -5.6418\n 3.3001 3.3001\n -0.678107 -0.678107\n -2.42254 -2.42254\n 0.439524 0.439524\n -0.417505 -0.417505\n -4.98938 -4.98938\n -6.34015 -6.34015\n -4.84203 -4.84203\n 2.86778 2.86778\n 3.29409 3.29409\n 2.59772 2.59772\n 5.20187 5.20187\n 3.55625 3.55625\n -7.065 -7.065\n -6.60792 -6.60792\n -3.20259 -3.20259\n 0.417062 0.417062\n -2.39846 -2.39846\n -5.762 -5.762\n 1.74843 1.74843\n 8.19239 8.19239\n -1.7349 -1.7349\n -0.0331415 -0.0331415\n 5.00712 5.00712\n 10.611 10.611\n 9.28817 9.28817\n -3.85324 -3.85324\n 2.29622 2.29622\n 10.962 10.962\n 4.44034 4.44034\n -3.2265 -3.2265\n 1.39326 1.39326\n -1.56539 -1.56539\n -8.78843 -8.78843\n -1.74101 -1.74101\n 8.51953 8.51953\n 3.31178 3.31178\n -1.20051 -1.20051\n -3.93224 -3.93224\n 2.4431 2.4431\n 3.69278 3.69278\n -10.2714 -10.2714\n -13.7579 -13.7579\n -1.76844 -1.76844\n -0.448193 -0.448193\n 1.48574 1.48574\n -0.831377 -0.831377\n 6.42657 6.42657\n 6.51848 6.51848\n 2.7764 2.7764\n 4.29448 4.29448\n -1.27173 -1.27173\n -7.14856 -7.14856\n 2.95751 2.95751\n 2.39789 2.39789\n 4.79429 4.79429\n 7.29216 7.29216\n -4.91502 -4.91502\n 2.38701 2.38701\n -2.34997 -2.34997\n -0.876115 -0.876115\n -0.672649 -0.672649\n 4.43884 4.43884\n 0.254258 0.254258\n -3.56471 -3.56471\n 0.161779 0.161779\n -10.1128 -10.1128\n 9.97279 9.97279\n -5.01498 -5.01498\n 1.10415 1.10415\n 1.37993 1.37993\n -3.32619 -3.32619\n 2.57257 2.57257\n -0.137478 -0.137478\n 1.49426 1.49426\n -0.805644 -0.805644\n 3.25356 3.25356\n 2.46332 2.46332\n 1.39266 1.39266\n 4.15167 4.15167\n -9.27164 -9.27164\n -2.29794 -2.29794\n 0.067971 0.067971\n 3.83697 3.83697\n 5.7385 5.7385\n -6.15176 -6.15176\n -4.08442 -4.08442\n -6.18563 -6.18563\n 6.44396 6.44396\n 5.63585 5.63585\n 1.21604 1.21604\n 11.1837 11.1837\n 2.29144 2.29144\n -0.995473 -0.995473\n 5.22826 5.22826\n 9.27205 9.27205\n -7.23457 -7.23457\n 6.29887 6.29887\n 2.48343 2.48343\n -4.96111 -4.96111\n -5.52811 -5.52811\n -4.40855 -4.40855\n -5.69429 -5.69429\n -1.12765 -1.12765\n -0.22142 -0.22142\n -5.96815 -5.96815\n 4.55923 4.55923\n -1.05719 -1.05719\n 2.07986 2.07986\n 7.77539 7.77539\n -2.03581 -2.03581\n 0.270705 0.270705\n 0.126658 0.126658\n -6.1672 -6.1672\n -16.0576 -16.0576\n 0.635198 0.635198\n 8.55006 8.55006\n -2.93081 -2.93081\n -1.7657 -1.7657\n -0.37886 -0.37886\n -2.4086 -2.4086\n 1.41889 1.41889\n -1.40539 -1.40539\n 0.963807 0.963807\n -2.14947 -2.14947\n -6.31832 -6.31832\n -4.30827 -4.30827\n 6.2609 6.2609\n -8.36351 -8.36351\n 4.28564 4.28564\n 0.646361 0.646361\n 4.60485 4.60485\n -3.1664 -3.1664\n -0.611618 -0.611618\n -9.53534 -9.53534\n 1.92275 1.92275\n -8.1521 -8.1521\n 0.101441 0.101441\n 0.399002 0.399002\n -2.04551 -2.04551\n -4.5564 -4.5564\n 3.0555 3.0555\n 0.992401 0.992401\n -5.62638 -5.62638\n -0.46873 -0.46873\n -6.86208 -6.86208\n -2.77108 -2.77108\n 3.51118 3.51118\n 0.885266 0.885266\n 3.65701 3.65701\n 6.88336 6.88336\n -7.25948 -7.25948\n 7.31435 7.31435\n -6.57357 -6.57357\n 3.67947 3.67947\n 4.80901 4.80901\n -2.80342 -2.80342\n 5.78724 5.78724\n 5.30985 5.30985\n 7.24724 7.24724\n -1.30439 -1.30439\n 2.50975 2.50975\n 5.28538 5.28538\n -3.91583 -3.91583\n 2.98722 2.98722\n 5.31167 5.31167\n -0.596966 -0.596966\n -4.94141 -4.94141\n 4.59005 4.59005\n 1.3813 1.3813\n 4.0611 4.0611\n -0.747616 -0.747616\n -3.1697 -3.1697\n -1.70787 -1.70787\n -2.43542 -2.43542\n -5.86823 -5.86823\n -10.9093 -10.9093\n 5.20087 5.20087\n -6.40378 -6.40378\n 1.5149 1.5149\n -6.52874 -6.52874\n -5.69743 -5.69743\n 1.06819 1.06819\n -7.31776 -7.31776\n 3.69649 3.69649\n -4.21319 -4.21319\n -4.91507 -4.91507\n 5.44776 5.44776\n -0.708927 -0.708927\n 1.94895 1.94895\n 2.90927 2.90927\n -2.82547 -2.82547\n -1.79858 -1.79858\n -15.6727 -15.6727\n -0.308918 -0.308918\n 2.61943 2.61943\n -3.89041 -3.89041\n -1.84684 -1.84684\n -6.80446 -6.80446\n 3.97398 3.97398\n 2.31201 2.31201\n 4.29417 4.29417\n -1.24479 -1.24479\n 4.25927 4.25927\n -1.96968 -1.96968\n 0.703519 0.703519\n 2.06517 2.06517\n 0.920347 0.920347\n 6.22843 6.22843\n 1.86167 1.86167\n 0.43407 0.43407\n 1.25225 1.25225\n -0.00512493 -0.00512493\n -1.70887 -1.70887\n 0.725693 0.725693\n 6.11604 6.11604\n -5.87059 -5.87059\n 3.26102 3.26102\n 2.0488 2.0488\n -0.0544172 -0.0544172\n 2.57295 2.57295\n -1.10578 -1.10578\n 2.43904 2.43904\n -2.34604 -2.34604\n 3.2098 3.2098\n 2.16089 2.16089\n -9.35001 -9.35001\n 9.43924 9.43924\n 0.916747 0.916747\n 2.59533 2.59533\n -1.84596 -1.84596\n 1.02889 1.02889\n 0.755944 0.755944\n 8.28274 8.28274\n -3.21136 -3.21136\n 1.24897 1.24897\n -0.363928 -0.363928\n 2.37533 2.37533\n -1.5794 -1.5794\n 6.67417 6.67417\n -4.4632 -4.4632\n 8.53731 8.53731\n -1.16526 -1.16526\n -0.51467 -0.51467\n -4.91688 -4.91688\n 7.17741 7.17741\n 4.61708 4.61708\n -2.41511 -2.41511\n -11.5234 -11.5234\n 2.61523 2.61523\n 4.7703 4.7703\n 6.72381 6.72381\n 5.65388 5.65388\n -4.23963 -4.23963\n 0.925176 0.925176\n 1.98862 1.98862\n -6.14466 -6.14466\n 2.76728 2.76728\n -0.83598 -0.83598\n -4.22593 -4.22593\n 5.99083 5.99083\n -4.886 -4.886\n 4.37801 4.37801\n 5.77761 5.77761\n 3.38352 3.38352\n -0.311291 -0.311291\n 8.26669 8.26669\n -4.94787 -4.94787\n -9.62034 -9.62034\n 2.37023 2.37023\n 3.41718 3.41718\n -2.43368 -2.43368\n 3.5898 3.5898\n -1.21973 -1.21973\n 0.0350305 0.0350305\n -4.33097 -4.33097\n -3.41432 -3.41432\n 2.59161 2.59161\n -2.11239 -2.11239\n -1.0801 -1.0801\n -3.27061 -3.27061\n -0.34025 -0.34025\n -6.40563 -6.40563\n -0.522305 -0.522305\n 4.63382 4.63382\n 1.5154 1.5154\n 0.968893 0.968893\n 2.79354 2.79354\n -0.829942 -0.829942\n -1.76388 -1.76388\n -6.64903 -6.64903\n -8.52588 -8.52588\n 2.70798 2.70798\n 6.78381 6.78381\n -5.67891 -5.67891\n -0.0588557 -0.0588557\n -4.12923 -4.12923\n -2.70431 -2.70431\n -0.12131 -0.12131\n 6.59494 6.59494\n 0.830427 0.830427\n 3.40436 3.40436\n 6.98828 6.98828\n -2.33332 -2.33332\n 5.85244 5.85244\n -10.0398 -10.0398\n -0.242519 -0.242519\n -3.38719 -3.38719\n 2.74288 2.74288\n 3.82961 3.82961\n -6.85166 -6.85166\n -0.345431 -0.345431\n -3.03082 -3.03082\n 1.68089 1.68089\n -0.785036 -0.785036\n -2.92804 -2.92804\n 1.03727 1.03727\n 5.51647 5.51647\n -2.15538 -2.15538\n -6.20918 -6.20918\n -0.986195 -0.986195\n -4.4207 -4.4207\n -0.314791 -0.314791\n -6.64843 -6.64843\n 1.255 1.255\n 4.39107 4.39107\n 2.20706 2.20706\n -1.894 -1.894\n -3.01471 -3.01471\n -0.0623641 -0.0623641\n -5.76316 -5.76316\n -2.45987 -2.45987\n -2.09262 -2.09262\n 0.0458748 0.0458748\n 5.09539 5.09539\n -3.80431 -3.80431\n -3.90738 -3.90738\n -6.48843 -6.48843\n -2.58373 -2.58373\n -6.38764 -6.38764\n 7.38858 7.38858\n 0.492176 0.492176\n 8.79347 8.79347\n 2.04442 2.04442\n -0.216083 -0.216083\n 11.3375 11.3375\n -3.4177 -3.4177\n 3.90111 3.90111\n 4.92081 4.92081\n 4.45964 4.45964\n 11.1458 11.1458\n -2.2688 -2.2688\n -4.43463 -4.43463\n -4.22186 -4.22186\n -5.93987 -5.93987\n 3.4437 3.4437\n -5.60816 -5.60816\n -8.04401 -8.04401\n -4.95256 -4.95256\n 3.88283 3.88283\n -0.173935 -0.173935\n -2.63243 -2.63243\n -1.03812 -1.03812\n -9.14078 -9.14078\n -6.1411 -6.1411\n 3.4284 3.4284\n -9.8305 -9.8305\n 6.76115 6.76115\n -11.3646 -11.3646\n -5.7296 -5.7296\n -2.41831 -2.41831\n -5.21505 -5.21505\n 10.4347 10.4347\n 2.06721 2.06721\n 1.02265 1.02265\n -6.93537 -6.93537\n 1.28707 1.28707\n 0.939615 0.939615\n 11.262 11.262\n 1.2805 1.2805\n 4.8619 4.8619\n 3.15836 3.15836\n -5.18747 -5.18747\n -2.98078 -2.98078\n -2.0489 -2.0489\n -2.85634 -2.85634\n -4.56059 -4.56059\n -4.0715 -4.0715\n 0.469543 0.469543\n -2.05188 -2.05188\n -2.79567 -2.79567\n 3.82027 3.82027\n 2.55175 2.55175\n -0.468207 -0.468207\n -5.65994 -5.65994\n 2.13508 2.13508\n -3.17019 -3.17019\n 6.53032 6.53032\n -4.98714 -4.98714\n -1.94956 -1.94956\n -3.08465 -3.08465\n 8.11664 8.11664\n 8.86283 8.86283\n 0.84108 0.84108\n 5.22353 5.22353\n -3.45671 -3.45671\n -1.38725 -1.38725\n 1.35206 1.35206\n -10.4407 -10.4407\n -2.20051 -2.20051\n -0.228019 -0.228019\n -1.38039 -1.38039\n 11.1342 11.1342\n 5.17568 5.17568\n -4.54852 -4.54852\n -1.26392 -1.26392\n 5.69792 5.69792\n -4.90866 -4.90866\n 2.84526 2.84526\n 10.9699 10.9699\n 12.9756 12.9756\n 8.48223 8.48223\n 2.11902 2.11902\n 3.74471 3.74471\n -5.14437 -5.14437\n -14.7206 -14.7206\n 3.01028 3.01028\n -2.67988 -2.67988\n -2.88296 -2.88296\n -4.95895 -4.95895\n -1.82286 -1.82286\n 5.23419 5.23419\n -2.23867 -2.23867\n 0.610838 0.610838\n 2.09177 2.09177\n 5.74677 5.74677\n 3.6242 3.6242\n 2.0758 2.0758\n -2.85159 -2.85159\n -3.93562 -3.93562\n 3.85649 3.85649\n -5.75638 -5.75638\n -7.07444 -7.07444\n 0.907402 0.907402\n -8.92532 -8.92532\n -4.09782 -4.09782\n 1.85777 1.85777\n 5.73041 5.73041\n -2.17118 -2.17118\n -3.4713 -3.4713\n 7.95825 7.95825\n 9.10838 9.10838\n 1.80182 1.80182\n -0.54593 -0.54593\n -4.89919 -4.89919\n -2.97982 -2.97982\n 0.807424 0.807424\n -2.27 -2.27\n -13.2338 -13.2338\n -3.94367 -3.94367\n -5.72938 -5.72938\n -2.42243 -2.42243\n -3.69581 -3.69581\n -4.71307 -4.71307\n 1.38983 1.38983\n -5.37869 -5.37869\n -6.82815 -6.82815\n 2.73203 2.73203\n 13.6495 13.6495\n -6.29731 -6.29731\n -8.43712 -8.43712\n 14.1567 14.1567\n -0.978804 -0.978804\n 1.26264 1.26264\n -9.25575 -9.25575\n -8.10968 -8.10968\n -3.98015 -3.98015\n 6.60273 6.60273\n -3.98373 -3.98373\n 1.35817 1.35817\n 1.20988 1.20988\n 1.53069 1.53069\n 4.08368 4.08368\n -2.38429 -2.38429\n -4.67381 -4.67381\n -5.49726 -5.49726\n 0.657715 0.657715\n -0.00123905 -0.00123905\n 4.62712 4.62712\n -0.317445 -0.317445\n -5.08829 -5.08829\n -9.85674 -9.85674\n 5.31787 5.31787\n 1.61793 1.61793\n 3.9901 3.9901\n -1.04243 -1.04243\n -3.73679 -3.73679\n 0.670282 0.670282\n 9.03148 9.03148\n -4.77058 -4.77058\n 8.60147 8.60147\n -0.664744 -0.664744\n 1.97711 1.97711\n -5.35794 -5.35794\n -9.70033 -9.70033\n 10.7781 10.7781\n 1.96443 1.96443\n 1.84069 1.84069\n -12.0109 -12.0109\n 2.08404 2.08404\n 3.64031 3.64031\n 8.65585 8.65585\n -11.8355 -11.8355\n 9.89404 9.89404\n 0.279063 0.279063\n -0.315296 -0.315296\n 3.74263 3.74263\n 6.54645 6.54645\n 5.43941 5.43941\n 4.83252 4.83252\n 1.70716 1.70716\n -3.27497 -3.27497\n -3.07764 -3.07764\n 9.25309 9.25309\n -1.69559 -1.69559\n 10.1694 10.1694\n -3.42523 -3.42523\n 6.39435 6.39435\n 2.18084 2.18084\n 1.33177 1.33177\n -0.709393 -0.709393\n 1.44799 1.44799\n 0.881759 0.881759\n -2.35085 -2.35085\n -1.91407 -1.91407\n 0.302603 0.302603\n 1.40288 1.40288\n -2.37323 -2.37323\n -7.74084 -7.74084\n -7.73224 -7.73224\n 2.8793 2.8793\n 6.62065 6.62065\n 1.4654 1.4654\n -0.982735 -0.982735\n -0.97328 -0.97328\n -8.38882 -8.38882\n 8.74643 8.74643\n -7.86996 -7.86996\n 3.25655 3.25655\n 2.78551 2.78551\n -5.17511 -5.17511\n 4.90515 4.90515\n 0.28899 0.28899\n 3.57292 3.57292\n -5.25376 -5.25376\n -8.57274 -8.57274\n -1.18267 -1.18267\n 37.4072 37.4072\n -4.00801 -4.00801\n 4.8073 4.8073\n -4.45001 -4.45001\n 7.66024 7.66024\n -4.47725 -4.47725\n -10.2209 -10.2209\n -4.80026 -4.80026\n -0.64446 -0.64446\n -0.899171 -0.899171\n 1.09833 1.09833\n -0.988097 -0.988097\n 2.82126 2.82126\n -8.19269 -8.19269\n -2.64922 -2.64922\n -9.16004 -9.16004\n -2.39588 -2.39588\n -4.72025 -4.72025\n 2.34077 2.34077\n 3.83879 3.83879\n 1.9499 1.9499\n -0.361603 -0.361603\n 7.79929 7.79929\n 2.34774 2.34774\n -8.21052 -8.21052\n -2.02077 -2.02077\n -1.58017 -1.58017\n -0.410542 -0.410542\n -10.7206 -10.7206\n 3.26874 3.26874\n 2.80972 2.80972\n 0.0906836 0.0906836\n -1.64773 -1.64773\n 6.49353 6.49353\n -0.791109 -0.791109\n 4.71404 4.71404\n 0.0741314 0.0741314\n -0.414415 -0.414415\n 6.84572 6.84572\n -0.367457 -0.367457\n 1.17563 1.17563\n 0.51039 0.51039\n 4.40348 4.40348\n 0.978932 0.978932\n 3.79206 3.79206\n 4.57632 4.57632\n 2.77883 2.77883\n 0.490867 0.490867\n -0.151798 -0.151798\n 6.72243 6.72243\n 4.77773 4.77773\n -0.50633 -0.50633\n -8.08639 -8.08639\n 4.88619 4.88619\n -2.07669 -2.07669\n -2.24093 -2.24093\n 1.72994 1.72994\n -7.45157 -7.45157\n -12.1192 -12.1192\n 1.4328 1.4328\n -8.14432 -8.14432\n -6.25485 -6.25485\n 0.516865 0.516865\n 7.11864 7.11864\n -0.616318 -0.616318\n -0.761916 -0.761916\n -5.99496 -5.99496\n 10.4321 10.4321\n -0.516052 -0.516052\n -5.68287 -5.68287\n -4.15541 -4.15541\n 1.56619 1.56619\n -20.8292 -20.8292\n 0.788033 0.788033\n 3.34264 3.34264\n 3.70493 3.70493\n -0.0822138 -0.0822138\n 2.31304 2.31304\n -1.69352 -1.69352\n 2.10396 2.10396\n 7.2613 7.2613\n -1.81799 -1.81799\n -2.09968 -2.09968\n -3.8336 -3.8336\n -3.93478 -3.93478\n 3.3059 3.3059\n 4.19189 4.19189\n -1.93794 -1.93794\n 2.7117 2.7117\n 9.43261 9.43261\n -1.83318 -1.83318\n -1.12685 -1.12685\n 2.40725 2.40725\n 7.50947 7.50947\n 7.65688 7.65688\n -5.02792 -5.02792\n -2.55777 -2.55777\n -1.9946 -1.9946\n -0.126192 -0.126192\n -3.30905 -3.30905\n -0.209775 -0.209775\n 9.06409 9.06409\n -3.79201 -3.79201\n 8.80185 8.80185\n -1.59367 -1.59367\n -2.49213 -2.49213\n -3.5242 -3.5242\n 2.4892 2.4892\n 5.68222 5.68222\n 4.29073 4.29073\n 0.490494 0.490494\n 3.31313 3.31313\n 8.27344 8.27344\n 1.44936 1.44936\n 5.94283 5.94283\n -5.90497 -5.90497\n 0.316931 0.316931\n 1.93975 1.93975\n -1.33405 -1.33405\n -4.17957 -4.17957\n 2.45999 2.45999\n -10.0965 -10.0965\n 0.648564 0.648564\n -0.745957 -0.745957\n -6.08922 -6.08922\n -6.45851 -6.45851\n 2.70093 2.70093\n -2.59331 -2.59331\n -2.73319 -2.73319\n -6.50584 -6.50584\n 4.14167 4.14167\n 6.78757 6.78757\n 4.63335 4.63335\n 2.01754 2.01754\n 3.97717 3.97717\n 2.73775 2.73775\n 2.04299 2.04299\n 7.03044 7.03044\n -8.59414 -8.59414\n -4.19956 -4.19956\n 0.0135157 0.0135157\n -5.45393 -5.45393\n 2.75578 2.75578\n 0.730278 0.730278\n -0.410035 -0.410035\n 10.7831 10.7831\n -2.82537 -2.82537\n 1.85601 1.85601\n 1.68496 1.68496\n 2.75249 2.75249\n 9.40848 9.40848\n 1.6032 1.6032\n -3.91263 -3.91263\n 1.12247 1.12247\n -3.46516 -3.46516\n -1.48668 -1.48668\n 6.7676 6.7676\n -5.76927 -5.76927\n -2.19943 -2.19943\n -1.61329 -1.61329\n 3.35791 3.35791\n -7.80737 -7.80737\n 3.06567 3.06567\n -12.2037 -12.2037\n 12.541 12.541\n 4.42316 4.42316\n 6.48419 6.48419\n 1.17664 1.17664\n 2.97986 2.97986\n -8.63966 -8.63966\n 0.241757 0.241757\n -5.03654 -5.03654\n -1.94594 -1.94594\n 12.8093 12.8093\n -3.58644 -3.58644\n -3.35952 -3.35952\n -0.864134 -0.864134\n -12.4807 -12.4807\n -1.69909 -1.69909\n -5.67676 -5.67676\n -10.6435 -10.6435\n -3.86815 -3.86815\n 4.20674 4.20674\n -4.94992 -4.94992\n 7.63289 7.63289\n -5.5226 -5.5226\n 1.58362 1.58362\n 1.14864 1.14864\n 5.98635 5.98635\n 11.9692 11.9692\n -0.208588 -0.208588\n -0.177219 -0.177219\n 6.35143 6.35143\n -2.21028 -2.21028\n 0.693657 0.693657\n 2.66882 2.66882\n -0.494413 -0.494413\n 10.9482 10.9482\n 2.9522 2.9522\n 1.69427 1.69427\n -5.54007 -5.54007\n -1.44208 -1.44208\n -2.75377 -2.75377\n 7.62773 7.62773\n -0.0991657 -0.0991657\n 0.541024 0.541024\n 0.383422 0.383422\n -6.28538 -6.28538\n -3.63239 -3.63239\n 5.54891 5.54891\n 4.38377 4.38377\n -4.21607 -4.21607\n -1.58462 -1.58462\n 1.99568 1.99568\n 1.70177 1.70177\n 1.65142 1.65142\n 1.79811 1.79811\n -6.82605 -6.82605\n 3.65159 3.65159\n 2.60935 2.60935\n -2.91237 -2.91237\n -1.56808 -1.56808\n -3.07334 -3.07334\n 0.883426 0.883426\n -1.59697 -1.59697\n 4.44432 4.44432\n -2.72255 -2.72255\n -0.853149 -0.853149\n -0.132598 -0.132598\n -0.63629 -0.63629\n -3.69308 -3.69308\n -7.18449 -7.18449\n 1.20547 1.20547\n 14.3427 14.3427\n 5.08288 5.08288\n 0.957041 0.957041\n 0.153537 0.153537\n -7.14906 -7.14906\n -8.78572 -8.78572\n 4.05049 4.05049\n 3.22929 3.22929\n -3.34601 -3.34601\n 3.86442 3.86442\n -2.80641 -2.80641\n 6.51055 6.51055\n -4.58706 -4.58706\n -1.51146 -1.51146\n 3.88212 3.88212\n 1.89549 1.89549\n 3.50062 3.50062\n -1.43005 -1.43005\n -2.91969 -2.91969\n -6.52573 -6.52573\n -3.8843 -3.8843\n -8.34716 -8.34716\n -7.42192 -7.42192\n -3.98985 -3.98985\n 15.526 15.526\n 8.70318 8.70318\n -1.10105 -1.10105\n 2.14694 2.14694\n 7.71484 7.71484\n -0.0260442 -0.0260442\n -3.31138 -3.31138\n 1.67906 1.67906\n -0.083112 -0.083112\n -8.42905 -8.42905\n -8.82729 -8.82729\n 11.2859 11.2859\n -8.07136 -8.07136\n -3.9371 -3.9371\n -4.63176 -4.63176\n -1.23605 -1.23605\n -2.08565 -2.08565\n 1.93918 1.93918\n -12.5031 -12.5031\n -0.442281 -0.442281\n -5.50289 -5.50289\n -0.815112 -0.815112\n 0.0898735 0.0898735\n 4.69373 4.69373\n -7.22004 -7.22004\n 0.543294 0.543294\n 4.2932 4.2932\n 2.12984 2.12984\n -4.42752 -4.42752\n 3.03694 3.03694\n -3.73337 -3.73337\n -12.0483 -12.0483\n -5.99704 -5.99704\n 0.0707967 0.0707967\n -4.52239 -4.52239\n 3.65625 3.65625\n -5.61903 -5.61903\n 9.78971 9.78971\n 8.47575 8.47575\n -0.320966 -0.320966\n -7.10339 -7.10339\n 0.485669 0.485669\n 3.19439 3.19439\n -0.411976 -0.411976\n -0.782875 -0.782875\n 16.4086 16.4086\n -2.67312 -2.67312\n 0.73424 0.73424\n 8.32014 8.32014\n -1.24665 -1.24665\n 3.70031 3.70031\n -6.22155 -6.22155\n -6.34804 -6.34804\n -4.84631 -4.84631\n 7.19111 7.19111\n 2.58937 2.58937\n 2.12044 2.12044\n -0.304369 -0.304369\n -11.5161 -11.5161\n -4.75933 -4.75933\n -5.40287 -5.40287\n -14.7511 -14.7511\n -11.3269 -11.3269\n -3.40961 -3.40961\n -8.36998 -8.36998\n -7.86816 -7.86816\n 3.46638 3.46638\n 5.10745 5.10745\n 9.12589 9.12589\n 4.53119 4.53119\n -0.0952322 -0.0952322\n -1.67069 -1.67069\n 1.48937 1.48937\n 2.1548 2.1548\n -0.680895 -0.680895\n 6.00943 6.00943\n -6.23597 -6.23597\n 15.2635 15.2635\n -5.39621 -5.39621\n 2.9004 2.9004\n -7.2031 -7.2031\n 0.188095 0.188095\n -5.65511 -5.65511\n 8.80472 8.80472\n 4.77116 4.77116\n -0.320718 -0.320718\n -0.094774 -0.094774\n 4.24892 4.24892\n -0.729715 -0.729715\n 3.46906 3.46906\n -4.86913 -4.86913\n -2.05092 -2.05092\n 3.24008 3.24008\n 2.67334 2.67334\n 5.41008 5.41008\n 4.61387 4.61387\n -11.9338 -11.9338\n 2.15538 2.15538\n 3.39914 3.39914\n 2.71216 2.71216\n 6.79031 6.79031\n -0.750493 -0.750493\n -0.683416 -0.683416\n 7.23875 7.23875\n 4.67949 4.67949\n -2.16467 -2.16467\n 3.64787 3.64787\n -1.27823 -1.27823\n -1.43992 -1.43992\n 3.183 3.183\n -8.60412 -8.60412\n -5.42757 -5.42757\n -0.564214 -0.564214\n -1.17837 -1.17837\n 2.45248 2.45248\n 3.60909 3.60909\n 2.61183 2.61183\n 5.20279 5.20279\n -1.07145 -1.07145\n -0.919519 -0.919519\n 3.89898 3.89898\n 3.72175 3.72175\n -9.9673 -9.9673\n 1.50607 1.50607\n -0.456562 -0.456562\n 10.9984 10.9984\n -2.18673 -2.18673\n -7.39159 -7.39159\n -5.54389 -5.54389\n 2.6353 2.6353\n 6.87535 6.87535\n -10.4019 -10.4019\n -5.51375 -5.51375\n -3.33244 -3.33244\n 7.60358 7.60358\n -9.48529 -9.48529\n -0.514099 -0.514099\n 6.20569 6.20569\n -4.60198 -4.60198\n -1.28686 -1.28686\n -0.383981 -0.383981\n -0.173934 -0.173934\n -7.97782 -7.97782\n 5.9926 5.9926\n -3.7357 -3.7357\n -7.77841 -7.77841\n 3.09245 3.09245\n -3.70421 -3.70421\n -1.50012 -1.50012\n -3.90181 -3.90181\n 0.183002 0.183002\n -4.72374 -4.72374\n -3.36966 -3.36966\n 8.23642 8.23642\n 0.387898 0.387898\n -2.53048 -2.53048\n 4.46348 4.46348\n -0.932844 -0.932844\n -1.76804 -1.76804\n -0.390175 -0.390175\n 8.28101 8.28101\n 8.66959 8.66959\n 2.47585 2.47585\n 6.33837 6.33837\n 3.05846 3.05846\n 6.43047 6.43047\n 0.167477 0.167477\n 0.615034 0.615034\n -8.467 -8.467\n 2.15566 2.15566\n 6.59172 6.59172\n -8.30068 -8.30068\n -2.92268 -2.92268\n -1.14616 -1.14616\n 3.864 3.864\n -8.07267 -8.07267\n 0.382952 0.382952\n 4.79087 4.79087\n 7.87692 7.87692\n -1.27352 -1.27352\n -0.439992 -0.439992\n -0.361056 -0.361056\n 5.51463 5.51463\n 4.10827 4.10827\n -1.36056 -1.36056\n -10.9063 -10.9063\n -3.12566 -3.12566\n -1.52612 -1.52612\n 2.47429 2.47429\n 1.92973 1.92973\n 6.05399 6.05399\n 6.35717 6.35717\n -6.54112 -6.54112\n 0.16752 0.16752\n -0.581192 -0.581192\n -3.91981 -3.91981\n 3.29046 3.29046\n -9.85289 -9.85289\n -1.68008 -1.68008\n -0.294261 -0.294261\n -2.33446 -2.33446\n 8.72203 8.72203\n -7.53754 -7.53754\n 1.8548 1.8548\n 0.0863562 0.0863562\n 3.71224 3.71224\n -2.72156 -2.72156\n 6.92717 6.92717\n 4.22066 4.22066\n 2.9384 2.9384\n -0.436476 -0.436476\n 7.94505 7.94505\n 3.35167 3.35167\n 4.57606 4.57606\n -1.94551 -1.94551\n 7.26891 7.26891\n 5.7114 5.7114\n -4.8975 -4.8975\n 0.24802 0.24802\n 4.4272 4.4272\n 3.21714 3.21714\n -2.75997 -2.75997\n 3.0239 3.0239\n 6.00743 6.00743\n 1.95157 1.95157\n -8.23524 -8.23524\n -0.0388194 -0.0388194\n -1.59723 -1.59723\n -15.7227 -15.7227\n 5.01363 5.01363\n 2.59661 2.59661\n 0.344503 0.344503\n 7.85727 7.85727\n 0.142462 0.142462\n -3.54743 -3.54743\n -4.18558 -4.18558\n 3.96172 3.96172\n -0.376684 -0.376684\n 3.78763 3.78763\n -1.58384 -1.58384\n 15.837 15.837\n -0.887404 -0.887404\n 0.855016 0.855016\n 11.1701 11.1701\n 5.15206 5.15206\n 6.83176 6.83176\n -0.91331 -0.91331\n -10.3398 -10.3398\n 2.48231 2.48231\n -2.03572 -2.03572\n 1.09096 1.09096\n -0.162198 -0.162198\n -7.32758 -7.32758\n -6.97941 -6.97941\n 5.98831 5.98831\n -7.43703 -7.43703\n -8.97936 -8.97936\n 0.676949 0.676949\n 1.37291 1.37291\n 4.41159 4.41159\n 2.45643 2.45643\n 2.79374 2.79374\n 2.36712 2.36712\n -7.74483 -7.74483\n 0.602922 0.602922\n -2.48544 -2.48544\n 0.299035 0.299035\n 6.77695 6.77695\n 1.44763 1.44763\n 1.94637 1.94637\n -4.04181 -4.04181\n 16.3509 16.3509\n 6.4273 6.4273\n 5.41235 5.41235\n -5.91387 -5.91387\n -6.06301 -6.06301\n 3.4536 3.4536\n -3.39128 -3.39128\n 11.299 11.299\n 2.62685 2.62685\n 1.00866 1.00866\n 10.6766 10.6766\n -0.805083 -0.805083\n 3.91073 3.91073\n 3.67201 3.67201\n -9.14116 -9.14116\n 15.6406 15.6406\n 3.22084 3.22084\n -2.90513 -2.90513\n 4.58966 4.58966\n 0.0983211 0.0983211\n 2.35908 2.35908\n 0.658109 0.658109\n 2.37478 2.37478\n -6.70679 -6.70679\n 6.08307 6.08307\n -29.6624 -29.6624\n 1.55578 1.55578\n 5.31311 5.31311\n -5.40681 -5.40681\n 1.80228 1.80228\n 4.50431 4.50431\n 7.25673 7.25673\n 5.89811 5.89811\n -2.92888 -2.92888\n 7.48853 7.48853\n -1.67318 -1.67318\n 0.974302 0.974302\n -8.10178 -8.10178\n 3.29435 3.29435\n -1.64519 -1.64519\n -7.08854 -7.08854\n 6.68891 6.68891\n -5.69927 -5.69927\n -3.51768 -3.51768\n 11.2895 11.2895\n -0.828568 -0.828568\n 5.53562 5.53562\n -0.358066 -0.358066\n -5.92559 -5.92559\n 4.39224 4.39224\n -5.1225 -5.1225\n -9.51174 -9.51174\n 9.80076 9.80076\n -1.85858 -1.85858\n 6.95181 6.95181\n -1.71297 -1.71297\n -0.275297 -0.275297\n -0.860135 -0.860135\n -0.484906 -0.484906\n 5.71425 5.71425\n 2.74639 2.74639\n -8.40417 -8.40417\n -1.84935 -1.84935\n 2.94526 2.94526\n 10.708 10.708\n 0.892511 0.892511\n -1.36773 -1.36773\n -7.25911 -7.25911\n 3.91428 3.91428\n -0.776027 -0.776027\n 3.44102 3.44102\n -4.87806 -4.87806\n 3.65101 3.65101\n -3.01077 -3.01077\n 1.17918 1.17918\n 5.82266 5.82266\n 8.52564 8.52564\n 4.35296 4.35296\n -2.94897 -2.94897\n -4.19366 -4.19366\n -4.7939 -4.7939\n 3.44038 3.44038\n -7.87089 -7.87089\n -3.18931 -3.18931\n -6.65708 -6.65708\n 1.09687 1.09687\n -4.36662 -4.36662\n 2.90783 2.90783\n 4.66889 4.66889\n -1.26146 -1.26146\n -2.01469 -2.01469\n -2.44566 -2.44566\n -2.15098 -2.15098\n 3.4006 3.4006\n 0.0396139 0.0396139\n 2.29469 2.29469\n -7.62709 -7.62709\n 7.18738 7.18738\n 1.45481 1.45481\n 2.37791 2.37791\n -5.37208 -5.37208\n -0.0612415 -0.0612415\n -1.46115 -1.46115\n 4.29624 4.29624\n 3.25993 3.25993\n 2.42986 2.42986\n 6.56133 6.56133\n -2.07349 -2.07349\n 5.61643 5.61643\n 5.48251 5.48251\n -0.703666 -0.703666\n -5.09456 -5.09456\n 0.57249 0.57249\n 4.28577 4.28577\n 2.468 2.468\n -10.013 -10.013\n -3.26046 -3.26046\n -7.91038 -7.91038\n -2.03302 -2.03302\n 3.49234 3.49234\n -1.2481 -1.2481\n -1.87417 -1.87417\n -1.93016 -1.93016\n 2.14307 2.14307\n -9.0722 -9.0722\n 2.03124 2.03124\n -0.938906 -0.938906\n 0.817464 0.817464\n 2.23636 2.23636\n 1.3076 1.3076\n 4.90629 4.90629\n 2.16603 2.16603\n 5.84398 5.84398\n -6.56748 -6.56748\n 7.22968 7.22968\n 0.664381 0.664381\n 11.2001 11.2001\n -4.98902 -4.98902\n 0.841822 0.841822\n -1.35522 -1.35522\n -2.43996 -2.43996\n 5.14732 5.14732\n -7.50974 -7.50974\n 5.73113 5.73113\n -2.72015 -2.72015\n -5.04474 -5.04474\n -13.1 -13.1\n 0.0777815 0.0777815\n 7.85631 7.85631\n -0.323243 -0.323243\n -2.97974 -2.97974\n 0.925187 0.925187\n 5.77219 5.77219\n 4.39868 4.39868\n 2.22326 2.22326\n 1.79052 1.79052\n -3.37507 -3.37507\n -4.08645 -4.08645\n 5.59349 5.59349\n 11.879 11.879\n -0.8099 -0.8099\n 16.6866 16.6866\n 2.85772 2.85772\n 3.73902 3.73902\n -0.406009 -0.406009\n 7.49033 7.49033\n -1.01733 -1.01733\n 4.03678 4.03678\n 4.91574 4.91574\n 14.6191 14.6191\n -1.18215 -1.18215\n -2.79895 -2.79895\n -5.16604 -5.16604\n -2.24596 -2.24596\n 1.83945 1.83945\n 1.72673 1.72673\n -23.2963 -23.2963\n -0.623748 -0.623748\n -2.8419 -2.8419\n 6.56374 6.56374\n 10.3431 10.3431\n 5.28302 5.28302\n 3.12716 3.12716\n 8.41242 8.41242\n 0.416003 0.416003\n -2.43236 -2.43236\n -1.63284 -1.63284\n 5.3806 5.3806\n 9.39975 9.39975\n 4.44496 4.44496\n -3.01441 -3.01441\n -1.33538 -1.33538\n 2.23541 2.23541\n -4.30131 -4.30131\n -1.20324 -1.20324\n 4.79406 4.79406\n 0.692551 0.692551\n -2.20403 -2.20403\n 0.12931 0.12931\n 0.842875 0.842875\n 0.29791 0.29791\n 6.59639 6.59639\n 8.6591 8.6591\n 2.07311 2.07311\n -6.48842 -6.48842\n 2.70007 2.70007\n -0.143695 -0.143695\n 3.99651 3.99651\n 6.86089 6.86089\n -2.54281 -2.54281\n -5.085 -5.085\n 3.61747 3.61747\n 2.09466 2.09466\n 3.35667 3.35667\n 7.38405 7.38405\n 0.816999 0.816999\n -0.564258 -0.564258\n 2.46281 2.46281\n -0.081471 -0.081471\n 12.0933 12.0933\n 9.45364 9.45364\n 0.303564 0.303564\n -2.20687 -2.20687\n 1.90101 1.90101\n -2.65606 -2.65606\n -11.3589 -11.3589\n -1.68249 -1.68249\n -1.25813 -1.25813\n -0.96125 -0.96125\n -2.84666 -2.84666\n 1.18914 1.18914\n 0.211945 0.211945\n -4.8988 -4.8988\n 0.894798 0.894798\n 3.9685 3.9685\n -0.852608 -0.852608\n 3.37537 3.37537\n -0.847579 -0.847579\n -4.37006 -4.37006\n -4.12787 -4.12787\n 4.37155 4.37155\n -7.86631 -7.86631\n -3.59755 -3.59755\n -2.55397 -2.55397\n 4.25921 4.25921\n 2.21721 2.21721\n 5.72299 5.72299\n 8.32362 8.32362\n 14.4057 14.4057\n 1.49376 1.49376\n 3.108 3.108\n -1.34388 -1.34388\n 3.77816 3.77816\n 5.69761 5.69761\n 0.255491 0.255491\n 4.15979 4.15979\n -14.6016 -14.6016\n 3.1475 3.1475\n 2.86732 2.86732\n -2.7875 -2.7875\n -8.78827 -8.78827\n -1.38068 -1.38068\n -2.74156 -2.74156\n -4.82257 -4.82257\n -4.64984 -4.64984\n -0.462036 -0.462036\n 2.36274 2.36274\n 2.73927 2.73927\n -4.01583 -4.01583\n -4.20256 -4.20256\n 7.33455 7.33455\n 7.53557 7.53557\n 3.2532 3.2532\n -0.556551 -0.556551\n 4.39618 4.39618\n 2.92025 2.92025\n -49.4395 -49.4395\n 1.84066 1.84066\n -6.03682 -6.03682\n 9.70956 9.70956\n 12.18 12.18\n -0.134471 -0.134471\n 0.388477 0.388477\n -4.30526 -4.30526\n 3.98614 3.98614\n -3.20351 -3.20351\n 3.81764 3.81764\n 5.34853 5.34853\n 0.382215 0.382215\n -0.473372 -0.473372\n -4.4073 -4.4073\n -10.1129 -10.1129\n -6.82482 -6.82482\n 5.39935 5.39935\n -0.664077 -0.664077\n 7.75577 7.75577\n -5.565 -5.565\n -2.28518 -2.28518\n -3.09472 -3.09472\n 6.0196 6.0196\n -1.32035 -1.32035\n 2.5721 2.5721\n -9.0201 -9.0201\n 6.87621 6.87621\n 7.57662 7.57662\n -2.42131 -2.42131\n -7.11 -7.11\n 1.5457 1.5457\n 1.38686 1.38686\n -1.67077 -1.67077\n 5.34357 5.34357\n -5.22992 -5.22992\n -5.50112 -5.50112\n -0.820436 -0.820436\n -6.85987 -6.85987\n 4.36935 4.36935\n 8.27737 8.27737\n 7.16613 7.16613\n 7.21538 7.21538\n 0.0297893 0.0297893\n -3.30991 -3.30991\n 1.18508 1.18508\n -0.745072 -0.745072\n -1.31153 -1.31153\n -2.57184 -2.57184\n -0.187369 -0.187369\n 6.79233 6.79233\n 8.04294 8.04294\n 3.06986 3.06986\n -5.13761 -5.13761\n 0.539648 0.539648\n 5.02007 5.02007\n 2.67737 2.67737\n -6.69984 -6.69984\n 6.76321 6.76321\n 6.25102 6.25102\n 3.80545 3.80545\n -2.16059 -2.16059\n 2.81803 2.81803\n 0.447194 0.447194\n 1.84756 1.84756\n -6.42528 -6.42528\n -2.23379 -2.23379\n -2.61151 -2.61151\n -2.86143 -2.86143\n -2.94039 -2.94039\n -3.38503 -3.38503\n 0.474985 0.474985\n -9.66389 -9.66389\n 4.96293 4.96293\n -5.6718 -5.6718\n 7.06422 7.06422\n -8.36354 -8.36354\n 0.0182466 0.0182466\n 9.20883 9.20883\n 8.23981 8.23981\n -1.41968 -1.41968\n -1.36057 -1.36057\n -3.99568 -3.99568\n 2.51484 2.51484\n 5.41846 5.41846\n -10.8511 -10.8511\n -8.41267 -8.41267\n 2.04668 2.04668\n -5.61525 -5.61525\n -9.73507 -9.73507\n -0.497102 -0.497102\n 4.29467 4.29467\n -1.61424 -1.61424\n -0.818494 -0.818494\n -7.02135 -7.02135\n 13.4836 13.4836\n -4.10115 -4.10115\n -8.11914 -8.11914\n -2.79895 -2.79895\n -4.39428 -4.39428\n -0.737467 -0.737467\n 1.37013 1.37013\n 9.56244 9.56244\n 2.92491 2.92491\n -7.13393 -7.13393\n -0.179291 -0.179291\n -6.00313 -6.00313\n 7.27104 7.27104\n -1.7103 -1.7103\n -7.84843 -7.84843\n 13.7304 13.7304\n 2.40973 2.40973\n -7.07755 -7.07755\n 1.31745 1.31745\n -9.99271 -9.99271\n -15.4753 -15.4753\n 4.38711 4.38711\n -5.41127 -5.41127\n -1.06491 -1.06491\n 1.09245 1.09245\n -1.33961 -1.33961\n -4.42681 -4.42681\n -4.44164 -4.44164\n -1.80772 -1.80772\n -5.06035 -5.06035\n 0.197369 0.197369\n 7.27798 7.27798\n -6.88382 -6.88382\n 3.21319 3.21319\n 8.04111 8.04111\n -3.94107 -3.94107\n 1.79716 1.79716\n -0.2134 -0.2134\n 1.36955 1.36955\n 13.7009 13.7009\n -7.3497 -7.3497\n 1.80078 1.80078\n 4.25352 4.25352\n -2.80092 -2.80092\n -3.81295 -3.81295\n -4.92036 -4.92036\n 0.856001 0.856001\n -1.26696 -1.26696\n 2.65207 2.65207\n -1.01876 -1.01876\n 1.50837 1.50837\n -11.5335 -11.5335\n 5.80989 5.80989\n 2.45606 2.45606\n 1.64394 1.64394\n 2.73651 2.73651\n -11.1653 -11.1653\n -1.66359 -1.66359\n -0.0317267 -0.0317267\n 0.115458 0.115458\n 4.43585 4.43585\n 1.24902 1.24902\n 7.30894 7.30894\n 16.7814 16.7814\n -0.456154 -0.456154\n -3.94033 -3.94033\n -4.4947 -4.4947\n -2.52048 -2.52048\n 0.0890704 0.0890704\n -4.66338 -4.66338\n 3.88142 3.88142\n 2.35984 2.35984\n 4.84037 4.84037\n 6.95444 6.95444\n 2.74408 2.74408\n -3.23958 -3.23958\n -0.467292 -0.467292\n 6.26367 6.26367\n -1.50588 -1.50588\n 4.13389 4.13389\n -2.53819 -2.53819\n -4.4987 -4.4987\n -10.3487 -10.3487\n -14.8297 -14.8297\n -8.48112 -8.48112\n 3.95155 3.95155\n 1.2289 1.2289\n -4.38025 -4.38025\n -0.61687 -0.61687\n 10.8511 10.8511\n 1.15556 1.15556\n -2.19768 -2.19768\n -7.66931 -7.66931\n 4.72919 4.72919\n -7.6738 -7.6738\n -0.688528 -0.688528\n 4.74928 4.74928\n 4.92126 4.92126\n 0.897546 0.897546\n 3.85735 3.85735\n 0.201364 0.201364\n -5.62425 -5.62425\n -3.83117 -3.83117\n 4.05866 4.05866\n 3.10063 3.10063\n 2.5224 2.5224\n -1.51274 -1.51274\n -0.683338 -0.683338\n -3.23147 -3.23147\n -4.21268 -4.21268\n -2.21401 -2.21401\n 1.57887 1.57887\n 0.848257 0.848257\n -5.83704 -5.83704\n -7.00011 -7.00011\n 3.16884 3.16884\n -4.44161 -4.44161\n -7.62482 -7.62482\n -0.266943 -0.266943\n 0.41761 0.41761\n -7.45144 -7.45144\n -0.211132 -0.211132\n 0.276707 0.276707\n 16.7781 16.7781\n 0.689757 0.689757\n -3.04049 -3.04049\n 2.91684 2.91684\n 1.97161 1.97161\n 3.7721 3.7721\n -1.60698 -1.60698\n -4.18868 -4.18868\n 7.66491 7.66491\n -0.64664 -0.64664\n -0.660623 -0.660623\n 8.68174 8.68174\n 0.282074 0.282074\n -2.85266 -2.85266\n -1.91293 -1.91293\n 7.18736 7.18736\n -10.3875 -10.3875\n -1.91603 -1.91603\n 6.29739 6.29739\n -0.0375388 -0.0375388\n -1.60576 -1.60576\n -3.22148 -3.22148\n -4.24549 -4.24549\n 1.30822 1.30822\n 2.52307 2.52307\n 0.403345 0.403345\n -0.744478 -0.744478\n 2.41241 2.41241\n -4.58098 -4.58098\n -0.791842 -0.791842\n 3.73626 3.73626\n -1.43002 -1.43002\n 4.30716 4.30716\n 3.30255 3.30255\n -4.08011 -4.08011\n -5.07282 -5.07282\n -1.54759 -1.54759\n -2.2305 -2.2305\n 6.8791 6.8791\n 9.7396 9.7396\n -6.50395 -6.50395\n 3.57178 3.57178\n 7.08987 7.08987\n 6.2669 6.2669\n 5.87329 5.87329\n 2.36823 2.36823\n -6.16 -6.16\n 1.96238 1.96238\n 7.31651 7.31651\n -1.5257 -1.5257\n -2.89061 -2.89061\n 0.407546 0.407546\n 5.10645 5.10645\n 11.0716 11.0716\n 4.7443 4.7443\n -8.77353 -8.77353\n -0.631177 -0.631177\n -4.36973 -4.36973\n 1.48666 1.48666\n 7.7678 7.7678\n -2.65407 -2.65407\n 4.56869 4.56869\n -0.541163 -0.541163\n 2.89543 2.89543\n 5.39424 5.39424\n -3.62954 -3.62954\n 3.77547 3.77547\n -5.96886 -5.96886\n -4.38947 -4.38947\n -2.96756 -2.96756\n 2.28222 2.28222\n -1.08489 -1.08489\n 1.74726 1.74726\n -3.46088 -3.46088\n 11.9371 11.9371\n -5.02359 -5.02359\n 2.51632 2.51632\n -0.0297022 -0.0297022\n -2.60011 -2.60011\n 0.254202 0.254202\n 9.7949 9.7949\n 3.64937 3.64937\n 10.0857 10.0857\n -5.36637 -5.36637\n 4.11127 4.11127\n 8.90571 8.90571\n -5.97219 -5.97219\n -7.21379 -7.21379\n -5.01561 -5.01561\n 2.98616 2.98616\n 1.99064 1.99064\n 0.16465 0.16465\n -4.07902 -4.07902\n 4.34018 4.34018\n -2.13528 -2.13528\n 2.39903 2.39903\n 4.00804 4.00804\n -1.85741 -1.85741\n -7.73083 -7.73083\n -4.21139 -4.21139\n 4.65743 4.65743\n 0.963549 0.963549\n 0.29506 0.29506\n 6.05798 6.05798\n 12.4428 12.4428\n -0.398651 -0.398651\n -0.584559 -0.584559\n 2.75445 2.75445\n -0.207975 -0.207975\n 6.11926 6.11926\n -8.66125 -8.66125\n 3.07568 3.07568\n -3.19358 -3.19358\n -2.53024 -2.53024\n 14.1187 14.1187\n -0.412049 -0.412049\n 12.5809 12.5809\n 6.26236 6.26236\n 5.23037 5.23037\n -0.11356 -0.11356\n -6.62321 -6.62321\n -1.29651 -1.29651\n -1.48734 -1.48734\n 13.0753 13.0753\n 4.21767 4.21767\n -2.4425 -2.4425\n -0.0901323 -0.0901323\n 9.79684 9.79684\n 4.74522 4.74522\n -3.34804 -3.34804\n 7.37816 7.37816\n 2.57938 2.57938\n 1.92968 1.92968\n 3.75166 3.75166\n 5.0617 5.0617\n 8.74324 8.74324\n -0.93703 -0.93703\n -1.36031 -1.36031\n -2.5439 -2.5439\n 1.56784 1.56784\n 2.56237 2.56237\n -1.02578 -1.02578\n 6.62085 6.62085\n 7.69745 7.69745\n 6.26864 6.26864\n -4.20046 -4.20046\n -2.30926 -2.30926\n 2.74598 2.74598\n 4.11078 4.11078\n 2.8455 2.8455\n -3.45407 -3.45407\n 2.82327 2.82327\n -1.00356 -1.00356\n 8.85974 8.85974\n 6.35864 6.35864\n -1.59146 -1.59146\n -0.361996 -0.361996\n -1.25198 -1.25198\n 8.2867 8.2867\n 0.981644 0.981644\n 2.68003 2.68003\n 1.10236 1.10236\n -1.63423 -1.63423\n -2.79552 -2.79552\n -6.5718 -6.5718\n -0.257779 -0.257779\n -4.49325 -4.49325\n 5.0455 5.0455\n 14.4508 14.4508\n 3.60407 3.60407\n 3.09003 3.09003\n -8.32962 -8.32962\n -1.41178 -1.41178\n 12.5777 12.5777\n -2.01342 -2.01342\n -1.48205 -1.48205\n 0.967158 0.967158\n -0.532548 -0.532548\n -5.23274 -5.23274\n -1.49702 -1.49702\n 0.739607 0.739607\n 3.49171 3.49171\n -1.0507 -1.0507\n -7.48299 -7.48299\n 7.57395 7.57395\n -3.04813 -3.04813\n 16.322 16.322\n 7.81441 7.81441\n -3.41529 -3.41529\n 2.05401 2.05401\n 1.08232 1.08232\n 12.5735 12.5735\n 0.126572 0.126572\n -6.92158 -6.92158\n -1.4651 -1.4651\n -3.19425 -3.19425\n -1.44093 -1.44093\n -3.82056 -3.82056\n 6.72914 6.72914\n -5.46583 -5.46583\n -1.43396 -1.43396\n 7.42164 7.42164\n 1.00438 1.00438\n -0.41415 -0.41415\n -2.54987 -2.54987\n 6.88491 6.88491\n 3.84807 3.84807\n -5.62245 -5.62245\n 5.24133 5.24133\n 7.99514 7.99514\n -2.51593 -2.51593\n 8.19568 8.19568\n 0.854985 0.854985\n -6.20478 -6.20478\n -2.58235 -2.58235\n -6.51346 -6.51346\n 12.8877 12.8877\n 8.6194 8.6194\n -6.82669 -6.82669\n -4.67379 -4.67379\n 8.13137 8.13137\n 0.733511 0.733511\n 5.66079 5.66079\n -2.94337 -2.94337\n -3.29462 -3.29462\n -6.3809 -6.3809\n -1.85613 -1.85613\n 0.635069 0.635069\n 0.432626 0.432626\n -14.6426 -14.6426\n 8.05825 8.05825\n 6.50637 6.50637\n 1.44014 1.44014\n -4.60602 -4.60602\n -6.49137 -6.49137\n 6.33163 6.33163\n -1.97616 -1.97616\n 0.573379 0.573379\n -2.78039 -2.78039\n -0.140087 -0.140087\n 1.52619 1.52619\n 6.83379 6.83379\n -0.197981 -0.197981\n -3.00849 -3.00849\n -2.09725 -2.09725\n -2.06883 -2.06883\n -0.328198 -0.328198\n -0.212338 -0.212338\n 5.4425 5.4425\n 6.48574 6.48574\n 2.00073 2.00073\n -3.15642 -3.15642\n -0.0673389 -0.0673389\n -4.19911 -4.19911\n 4.5466 4.5466\n 3.73221 3.73221\n -1.01059 -1.01059\n -4.29015 -4.29015\n 4.9909 4.9909\n 3.22397 3.22397\n -1.27984 -1.27984\n 2.83358 2.83358\n 2.25695 2.25695\n 7.2879 7.2879\n -1.47955 -1.47955\n 12.7627 12.7627\n -3.72449 -3.72449\n 3.97719 3.97719\n 14.2197 14.2197\n -1.24031 -1.24031\n -7.41824 -7.41824\n 1.90207 1.90207\n 1.10939 1.10939\n -7.47202 -7.47202\n 3.85738 3.85738\n -4.12085 -4.12085\n 1.12097 1.12097\n -0.545646 -0.545646\n 3.04129 3.04129\n 1.05043 1.05043\n 0.993448 0.993448\n -5.78424 -5.78424\n -1.97199 -1.97199\n -5.74806 -5.74806\n 2.70835 2.70835\n -8.09729 -8.09729\n -6.36035 -6.36035\n -1.24361 -1.24361\n -2.44813 -2.44813\n 7.48353 7.48353\n 2.0202 2.0202\n 3.04366 3.04366\n -3.98778 -3.98778\n 4.80106 4.80106\n 0.926552 0.926552\n 3.35253 3.35253\n -4.10577 -4.10577\n -3.57853 -3.57853\n 4.03372 4.03372\n -2.38792 -2.38792\n 0.12177 0.12177\n -0.761671 -0.761671\n -4.25652 -4.25652\n 7.27933 7.27933\n 0.165182 0.165182\n 1.34367 1.34367\n -7.36923 -7.36923\n 2.38548 2.38548\n 0.117217 0.117217\n 2.02002 2.02002\n -4.60023 -4.60023\n 2.78 2.78\n -1.34604 -1.34604\n 4.7234 4.7234\n 7.37673 7.37673\n 2.07986 2.07986\n -5.72573 -5.72573\n -6.66143 -6.66143\n 2.43072 2.43072\n 1.34782 1.34782\n -0.114238 -0.114238\n 2.32103 2.32103\n 1.84042 1.84042\n 1.07005 1.07005\n 3.88182 3.88182\n -0.752264 -0.752264\n -2.43517 -2.43517\n -5.29216 -5.29216\n -0.13527 -0.13527\n 1.40188 1.40188\n -5.87815 -5.87815\n -1.90167 -1.90167\n 2.88562 2.88562\n -2.29028 -2.29028\n 2.35477 2.35477\n -3.50731 -3.50731\n 6.0621 6.0621\n 3.2011 3.2011\n 2.19115 2.19115\n -3.03557 -3.03557\n -8.49394 -8.49394\n 0.936501 0.936501\n 7.19188 7.19188\n 4.50162 4.50162\n 0.341394 0.341394\n 2.54484 2.54484\n 1.67305 1.67305\n 3.05008 3.05008\n -2.0266 -2.0266\n 7.28431 7.28431\n -7.70924 -7.70924\n 2.60851 2.60851\n 6.8054 6.8054\n 1.8878 1.8878\n 1.87624 1.87624\n -5.13611 -5.13611\n -3.23698 -3.23698\n 4.03201 4.03201\n -5.27165 -5.27165\n -4.95817 -4.95817\n -0.200461 -0.200461\n 4.27259 4.27259\n 0.449661 0.449661\n 7.49752 7.49752\n -5.47923 -5.47923\n -2.40934 -2.40934\n 25.0066 25.0066\n -3.14511 -3.14511\n -1.62587 -1.62587\n -1.67652 -1.67652\n -2.17888 -2.17888\n 2.37296 2.37296\n -4.41408 -4.41408\n 0.65204 0.65204\n 10.849 10.849\n -2.3021 -2.3021\n 2.20417 2.20417\n 10.0579 10.0579\n -4.03489 -4.03489\n 7.60982 7.60982\n -5.74951 -5.74951\n -2.97582 -2.97582\n -8.61382 -8.61382\n -1.90903 -1.90903\n -3.64556 -3.64556\n -16.2304 -16.2304\n -15.9793 -15.9793\n -4.59448 -4.59448\n -2.67688 -2.67688\n -1.67148 -1.67148\n 5.57026 5.57026\n 0.846445 0.846445\n -7.54149 -7.54149\n -3.61401 -3.61401\n 4.03723 4.03723\n 0.711821 0.711821\n 8.99009 8.99009\n -6.15866 -6.15866\n -1.36865 -1.36865\n -4.31058 -4.31058\n 6.31659 6.31659\n -6.23773 -6.23773\n 0.857388 0.857388\n 3.6152 3.6152\n -1.28774 -1.28774\n -4.92094 -4.92094\n 3.08527 3.08527\n -5.74582 -5.74582\n -4.20897 -4.20897\n -5.19406 -5.19406\n -4.06851 -4.06851\n 5.73867 5.73867\n 3.32767 3.32767\n -11.2588 -11.2588\n -7.94126 -7.94126\n 5.38746 5.38746\n -0.0253579 -0.0253579\n -1.7856 -1.7856\n -1.31209 -1.31209\n 6.85519 6.85519\n 2.71496 2.71496\n -2.58838 -2.58838\n -6.86996 -6.86996\n 1.01204 1.01204\n 3.43433 3.43433\n -0.249192 -0.249192\n 7.96322 7.96322\n 14.3414 14.3414\n 2.44774 2.44774\n 4.73731 4.73731\n -9.14288 -9.14288\n 2.70325 2.70325\n 6.48202 6.48202\n -2.58391 -2.58391\n -4.52079 -4.52079\n -0.64105 -0.64105\n -3.75531 -3.75531\n -3.93321 -3.93321\n -2.5879 -2.5879\n 2.34697 2.34697\n -3.89721 -3.89721\n -1.60712 -1.60712\n -7.49452 -7.49452\n -0.518596 -0.518596\n 0.996693 0.996693\n 2.83468 2.83468\n -6.19363 -6.19363\n -7.25683 -7.25683\n 0.391546 0.391546\n -7.52756 -7.52756\n -0.810817 -0.810817\n -2.64942 -2.64942\n -2.95081 -2.95081\n -6.34989 -6.34989\n 3.9961 3.9961\n 1.36755 1.36755\n -0.335808 -0.335808\n -11.7919 -11.7919\n 1.16904 1.16904\n 6.26031 6.26031\n -4.68064 -4.68064\n 5.55008 5.55008\n 3.65873 3.65873\n -3.95177 -3.95177\n 7.62708 7.62708\n -2.4932 -2.4932\n -0.713266 -0.713266\n 6.76214 6.76214\n -0.802523 -0.802523\n -0.327543 -0.327543\n -6.9053 -6.9053\n -2.69604 -2.69604\n 9.729 9.729\n -7.61691 -7.61691\n -0.658653 -0.658653\n 1.62531 1.62531\n 0.532107 0.532107\n 1.71729 1.71729\n -10.1795 -10.1795\n 5.54208 5.54208\n 4.02502 4.02502\n -1.47596 -1.47596\n 11.818 11.818\n 4.40414 4.40414\n 5.64827 5.64827\n 5.89386 5.89386\n -6.19187 -6.19187\n 4.77889 4.77889\n -0.261731 -0.261731\n -0.570525 -0.570525\n 3.80941 3.80941\n -3.95414 -3.95414\n 0.642971 0.642971\n -7.23493 -7.23493\n 0.744423 0.744423\n 11.5682 11.5682\n -3.17145 -3.17145\n 9.02877 9.02877\n 10.5452 10.5452\n -7.05642 -7.05642\n -6.01952 -6.01952\n -5.61355 -5.61355\n 1.28759 1.28759\n 3.44186 3.44186\n -2.52363 -2.52363\n 8.95712 8.95712\n -1.33999 -1.33999\n -3.25858 -3.25858\n 2.33509 2.33509\n 2.16314 2.16314\n 14.4002 14.4002\n -5.22345 -5.22345\n -5.6232 -5.6232\n -4.20801 -4.20801\n 0.677359 0.677359\n 1.92688 1.92688\n 2.4265 2.4265\n -3.47901 -3.47901\n -3.35004 -3.35004\n -5.32445 -5.32445\n 0.817822 0.817822\n 5.9241 5.9241\n 2.13342 2.13342\n 9.30726 9.30726\n -6.00328 -6.00328\n 5.10125 5.10125\n 16.6941 16.6941\n -1.41774 -1.41774\n 0.843709 0.843709\n 3.71326 3.71326\n -12.7315 -12.7315\n -1.58947 -1.58947\n 2.7713 2.7713\n -5.89993 -5.89993\n -10.1427 -10.1427\n -1.60823 -1.60823\n -4.98621 -4.98621\n -10.6258 -10.6258\n 0.255858 0.255858\n 5.87781 5.87781\n 0.549239 0.549239\n -0.361649 -0.361649\n 2.89543 2.89543\n -1.56252 -1.56252\n -7.04269 -7.04269\n 0.360599 0.360599\n -0.80318 -0.80318\n -8.15537 -8.15537\n 7.86106 7.86106\n 4.25906 4.25906\n 1.78474 1.78474\n 4.15764 4.15764\n -1.8884 -1.8884\n -7.16959 -7.16959\n 2.84539 2.84539\n -3.33161 -3.33161\n 4.89863 4.89863\n -3.36503 -3.36503\n -4.68013 -4.68013\n 5.18058 5.18058\n -9.69276 -9.69276\n -1.56116 -1.56116\n -3.58275 -3.58275\n -2.73766 -2.73766\n 6.64492 6.64492\n -3.78966 -3.78966\n 2.63467 2.63467\n -12.4868 -12.4868\n -3.4241 -3.4241\n 3.2898 3.2898\n 2.20265 2.20265\n -1.36672 -1.36672\n 2.71448 2.71448\n 5.87839 5.87839\n 0.160837 0.160837\n -2.64458 -2.64458\n -3.8078 -3.8078\n 5.08743 5.08743\n -14.014 -14.014\n 4.44746 4.44746\n 6.61584 6.61584\n -0.916513 -0.916513\n -8.08277 -8.08277\n -8.088 -8.088\n -5.14152 -5.14152\n -4.30739 -4.30739\n -8.76727 -8.76727\n -4.53313 -4.53313\n 11.0356 11.0356\n -2.37348 -2.37348\n -8.71711 -8.71711\n -2.22971 -2.22971\n 8.19346 8.19346\n -0.330962 -0.330962\n 1.10067 1.10067\n 1.01878 1.01878\n -10.2666 -10.2666\n 8.15909 8.15909\n 9.09316 9.09316\n -0.862864 -0.862864\n -7.54443 -7.54443\n -3.44703 -3.44703\n 5.21819 5.21819\n -2.06834 -2.06834\n 9.55442 9.55442\n -1.89649 -1.89649\n -5.57892 -5.57892\n 4.22421 4.22421\n -4.06375 -4.06375\n 3.81452 3.81452\n 3.09071 3.09071\n -7.34297 -7.34297\n -1.67899 -1.67899\n 0.58489 0.58489\n -5.33824 -5.33824\n 2.82705 2.82705\n -3.70864 -3.70864\n 4.21641 4.21641\n 3.82508 3.82508\n -4.04356 -4.04356\n 20.0249 20.0249\n -13.1531 -13.1531\n 2.98603 2.98603\n 5.54713 5.54713\n -1.39722 -1.39722\n 2.13016 2.13016\n -2.40215 -2.40215\n 0.168123 0.168123\n 2.77021 2.77021\n -2.32327 -2.32327\n -1.06731 -1.06731\n 2.53877 2.53877\n -1.94325 -1.94325\n 1.47106 1.47106\n 0.294436 0.294436\n -0.547055 -0.547055\n 0.116016 0.116016\n 1.56148 1.56148\n 3.21789 3.21789\n -2.89007 -2.89007\n -4.33765 -4.33765\n 0.566163 0.566163\n 0.402729 0.402729\n -7.80674 -7.80674\n 4.72058 4.72058\n 3.97584 3.97584\n 1.91646 1.91646\n 2.09298 2.09298\n 1.88552 1.88552\n -2.37581 -2.37581\n -18.2615 -18.2615\n 2.68651 2.68651\n 5.5 5.5\n 0.355051 0.355051\n 5.6052 5.6052\n 7.74854 7.74854\n -0.512378 -0.512378\n 1.60299 1.60299\n -5.49563 -5.49563\n -1.96455 -1.96455\n -16.3228 -16.3228\n -6.87737 -6.87737\n -4.60755 -4.60755\n -1.32116 -1.32116\n 2.87263 2.87263\n -2.09541 -2.09541\n 3.43595 3.43595\n 3.63528 3.63528\n 3.52056 3.52056\n -3.59484 -3.59484\n 1.03764 1.03764\n -7.14947 -7.14947\n -5.80634 -5.80634\n 4.71397 4.71397\n 0.720588 0.720588\n -2.24074 -2.24074\n 5.82418 5.82418\n -3.22013 -3.22013\n 3.68858 3.68858\n -1.43166 -1.43166\n 4.47978 4.47978\n -4.83356 -4.83356\n -3.96257 -3.96257\n -5.95512 -5.95512\n 0.496691 0.496691\n -7.58825 -7.58825\n -6.47331 -6.47331\n -1.14446 -1.14446\n 3.91615 3.91615\n -0.588841 -0.588841\n 6.56683 6.56683\n 3.97252 3.97252\n -4.3126 -4.3126\n -8.20913 -8.20913\n 0.310182 0.310182\n -7.3006 -7.3006\n 7.92805 7.92805\n 2.1756 2.1756\n 1.06404 1.06404\n 1.14471 1.14471\n -1.50242 -1.50242\n 0.00723557 0.00723557\n 5.76841 5.76841\n -1.96707 -1.96707\n 8.87243 8.87243\n -3.23281 -3.23281\n 12.3087 12.3087\n 3.3245 3.3245\n 3.00334 3.00334\n -5.74048 -5.74048\n 7.43939 7.43939\n -0.906001 -0.906001\n 2.24067 2.24067\n -6.23989 -6.23989\n 2.81483 2.81483\n -1.62648 -1.62648\n -7.26368 -7.26368\n 1.69171 1.69171\n -11.2631 -11.2631\n -2.32992 -2.32992\n -6.07361 -6.07361\n -7.56822 -7.56822\n -7.56737 -7.56737\n 5.97037 5.97037\n 6.74398 6.74398\n -2.24599 -2.24599\n 2.95213 2.95213\n -12.7864 -12.7864\n 0.680035 0.680035\n -1.39988 -1.39988\n -4.74028 -4.74028\n 3.01887 3.01887\n 1.89636 1.89636\n 4.46014 4.46014\n -4.38308 -4.38308\n 11.7633 11.7633\n -3.54671 -3.54671\n -3.47584 -3.47584\n 3.80037 3.80037\n 7.77849 7.77849\n -7.00006 -7.00006\n -4.87665 -4.87665\n -4.54736 -4.54736\n -7.81752 -7.81752\n -0.0654465 -0.0654465\n -3.70587 -3.70587\n -2.24231 -2.24231\n 5.58005 5.58005\n -3.09415 -3.09415\n -5.55063 -5.55063\n -4.19666 -4.19666\n -6.83328 -6.83328\n -6.9216 -6.9216\n -3.72782 -3.72782\n -2.18574 -2.18574\n 1.28076 1.28076\n -3.40691 -3.40691\n 0.486964 0.486964\n -2.11025 -2.11025\n -1.42349 -1.42349\n 6.06854 6.06854\n -1.37534 -1.37534\n 9.47832 9.47832\n -0.567045 -0.567045\n -6.98328 -6.98328\n 6.73139 6.73139\n -1.56812 -1.56812\n 0.141683 0.141683\n 1.78697 1.78697\n -2.03874 -2.03874\n 1.28356 1.28356\n 6.9912 6.9912\n -3.8858 -3.8858\n -1.38808 -1.38808\n -2.16632 -2.16632\n 3.57955 3.57955\n 2.73506 2.73506\n -3.03108 -3.03108\n -3.44677 -3.44677\n 1.37111 1.37111\n -10.0008 -10.0008\n -3.61651 -3.61651\n 1.97313 1.97313\n 2.11298 2.11298\n 0.174957 0.174957\n -0.131546 -0.131546\n 7.58484 7.58484\n 4.27907 4.27907\n 0.855439 0.855439\n 4.44153 4.44153\n -1.04577 -1.04577\n -7.49625 -7.49625\n 2.1572 2.1572\n 13.0815 13.0815\n 4.57025 4.57025\n 0.704658 0.704658\n 3.25079 3.25079\n -0.682139 -0.682139\n -4.17209 -4.17209\n -1.38547 -1.38547\n 5.52688 5.52688\n -4.90717 -4.90717\n 2.56402 2.56402\n -1.37164 -1.37164\n -6.05044 -6.05044\n 8.3158 8.3158\n -0.640461 -0.640461\n -2.40145 -2.40145\n -1.02959 -1.02959\n -6.75028 -6.75028\n 4.20206 4.20206\n 0.615412 0.615412\n -0.389435 -0.389435\n -5.07439 -5.07439\n -5.34136 -5.34136\n -1.88522 -1.88522\n -4.82628 -4.82628\n 0.54435 0.54435\n -3.28948 -3.28948\n 5.0051 5.0051\n -8.5501 -8.5501\n 7.31448 7.31448\n 0.145651 0.145651\n 3.28586 3.28586\n -1.8624 -1.8624\n -8.9235 -8.9235\n 3.15894 3.15894\n -9.9459 -9.9459\n 0.517233 0.517233\n -4.59899 -4.59899\n 0.641116 0.641116\n 10.3809 10.3809\n 2.39935 2.39935\n -0.378496 -0.378496\n 0.680329 0.680329\n 2.35584 2.35584\n -2.24714 -2.24714\n -4.8742 -4.8742\n -3.96429 -3.96429\n 1.29263 1.29263\n 0.618875 0.618875\n -0.611961 -0.611961\n 1.06612 1.06612\n -3.39289 -3.39289\n -0.226022 -0.226022\n 4.24418 4.24418\n 0.884239 0.884239\n 8.25747 8.25747\n -3.23019 -3.23019\n -9.99374 -9.99374\n 8.54414 8.54414\n -6.06374 -6.06374\n -4.92601 -4.92601\n 7.22101 7.22101\n 11.5756 11.5756\n 13.436 13.436\n 4.13522 4.13522\n 9.67412 9.67412\n -3.13805 -3.13805\n 7.50856 7.50856\n -7.98069 -7.98069\n 4.92059 4.92059\n -6.72969 -6.72969\n -4.48762 -4.48762\n -3.60328 -3.60328\n -1.75053 -1.75053\n 1.5638 1.5638\n 4.74213 4.74213\n 5.16046 5.16046\n -1.9857 -1.9857\n -6.34885 -6.34885\n -3.58963 -3.58963\n 4.96795 4.96795\n 1.44405 1.44405\n -2.74682 -2.74682\n -0.545296 -0.545296\n -10.7507 -10.7507\n -0.117477 -0.117477\n -0.436907 -0.436907\n -1.11656 -1.11656\n 1.64789 1.64789\n -4.08799 -4.08799\n -1.04262 -1.04262\n 6.06007 6.06007\n -6.68208 -6.68208\n 6.81976 6.81976\n -6.89836 -6.89836\n -0.555115 -0.555115\n -2.85307 -2.85307\n -7.76567 -7.76567\n -5.65104 -5.65104\n 8.93521 8.93521\n -5.0663 -5.0663\n 2.52214 2.52214\n 0.382824 0.382824\n -0.398468 -0.398468\n 5.05183 5.05183\n 4.134 4.134\n 1.42909 1.42909\n 2.99357 2.99357\n 10.7821 10.7821\n -4.54764 -4.54764\n -0.0440308 -0.0440308\n 0.647161 0.647161\n 3.27569 3.27569\n -32.9478 -32.9478\n 6.92399 6.92399\n -3.05953 -3.05953\n -2.29742 -2.29742\n -0.41863 -0.41863\n 2.99125 2.99125\n 3.40805 3.40805\n -1.36651 -1.36651\n -3.25561 -3.25561\n 5.11504 5.11504\n -0.532291 -0.532291\n 9.93341 9.93341\n -2.2806 -2.2806\n 10.9617 10.9617\n -2.53642 -2.53642\n 0.995763 0.995763\n -1.28898 -1.28898\n -2.99921 -2.99921\n -2.46773 -2.46773\n -11.0849 -11.0849\n -11.64 -11.64\n -3.73617 -3.73617\n 2.74223 2.74223\n -0.976817 -0.976817\n -0.384814 -0.384814\n -3.38815 -3.38815\n 2.27591 2.27591\n -5.25732 -5.25732\n -1.65764 -1.65764\n -5.8501 -5.8501\n -4.85863 -4.85863\n 2.78987 2.78987\n 5.3324 5.3324\n -9.16758 -9.16758\n 7.90047 7.90047\n 5.68696 5.68696\n 7.2668 7.2668\n -0.857072 -0.857072\n 0.0834347 0.0834347\n 1.11833 1.11833\n 0.88212 0.88212\n -4.40785 -4.40785\n 5.25846 5.25846\n 7.46283 7.46283\n 6.26981 6.26981\n -10.8935 -10.8935\n -0.226332 -0.226332\n -1.64568 -1.64568\n -0.389003 -0.389003\n -0.854872 -0.854872\n -3.38063 -3.38063\n -4.74874 -4.74874\n -1.81717 -1.81717\n -6.03338 -6.03338\n 9.41153 9.41153\n -2.75636 -2.75636\n -4.03638 -4.03638\n -2.82527 -2.82527\n 0.641039 0.641039\n -3.08939 -3.08939\n -1.04523 -1.04523\n -4.17379 -4.17379\n 0.453503 0.453503\n 5.64541 5.64541\n 2.72225 2.72225\n -1.67354 -1.67354\n -6.68729 -6.68729\n -1.20785 -1.20785\n 3.51562 3.51562\n 2.38257 2.38257\n 2.75735 2.75735\n -4.62925 -4.62925\n 7.98247 7.98247\n 6.254 6.254\n 3.85448 3.85448\n -4.40298 -4.40298\n -8.28751 -8.28751\n -7.28055 -7.28055\n 7.31675 7.31675\n 3.53957 3.53957\n 2.94378 2.94378\n 1.41268 1.41268\n 5.2878 5.2878\n -0.807317 -0.807317\n -13.141 -13.141\n 5.71505 5.71505\n -3.86739 -3.86739\n 0.922435 0.922435\n -4.52167 -4.52167\n 0.82741 0.82741\n 4.1254 4.1254\n -3.64229 -3.64229\n -4.34879 -4.34879\n -5.69361 -5.69361\n 10.0503 10.0503\n -6.20878 -6.20878\n -5.70531 -5.70531\n -0.265037 -0.265037\n 4.91217 4.91217\n -9.85839 -9.85839\n 9.14639 9.14639\n 0.78426 0.78426\n -6.03581 -6.03581\n -1.225 -1.225\n -1.82514 -1.82514\n -4.38257 -4.38257\n -4.14898 -4.14898\n 1.30056 1.30056\n -4.04361 -4.04361\n -10.7862 -10.7862\n -1.71033 -1.71033\n -5.3235 -5.3235\n -5.05158 -5.05158\n 2.03088 2.03088\n -4.639 -4.639\n -8.90379 -8.90379\n -1.46286 -1.46286\n 4.78737 4.78737\n 2.84292 2.84292\n -4.60125 -4.60125\n -0.454598 -0.454598\n -3.54703 -3.54703\n -3.15574 -3.15574\n -5.66794 -5.66794\n -0.499733 -0.499733\n 4.80394 4.80394\n 7.0018 7.0018\n -12.2494 -12.2494\n -0.705371 -0.705371\n 0.0740021 0.0740021\n -2.66987 -2.66987\n 2.48263 2.48263\n -9.06332 -9.06332\n -1.01261 -1.01261\n 3.84118 3.84118\n 4.21216 4.21216\n -1.18673 -1.18673\n -11.0005 -11.0005\n -9.71638 -9.71638\n 1.76212 1.76212\n -2.83766 -2.83766\n -9.13768 -9.13768\n -1.05015 -1.05015\n 2.53008 2.53008\n 0.379504 0.379504\n 5.28803 5.28803\n -6.17221 -6.17221\n 5.75619 5.75619\n 2.3737 2.3737\n -9.0974 -9.0974\n -7.85433 -7.85433\n -10.9094 -10.9094\n 1.20756 1.20756\n 2.61486 2.61486\n 1.23359 1.23359\n 43.6151 43.6151\n -1.72859 -1.72859\n -0.965831 -0.965831\n -0.482239 -0.482239\n -1.82159 -1.82159\n 1.661 1.661\n 1.93636 1.93636\n -11.9999 -11.9999\n 0.104367 0.104367\n -1.70555 -1.70555\n -9.81074 -9.81074\n 12.7941 12.7941\n -3.36221 -3.36221\n -6.06523 -6.06523\n 0.47411 0.47411\n -6.64475 -6.64475\n -0.763006 -0.763006\n -3.9763 -3.9763\n -2.86732 -2.86732\n -20.6937 -20.6937\n 1.84418 1.84418\n 5.65243 5.65243\n 10.7255 10.7255\n -1.21293 -1.21293\n 3.15057 3.15057\n 8.96094 8.96094\n -0.205015 -0.205015\n 8.44579 8.44579\n 2.01362 2.01362\n 2.36648 2.36648\n 11.6752 11.6752\n 2.19072 2.19072\n -13.9182 -13.9182\n 3.3257 3.3257\n -6.60627 -6.60627\n 1.62083 1.62083\n -2.00847 -2.00847\n 11.6978 11.6978\n 5.93254 5.93254\n 4.93134 4.93134\n -2.50847 -2.50847\n -5.92846 -5.92846\n 1.16717 1.16717\n 6.9673 6.9673\n -1.21182 -1.21182\n 7.25413 7.25413\n -4.24031 -4.24031\n -3.12368 -3.12368\n 1.73734 1.73734\n -2.6551 -2.6551\n 5.01063 5.01063\n 10.9923 10.9923\n 3.08502 3.08502\n -1.67866 -1.67866\n 10.7003 10.7003\n -0.982895 -0.982895\n 1.97681 1.97681\n -1.29045 -1.29045\n 1.64227 1.64227\n 3.21157 3.21157\n -4.63376 -4.63376\n 4.47725 4.47725\n 7.77208 7.77208\n 0.332548 0.332548\n 2.82084 2.82084\n 0.958649 0.958649\n 1.21302 1.21302\n -3.16936 -3.16936\n 0.0672417 0.0672417\n 0.563038 0.563038\n -1.87542 -1.87542\n -3.01753 -3.01753\n 2.73107 2.73107\n -3.68276 -3.68276\n 4.64376 4.64376\n -12.4341 -12.4341\n 4.43429 4.43429\n 5.72878 5.72878\n 2.39332 2.39332\n 1.91106 1.91106\n 2.50458 2.50458\n 0.942479 0.942479\n -0.489758 -0.489758\n 0.311101 0.311101\n -2.74953 -2.74953\n 4.95959 4.95959\n 1.26862 1.26862\n 10.3622 10.3622\n 3.61213 3.61213\n -2.19285 -2.19285\n 1.28587 1.28587\n -1.85274 -1.85274\n -1.62541 -1.62541\n 2.00382 2.00382\n -5.8959 -5.8959\n -0.918042 -0.918042\n 6.43711 6.43711\n 0.419441 0.419441\n -2.61133 -2.61133\n -0.0277654 -0.0277654\n 2.77443 2.77443\n 3.83764 3.83764\n -1.44486 -1.44486\n -0.611288 -0.611288\n -4.30436 -4.30436\n 5.29466 5.29466\n 1.56058 1.56058\n 1.88962 1.88962\n 0.761408 0.761408\n 1.76505 1.76505\n 1.18453 1.18453\n 1.71559 1.71559\n -3.14851 -3.14851\n 2.73145 2.73145\n -1.23904 -1.23904\n 0.00672958 0.00672958\n 3.40979 3.40979\n -1.77498 -1.77498\n -7.12266 -7.12266\n -9.24697 -9.24697\n -4.12038 -4.12038\n -2.77817 -2.77817\n 8.23453 8.23453\n -1.29818 -1.29818\n -7.02203 -7.02203\n -5.8994 -5.8994\n 8.20499 8.20499\n 0.356509 0.356509\n -0.515947 -0.515947\n -6.23904 -6.23904\n 5.59801 5.59801\n -4.44281 -4.44281\n -2.28591 -2.28591\n -3.31819 -3.31819\n 2.39253 2.39253\n 3.18355 3.18355\n -2.73303 -2.73303\n -0.0346074 -0.0346074\n -10.2692 -10.2692\n 6.74308 6.74308\n 5.72055 5.72055\n -4.49033 -4.49033\n 1.99176 1.99176\n 6.10782 6.10782\n 2.65759 2.65759\n 1.97884 1.97884\n 0.927606 0.927606\n 1.25006 1.25006\n 9.3695 9.3695\n -2.75726 -2.75726\n -0.580415 -0.580415\n 2.92463 2.92463\n -4.49535 -4.49535\n -1.61397 -1.61397\n 3.26733 3.26733\n -3.61505 -3.61505\n -2.46453 -2.46453\n 2.42436 2.42436\n 5.68683 5.68683\n 6.07494 6.07494\n 4.35205 4.35205\n -5.29467 -5.29467\n -3.90039 -3.90039\n -1.70776 -1.70776\n -6.3172 -6.3172\n 4.03858 4.03858\n -2.58786 -2.58786\n -1.1514 -1.1514\n -0.632569 -0.632569\n -0.343314 -0.343314\n -12.2115 -12.2115\n 0.405742 0.405742\n -6.46017 -6.46017\n -2.30808 -2.30808\n 1.1336 1.1336\n 1.47556 1.47556\n 1.98494 1.98494\n 2.24865 2.24865\n -1.65786 -1.65786\n -4.62769 -4.62769\n 4.43717 4.43717\n 8.75249 8.75249\n 4.29167 4.29167\n -3.96876 -3.96876\n -3.52244 -3.52244\n 0.161164 0.161164\n -4.13202 -4.13202\n 1.42269 1.42269\n -3.05155 -3.05155\n 1.81371 1.81371\n -1.03765 -1.03765\n 0.696656 0.696656\n 2.95359 2.95359\n -4.74837 -4.74837\n -9.03481 -9.03481\n 4.8852 4.8852\n 9.47173 9.47173\n 11.3037 11.3037\n -3.88084 -3.88084\n -5.99356 -5.99356\n 7.81639 7.81639\n -6.51949 -6.51949\n 7.801 7.801\n -0.795429 -0.795429\n -0.801046 -0.801046\n 2.70658 2.70658\n 5.51012 5.51012\n 1.8181 1.8181\n -0.452854 -0.452854\n -10.1558 -10.1558\n 1.95877 1.95877\n -3.88197 -3.88197\n 1.72033 1.72033\n -1.8939 -1.8939\n -1.64082 -1.64082\n -0.409815 -0.409815\n 9.98658 9.98658\n -0.115277 -0.115277\n 1.49827 1.49827\n 1.6696 1.6696\n 2.29297 2.29297\n -2.14941 -2.14941\n 2.43318 2.43318\n 3.59845 3.59845\n -4.58877 -4.58877\n -9.25371 -9.25371\n 2.03609 2.03609\n 5.5921 5.5921\n -0.532859 -0.532859\n 4.34937 4.34937\n 1.57036 1.57036\n 2.30747 2.30747\n 7.5055 7.5055\n 3.41771 3.41771\n 0.589402 0.589402\n 1.55834 1.55834\n 5.12407 5.12407\n -1.41727 -1.41727\n 1.03223 1.03223\n -2.06257 -2.06257\n 3.11532 3.11532\n 1.90042 1.90042\n 8.66814 8.66814\n 5.36716 5.36716\n 2.38085 2.38085\n 5.72834 5.72834\n -6.5998 -6.5998\n 0.852569 0.852569\n -7.5648 -7.5648\n 2.98063 2.98063\n 7.81573 7.81573\n 1.82276 1.82276\n -1.81083 -1.81083\n 5.48043 5.48043\n -1.85315 -1.85315\n -1.62277 -1.62277\n -10.4951 -10.4951\n 5.34799 5.34799\n -1.77515 -1.77515\n 5.88005 5.88005\n 0.0799242 0.0799242\n 1.23264 1.23264\n -11.835 -11.835\n 3.56828 3.56828\n 7.53741 7.53741\n -5.24051 -5.24051\n -0.206917 -0.206917\n 4.36865 4.36865\n -4.10348 -4.10348\n 0.857712 0.857712\n -5.09677 -5.09677\n 7.37208 7.37208\n -3.14614 -3.14614\n 12.061 12.061\n 4.80096 4.80096\n 2.82421 2.82421\n -4.97446 -4.97446\n -11.0289 -11.0289\n -8.33282 -8.33282\n 0.69922 0.69922\n 5.08771 5.08771\n 2.65174 2.65174\n -3.30182 -3.30182\n 5.21741 5.21741\n 8.85373 8.85373\n 8.36416 8.36416\n 2.54295 2.54295\n -1.61657 -1.61657\n 1.12017 1.12017\n -7.33205 -7.33205\n 3.82582 3.82582\n -0.858026 -0.858026\n 1.40304 1.40304\n 1.35079 1.35079\n 4.19532 4.19532\n -1.77923 -1.77923\n -10.5119 -10.5119\n 10.8061 10.8061\n -3.49603 -3.49603\n 3.12404 3.12404\n -3.93328 -3.93328\n -6.73356 -6.73356\n 1.80532 1.80532\n -0.368024 -0.368024\n -3.47875 -3.47875\n -4.22893 -4.22893\n 2.52519 2.52519\n -3.54943 -3.54943\n -2.39869 -2.39869\n 4.22126 4.22126\n -0.253856 -0.253856\n 7.51866 7.51866\n -4.54093 -4.54093\n 3.44497 3.44497\n 4.77417 4.77417\n 4.49646 4.49646\n -5.78678 -5.78678\n 0.745013 0.745013\n 1.69763 1.69763\n -2.64759 -2.64759\n 1.66108 1.66108\n -4.68276 -4.68276\n 5.31823 5.31823\n 3.52288 3.52288\n 4.9695 4.9695\n 12.2016 12.2016\n 2.46849 2.46849\n -7.60038 -7.60038\n 8.21628 8.21628\n 5.99856 5.99856\n -6.80947 -6.80947\n 7.22522 7.22522\n -2.00065 -2.00065\n -8.24049 -8.24049\n -0.0804049 -0.0804049\n -2.06638 -2.06638\n -2.82884 -2.82884\n -4.25891 -4.25891\n -5.20258 -5.20258\n -3.19396 -3.19396\n -5.14527 -5.14527\n -4.28244 -4.28244\n 4.70805 4.70805\n -3.08065 -3.08065\n -4.86906 -4.86906\n -29.0266 -29.0266\n -1.22941 -1.22941\n -1.30928 -1.30928\n -6.35234 -6.35234\n 1.87904 1.87904\n 8.37797 8.37797\n -5.8821 -5.8821\n 3.10138 3.10138\n -3.27553 -3.27553\n -0.208451 -0.208451\n -2.28999 -2.28999\n 12.2896 12.2896\n -1.27394 -1.27394\n -3.41924 -3.41924\n -0.289592 -0.289592\n 1.79867 1.79867\n 1.98504 1.98504\n 1.55159 1.55159\n 1.10858 1.10858\n 0.352842 0.352842\n -0.309044 -0.309044\n -0.165336 -0.165336\n 1.15822 1.15822\n -1.39342 -1.39342\n -0.162562 -0.162562\n -8.06055 -8.06055\n -5.02776 -5.02776\n -8.66927 -8.66927\n 1.14576 1.14576\n -1.52122 -1.52122\n -1.29436 -1.29436\n 3.26421 3.26421\n 7.55561 7.55561\n 7.7265 7.7265\n -0.48821 -0.48821\n 12.439 12.439\n 7.0264 7.0264\n -11.9855 -11.9855\n -3.74151 -3.74151\n -0.200302 -0.200302\n 5.39515 5.39515\n -4.3468 -4.3468\n 9.25599 9.25599\n 3.37455 3.37455\n -6.15424 -6.15424\n -6.6271 -6.6271\n 0.000272481 0.000272481\n -5.48117 -5.48117\n -0.493191 -0.493191\n -3.46473 -3.46473\n 2.33812 2.33812\n 0.885965 0.885965\n 4.74926 4.74926\n 1.51959 1.51959\n 2.50956 2.50956\n -0.728024 -0.728024\n 1.0381 1.0381\n 5.48121 5.48121\n -1.68033 -1.68033\n -5.05915 -5.05915\n -0.646233 -0.646233\n 0.614062 0.614062\n 4.54219 4.54219\n -1.63006 -1.63006\n -3.10589 -3.10589\n -3.12801 -3.12801\n -5.98177 -5.98177\n -3.59188 -3.59188\n 1.7066 1.7066\n -7.43935 -7.43935\n 10.6141 10.6141\n 12.6478 12.6478\n -1.7222 -1.7222\n -2.1519 -2.1519\n -7.16573 -7.16573\n 0.887314 0.887314\n -8.59735 -8.59735\n -1.3609 -1.3609\n 4.47651 4.47651\n 0.900892 0.900892\n 7.81857 7.81857\n 6.19857 6.19857\n 2.12844 2.12844\n 3.08551 3.08551\n 4.15866 4.15866\n 2.09657 2.09657\n -2.27786 -2.27786\n 1.33571 1.33571\n 4.46899 4.46899\n 4.46674 4.46674\n 3.20736 3.20736\n 5.68287 5.68287\n 10.1058 10.1058\n 5.1894 5.1894\n 3.5452 3.5452\n 10.06 10.06\n 7.02935 7.02935\n -1.06066 -1.06066\n 10.32 10.32\n -0.860463 -0.860463\n 5.95992 5.95992\n -6.30137 -6.30137\n -5.01947 -5.01947\n 5.75187 5.75187\n -1.10079 -1.10079\n -1.91783 -1.91783\n 0.815744 0.815744\n -0.958663 -0.958663\n -3.28825 -3.28825\n -6.37854 -6.37854\n 6.91577 6.91577\n 2.54565 2.54565\n 1.39487 1.39487\n 1.59679 1.59679\n 4.72347 4.72347\n 2.49221 2.49221\n 1.29896 1.29896\n -4.08232 -4.08232\n -0.648436 -0.648436\n -6.43531 -6.43531\n -0.556197 -0.556197\n -1.40304 -1.40304\n 0.699818 0.699818\n -5.29777 -5.29777\n -3.44335 -3.44335\n 7.35309 7.35309\n 8.846 8.846\n 8.39833 8.39833\n -2.71436 -2.71436\n 3.37063 3.37063\n -3.18723 -3.18723\n 1.32256 1.32256\n -3.09485 -3.09485\n 8.78146 8.78146\n -1.30004 -1.30004\n 3.03526 3.03526\n -1.4592 -1.4592\n 3.90288 3.90288\n -13.5124 -13.5124\n 1.35105 1.35105\n 3.37337 3.37337\n 2.5171 2.5171\n -4.22085 -4.22085\n 13.1858 13.1858\n -6.02839 -6.02839\n 5.75692 5.75692\n 2.46171 2.46171\n -0.950315 -0.950315\n -3.63255 -3.63255\n 1.88 1.88\n 5.48758 5.48758\n 4.96786 4.96786\n -6.17199 -6.17199\n -0.284244 -0.284244\n -1.80256 -1.80256\n 3.03221 3.03221\n -8.90171 -8.90171\n -8.66084 -8.66084\n -9.06366 -9.06366\n -3.02007 -3.02007\n -8.2276 -8.2276\n 8.10032 8.10032\n -4.11364 -4.11364\n -3.39291 -3.39291\n 3.64208 3.64208\n -0.739833 -0.739833\n -2.84156 -2.84156\n -0.843081 -0.843081\n -0.249744 -0.249744\n 7.05075 7.05075\n 0.369632 0.369632\n -1.90893 -1.90893\n 9.79465 9.79465\n 3.52356 3.52356\n 4.14091 4.14091\n 1.66568 1.66568\n -10.7162 -10.7162\n -7.64522 -7.64522\n 1.54688 1.54688\n 7.84479 7.84479\n 0.466458 0.466458\n 4.03315 4.03315\n 0.472926 0.472926\n 1.73319 1.73319\n 1.79317 1.79317\n 1.46234 1.46234\n -8.45267 -8.45267\n 7.30327 7.30327\n 3.08869 3.08869\n 5.27442 5.27442\n 2.92876 2.92876\n -1.6673 -1.6673\n 14.4442 14.4442\n 13.4055 13.4055\n -1.47522 -1.47522\n -3.57821 -3.57821\n 9.00659 9.00659\n -9.6723 -9.6723\n 2.8818 2.8818\n -2.61898 -2.61898\n 1.17927 1.17927\n -3.15135 -3.15135\n -0.976968 -0.976968\n 1.45062 1.45062\n 4.66687 4.66687\n 4.94346 4.94346\n -2.20375 -2.20375\n 2.93643 2.93643\n 7.51365 7.51365\n 6.50034 6.50034\n 1.74088 1.74088\n -4.43403 -4.43403\n 0.796894 0.796894\n -1.23803 -1.23803\n 5.33941 5.33941\n 4.90517 4.90517\n 0.569053 0.569053\n -0.609673 -0.609673\n 5.091 5.091\n 1.76184 1.76184\n -3.81174 -3.81174\n -5.39095 -5.39095\n -3.09718 -3.09718\n -1.87868 -1.87868\n -4.85278 -4.85278\n -1.05327 -1.05327\n -1.11892 -1.11892\n -3.52006 -3.52006\n -2.8466 -2.8466\n 3.03494 3.03494\n 3.7605 3.7605\n -1.8123 -1.8123\n -5.10186 -5.10186\n 2.85973 2.85973\n -3.6241 -3.6241\n 1.78302 1.78302\n -12.3108 -12.3108\n 0.378043 0.378043\n -1.70182 -1.70182\n -0.91773 -0.91773\n -5.37355 -5.37355","category":"page"},{"location":"examples/working_with_ollama/#Using-postprocessing-function","page":"Local models with Ollama.ai","title":"Using postprocessing function","text":"","category":"section"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"Add normalization as postprocessing function to normalize embeddings on reception (for easy cosine similarity later)","category":"page"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"using LinearAlgebra\nschema = PT.OllamaManagedSchema()\n\nmsg = aiembed(schema,\n [\"embed me\", \"and me too\"],\n LinearAlgebra.normalize;\n model = \"openhermes2.5-mistral\")","category":"page"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"DataMessage(Matrix{Float64} of size (4096, 2))","category":"page"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"Cosine similarity is then a simple multiplication","category":"page"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"msg.content' * msg.content[:, 1]","category":"page"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"2-element Vector{Float64}:\n 0.9999999999999946\n 0.34130017815042357","category":"page"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"","category":"page"},{"location":"examples/working_with_ollama/","page":"Local models with Ollama.ai","title":"Local models with Ollama.ai","text":"This page was generated using Literate.jl.","category":"page"},{"location":"frequently_asked_questions/#Frequently-Asked-Questions","page":"F.A.Q.","title":"Frequently Asked Questions","text":"","category":"section"},{"location":"frequently_asked_questions/#Why-OpenAI","page":"F.A.Q.","title":"Why OpenAI","text":"","category":"section"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"OpenAI's models are at the forefront of AI research and provide robust, state-of-the-art capabilities for many tasks.","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"There will be situations not or cannot use it (eg, privacy, cost, etc.). In that case, you can use local models (eg, Ollama) or other APIs (eg, Anthropic).","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"Note: To get started with Ollama.ai, see the Setup Guide for Ollama section below.","category":"page"},{"location":"frequently_asked_questions/#Data-Privacy-and-OpenAI","page":"F.A.Q.","title":"Data Privacy and OpenAI","text":"","category":"section"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"At the time of writing, OpenAI does NOT use the API calls for training their models.","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"APIOpenAI does not use data submitted to and generated by our API to train OpenAI models or improve OpenAI’s service offering. In order to support the continuous improvement of our models, you can fill out this form to opt-in to share your data with us. – How your data is used to improve our models","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"You can always double-check the latest information on the OpenAI's How we use your data page.","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"Resources:","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"OpenAI's How we use your data\nData usage for consumer services FAQ\nHow your data is used to improve our models","category":"page"},{"location":"frequently_asked_questions/#Creating-OpenAI-API-Key","page":"F.A.Q.","title":"Creating OpenAI API Key","text":"","category":"section"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"You can get your API key from OpenAI by signing up for an account and accessing the API section of the OpenAI website.","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"Create an account with OpenAI\nGo to API Key page\nClick on “Create new secret key”","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"!!! Do not share it with anyone and do NOT save it to any files that get synced online.","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"Resources:","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"OpenAI Documentation\nVisual tutorial","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"Pro tip: Always set the spending limits!","category":"page"},{"location":"frequently_asked_questions/#Setting-OpenAI-Spending-Limits","page":"F.A.Q.","title":"Setting OpenAI Spending Limits","text":"","category":"section"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"OpenAI allows you to set spending limits directly on your account dashboard to prevent unexpected costs.","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"Go to OpenAI Billing\nSet Soft Limit (you’ll receive a notification) and Hard Limit (API will stop working not to spend more money)","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"A good start might be a soft limit of c.5 and a hard limit of c10 - you can always increase it later in the month.","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"Resources:","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"OpenAI Forum","category":"page"},{"location":"frequently_asked_questions/#How-much-does-it-cost?-Is-it-worth-paying-for?","page":"F.A.Q.","title":"How much does it cost? Is it worth paying for?","text":"","category":"section"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"If you use a local model (eg, with Ollama), it's free. If you use any commercial APIs (eg, OpenAI), you will likely pay per \"token\" (a sub-word unit).","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"For example, a simple request with a simple question and 1 sentence response in return (”Is statement XYZ a positive comment”) will cost you ~0.0001 (ie, one hundredth of a cent)","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"Is it worth paying for?","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"GenAI is a way to buy time! You can pay cents to save tens of minutes every day.","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"Continuing the example above, imagine you have a table with 200 comments. Now, you can parse each one of them with an LLM for the features/checks you need. Assuming the price per call was 0.0001, you'd pay 2 cents for the job and save 30-60 minutes of your time!","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"Resources:","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"OpenAI Pricing per 1000 tokens","category":"page"},{"location":"frequently_asked_questions/#Configuring-the-Environment-Variable-for-API-Key","page":"F.A.Q.","title":"Configuring the Environment Variable for API Key","text":"","category":"section"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"To use the OpenAI API with PromptingTools.jl, set your API key as an environment variable:","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"ENV[\"OPENAI_API_KEY\"] = \"your-api-key\"","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"As a one-off, you can: ","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"set it in the terminal before launching Julia: export OPENAI_API_KEY = \nset it in your setup.jl (make sure not to commit it to GitHub!)","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"Make sure to start Julia from the same terminal window where you set the variable. Easy check in Julia, run ENV[\"OPENAI_API_KEY\"] and you should see your key!","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"A better way:","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"On a Mac, add the configuration line to your terminal's configuration file (eg, ~/.zshrc). It will get automatically loaded every time you launch the terminal\nOn Windows, set it as a system variable in \"Environment Variables\" settings (see the Resources)","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"Resources: ","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"OpenAI Guide","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"Note: In the future, we hope to add Preferences.jl-based workflow to set the API key and other preferences.","category":"page"},{"location":"frequently_asked_questions/#Understanding-the-API-Keyword-Arguments-in-aigenerate-(api_kwargs)","page":"F.A.Q.","title":"Understanding the API Keyword Arguments in aigenerate (api_kwargs)","text":"","category":"section"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"See OpenAI API reference for more information.","category":"page"},{"location":"frequently_asked_questions/#Instant-Access-from-Anywhere","page":"F.A.Q.","title":"Instant Access from Anywhere","text":"","category":"section"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"For easy access from anywhere, add PromptingTools into your startup.jl (can be found in ~/.julia/config/startup.jl).","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"Add the following snippet:","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"using PromptingTools\nconst PT = PromptingTools # to access unexported functions and types","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"Now, you can just use ai\"Help me do X to achieve Y\" from any REPL session!","category":"page"},{"location":"frequently_asked_questions/#Open-Source-Alternatives","page":"F.A.Q.","title":"Open Source Alternatives","text":"","category":"section"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"The ethos of PromptingTools.jl is to allow you to use whatever model you want, which includes Open Source LLMs. The most popular and easiest to setup is Ollama.ai - see below for more information.","category":"page"},{"location":"frequently_asked_questions/#Setup-Guide-for-Ollama","page":"F.A.Q.","title":"Setup Guide for Ollama","text":"","category":"section"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"Ollama runs a background service hosting LLMs that you can access via a simple API. It's especially useful when you're working with some sensitive data that should not be sent anywhere.","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"Installation is very easy, just download the latest version here.","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"Once you've installed it, just launch the app and you're ready to go!","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"To check if it's running, go to your browser and open 127.0.0.1:11434. You should see the message \"Ollama is running\". Alternatively, you can run ollama serve in your terminal and you'll get a message that it's already running.","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"There are many models available in Ollama Library, including Llama2, CodeLlama, SQLCoder, or my personal favorite openhermes2.5-mistral.","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"Download new models with ollama pull (eg, ollama pull openhermes2.5-mistral). ","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"Show currently available models with ollama list.","category":"page"},{"location":"frequently_asked_questions/","page":"F.A.Q.","title":"F.A.Q.","text":"See Ollama.ai for more information.","category":"page"},{"location":"reference/#Reference","page":"Reference","title":"Reference","text":"","category":"section"},{"location":"reference/","page":"Reference","title":"Reference","text":"","category":"page"},{"location":"reference/","page":"Reference","title":"Reference","text":"Modules = [PromptingTools]","category":"page"},{"location":"reference/#PromptingTools.AITemplate","page":"Reference","title":"PromptingTools.AITemplate","text":"AITemplate\n\nAITemplate is a template for a conversation prompt. This type is merely a container for the template name, which is resolved into a set of messages (=prompt) by render.\n\nNaming Convention\n\nTemplate names should be in CamelCase\nFollow the format ...... where possible, eg, JudgeIsItTrue, ``\nStarting with the Persona (=System prompt), eg, Judge = persona is meant to judge some provided information\nVariable to be filled in with context, eg, It = placeholder it\nEnding with the variable name is helpful, eg, JuliaExpertTask for a persona to be an expert in Julia language and task is the placeholder name\nIdeally, the template name should be self-explanatory, eg, JudgeIsItTrue = persona is meant to judge some provided information where it is true or false\n\nExamples\n\nSave time by re-using pre-made templates, just fill in the placeholders with the keyword arguments:\n\nmsg = aigenerate(:JuliaExpertAsk; ask = \"How do I add packages?\")\n\nThe above is equivalent to a more verbose version that explicitly uses the dispatch on AITemplate:\n\nmsg = aigenerate(AITemplate(:JuliaExpertAsk); ask = \"How do I add packages?\")\n\nFind available templates with aitemplates:\n\ntmps = aitemplates(\"JuliaExpertAsk\")\n# Will surface one specific template\n# 1-element Vector{AITemplateMetadata}:\n# PromptingTools.AITemplateMetadata\n# name: Symbol JuliaExpertAsk\n# description: String \"For asking questions about Julia language. Placeholders: `ask`\"\n# version: String \"1\"\n# wordcount: Int64 237\n# variables: Array{Symbol}((1,))\n# system_preview: String \"You are a world-class Julia language programmer with the knowledge of the latest syntax. Your commun\"\n# user_preview: String \"# Question\n\n{{ask}}\"\n# source: String \"\"\n\nThe above gives you a good idea of what the template is about, what placeholders are available, and how much it would cost to use it (=wordcount).\n\nSearch for all Julia-related templates:\n\ntmps = aitemplates(\"Julia\")\n# 2-element Vector{AITemplateMetadata}... -> more to come later!\n\nIf you are on VSCode, you can leverage nice tabular display with vscodedisplay:\n\nusing DataFrames\ntmps = aitemplates(\"Julia\") |> DataFrame |> vscodedisplay\n\nI have my selected template, how do I use it? Just use the \"name\" in aigenerate or aiclassify like you see in the first example!\n\nYou can inspect any template by \"rendering\" it (this is what the LLM will see):\n\njulia> AITemplate(:JudgeIsItTrue) |> PromptingTools.render\n\nSee also: save_template, load_template, load_templates! for more advanced use cases (and the corresponding script in examples/ folder)\n\n\n\n\n\n","category":"type"},{"location":"reference/#PromptingTools.AITemplateMetadata","page":"Reference","title":"PromptingTools.AITemplateMetadata","text":"Helper for easy searching and reviewing of templates. Defined on loading of each template.\n\n\n\n\n\n","category":"type"},{"location":"reference/#PromptingTools.AbstractPromptSchema","page":"Reference","title":"PromptingTools.AbstractPromptSchema","text":"Defines different prompting styles based on the model training and fine-tuning.\n\n\n\n\n\n","category":"type"},{"location":"reference/#PromptingTools.ChatMLSchema","page":"Reference","title":"PromptingTools.ChatMLSchema","text":"ChatMLSchema is used by many open-source chatbots, by OpenAI models (under the hood) and by several models and inferfaces (eg, Ollama, vLLM)\n\nYou can explore it on tiktokenizer\n\nIt uses the following conversation structure:\n\nsystem\n...\n<|im_start|>user\n...<|im_end|>\n<|im_start|>assistant\n...<|im_end|>\n\n\n\n\n\n","category":"type"},{"location":"reference/#PromptingTools.MaybeExtract","page":"Reference","title":"PromptingTools.MaybeExtract","text":"Extract a result from the provided data, if any, otherwise set the error and message fields.\n\nArguments\n\nerror::Bool: true if a result is found, false otherwise.\nmessage::String: Only present if no result is found, should be short and concise.\n\n\n\n\n\n","category":"type"},{"location":"reference/#PromptingTools.OllamaManagedSchema","page":"Reference","title":"PromptingTools.OllamaManagedSchema","text":"Ollama by default manages different models and their associated prompt schemas when you pass system_prompt and prompt fields to the API.\n\nWarning: It works only for 1 system message and 1 user message, so anything more than that has to be rejected.\n\nIf you need to pass more messagese / longer conversational history, you can use define the model-specific schema directly and pass your Ollama requests with raw=true, which disables and templating and schema management by Ollama.\n\n\n\n\n\n","category":"type"},{"location":"reference/#PromptingTools.OpenAISchema","page":"Reference","title":"PromptingTools.OpenAISchema","text":"OpenAISchema is the default schema for OpenAI models.\n\nIt uses the following conversation template:\n\n[Dict(role=\"system\",content=\"...\"),Dict(role=\"user\",content=\"...\"),Dict(role=\"assistant\",content=\"...\")]\n\nIt's recommended to separate sections in your prompt with markdown headers (e.g. `##Answer\n\n`).\n\n\n\n\n\n","category":"type"},{"location":"reference/#PromptingTools.TestEchoOllamaManagedSchema","page":"Reference","title":"PromptingTools.TestEchoOllamaManagedSchema","text":"Echoes the user's input back to them. Used for testing the implementation\n\n\n\n\n\n","category":"type"},{"location":"reference/#PromptingTools.TestEchoOpenAISchema","page":"Reference","title":"PromptingTools.TestEchoOpenAISchema","text":"Echoes the user's input back to them. Used for testing the implementation\n\n\n\n\n\n","category":"type"},{"location":"reference/#PromptingTools.UserMessageWithImages-Tuple{AbstractString}","page":"Reference","title":"PromptingTools.UserMessageWithImages","text":"Construct UserMessageWithImages with 1 or more images. Images can be either URLs or local paths.\n\n\n\n\n\n","category":"method"},{"location":"reference/#PromptingTools.X123","page":"Reference","title":"PromptingTools.X123","text":"With docstring\n\n\n\n\n\n","category":"type"},{"location":"reference/#PromptingTools.aiclassify-Tuple{PromptingTools.AbstractOpenAISchema, Union{AbstractString, PromptingTools.AbstractMessage, Vector{<:PromptingTools.AbstractMessage}}}","page":"Reference","title":"PromptingTools.aiclassify","text":"aiclassify(prompt_schema::AbstractOpenAISchema, prompt::ALLOWED_PROMPT_TYPE;\napi_kwargs::NamedTuple = (logit_bias = Dict(837 => 100, 905 => 100, 9987 => 100),\n max_tokens = 1, temperature = 0),\nkwargs...)\n\nClassifies the given prompt/statement as true/false/unknown.\n\nNote: this is a very simple classifier, it is not meant to be used in production. Credit goes to AAAzzam.\n\nIt uses Logit bias trick and limits the output to 1 token to force the model to output only true/false/unknown.\n\nOutput tokens used (via api_kwargs):\n\n837: ' true'\n905: ' false'\n9987: ' unknown'\n\nArguments\n\nprompt_schema::AbstractOpenAISchema: The schema for the prompt.\nprompt: The prompt/statement to classify if it's a String. If it's a Symbol, it is expanded as a template via render(schema,template).\n\nExample\n\naiclassify(\"Is two plus two four?\") # true\naiclassify(\"Is two plus three a vegetable on Mars?\") # false\n\naiclassify returns only true/false/unknown. It's easy to get the proper Bool output type out with tryparse, eg,\n\ntryparse(Bool, aiclassify(\"Is two plus two four?\")) isa Bool # true\n\nOutput of type Nothing marks that the model couldn't classify the statement as true/false.\n\nIdeally, we would like to re-use some helpful system prompt to get more accurate responses. For this reason we have templates, eg, :JudgeIsItTrue. By specifying the template, we can provide our statement as the expected variable (it in this case) See that the model now correctly classifies the statement as \"unknown\".\n\naiclassify(:JudgeIsItTrue; it = \"Is two plus three a vegetable on Mars?\") # unknown\n\nFor better results, use higher quality models like gpt4, eg, \n\naiclassify(:JudgeIsItTrue;\n it = \"If I had two apples and I got three more, I have five apples now.\",\n model = \"gpt4\") # true\n\n\n\n\n\n","category":"method"},{"location":"reference/#PromptingTools.aiembed-Union{Tuple{F}, Tuple{PromptingTools.AbstractOllamaManagedSchema, AbstractString}, Tuple{PromptingTools.AbstractOllamaManagedSchema, AbstractString, F}} where F<:Function","page":"Reference","title":"PromptingTools.aiembed","text":"aiembed(prompt_schema::AbstractOllamaManagedSchema,\n doc_or_docs::Union{AbstractString, Vector{<:AbstractString}},\n postprocess::F = identity;\n verbose::Bool = true,\n api_key::String = API_KEY,\n model::String = MODEL_EMBEDDING,\n http_kwargs::NamedTuple = (retry_non_idempotent = true,\n retries = 5,\n readtimeout = 120),\n api_kwargs::NamedTuple = NamedTuple(),\n kwargs...) where {F <: Function}\n\nThe aiembed function generates embeddings for the given input using a specified model and returns a message object containing the embeddings, status, token count, and elapsed time.\n\nArguments\n\nprompt_schema::AbstractOllamaManagedSchema: The schema for the prompt.\ndoc_or_docs::Union{AbstractString, Vector{<:AbstractString}}: The document or list of documents to generate embeddings for. The list of documents is processed sequentially, so users should consider implementing an async version with with Threads.@spawn\npostprocess::F: The post-processing function to apply to each embedding. Defaults to the identity function, but could be LinearAlgebra.normalize.\nverbose::Bool: A flag indicating whether to print verbose information. Defaults to true.\napi_key::String: The API key to use for the OpenAI API. Defaults to API_KEY.\nmodel::String: The model to use for generating embeddings. Defaults to MODEL_EMBEDDING.\nhttp_kwargs::NamedTuple: Additional keyword arguments for the HTTP request. Defaults to empty NamedTuple.\napi_kwargs::NamedTuple: Additional keyword arguments for the Ollama API. Defaults to an empty NamedTuple.\nkwargs: Prompt variables to be used to fill the prompt/template\n\nReturns\n\nmsg: A DataMessage object containing the embeddings, status, token count, and elapsed time.\n\nNote: Ollama API currently does not return the token count, so it's set to (0,0)\n\nExample\n\nconst PT = PromptingTools\nschema = PT.OllamaManagedSchema()\n\nmsg = aiembed(schema, \"Hello World\"; model=\"openhermes2.5-mistral\")\nmsg.content # 4096-element JSON3.Array{Float64...\n\nWe can embed multiple strings at once and they will be hcat into a matrix (ie, each column corresponds to one string)\n\nconst PT = PromptingTools\nschema = PT.OllamaManagedSchema()\n\nmsg = aiembed(schema, [\"Hello World\", \"How are you?\"]; model=\"openhermes2.5-mistral\")\nmsg.content # 4096×2 Matrix{Float64}:\n\nIf you plan to calculate the cosine distance between embeddings, you can normalize them first:\n\nconst PT = PromptingTools\nusing LinearAlgebra\nschema = PT.OllamaManagedSchema()\n\nmsg = aiembed(schema, [\"embed me\", \"and me too\"], LinearAlgebra.normalize; model=\"openhermes2.5-mistral\")\n\n# calculate cosine distance between the two normalized embeddings as a simple dot product\nmsg.content' * msg.content[:, 1] # [1.0, 0.34]\n\nSimilarly, you can use the postprocess argument to materialize the data from JSON3.Object by using postprocess = copy\n\nconst PT = PromptingTools\nschema = PT.OllamaManagedSchema()\n\nmsg = aiembed(schema, \"Hello World\", copy; model=\"openhermes2.5-mistral\")\nmsg.content # 4096-element Vector{Float64}\n\n\n\n\n\n","category":"method"},{"location":"reference/#PromptingTools.aiembed-Union{Tuple{F}, Tuple{PromptingTools.AbstractOpenAISchema, Union{AbstractString, Vector{<:AbstractString}}}, Tuple{PromptingTools.AbstractOpenAISchema, Union{AbstractString, Vector{<:AbstractString}}, F}} where F<:Function","page":"Reference","title":"PromptingTools.aiembed","text":"aiembed(prompt_schema::AbstractOpenAISchema,\n doc_or_docs::Union{AbstractString, Vector{<:AbstractString}},\n postprocess::F = identity;\n verbose::Bool = true,\n api_key::String = API_KEY,\n model::String = MODEL_EMBEDDING,\n http_kwargs::NamedTuple = (retry_non_idempotent = true,\n retries = 5,\n readtimeout = 120),\n api_kwargs::NamedTuple = NamedTuple(),\n kwargs...) where {F <: Function}\n\nThe aiembed function generates embeddings for the given input using a specified model and returns a message object containing the embeddings, status, token count, and elapsed time.\n\nArguments\n\nprompt_schema::AbstractOpenAISchema: The schema for the prompt.\ndoc_or_docs::Union{AbstractString, Vector{<:AbstractString}}: The document or list of documents to generate embeddings for.\npostprocess::F: The post-processing function to apply to each embedding. Defaults to the identity function.\nverbose::Bool: A flag indicating whether to print verbose information. Defaults to true.\napi_key::String: The API key to use for the OpenAI API. Defaults to API_KEY.\nmodel::String: The model to use for generating embeddings. Defaults to MODEL_EMBEDDING.\nhttp_kwargs::NamedTuple: Additional keyword arguments for the HTTP request. Defaults to (retry_non_idempotent = true, retries = 5, readtimeout = 120).\napi_kwargs::NamedTuple: Additional keyword arguments for the OpenAI API. Defaults to an empty NamedTuple.\nkwargs...: Additional keyword arguments.\n\nReturns\n\nmsg: A DataMessage object containing the embeddings, status, token count, and elapsed time.\n\nExample\n\nmsg = aiembed(\"Hello World\")\nmsg.content # 1536-element JSON3.Array{Float64...\n\nWe can embed multiple strings at once and they will be hcat into a matrix (ie, each column corresponds to one string)\n\nmsg = aiembed([\"Hello World\", \"How are you?\"])\nmsg.content # 1536×2 Matrix{Float64}:\n\nIf you plan to calculate the cosine distance between embeddings, you can normalize them first:\n\nusing LinearAlgebra\nmsg = aiembed([\"embed me\", \"and me too\"], LinearAlgebra.normalize)\n\n# calculate cosine distance between the two normalized embeddings as a simple dot product\nmsg.content' * msg.content[:, 1] # [1.0, 0.787]\n\n\n\n\n\n","category":"method"},{"location":"reference/#PromptingTools.aiextract-Tuple{PromptingTools.AbstractOpenAISchema, Union{AbstractString, PromptingTools.AbstractMessage, Vector{<:PromptingTools.AbstractMessage}}}","page":"Reference","title":"PromptingTools.aiextract","text":"aiextract([prompt_schema::AbstractOpenAISchema,] prompt::ALLOWED_PROMPT_TYPE; \nreturn_type::Type,\nverbose::Bool = true,\n model::String = MODEL_CHAT,\n http_kwargs::NamedTuple = (;\n retry_non_idempotent = true,\n retries = 5,\n readtimeout = 120), api_kwargs::NamedTuple = NamedTuple(),\n kwargs...)\n\nExtract required information (defined by a struct return_type) from the provided prompt by leveraging OpenAI function calling mode.\n\nThis is a perfect solution for extracting structured information from text (eg, extract organization names in news articles, etc.)\n\nIt's effectively a light wrapper around aigenerate call, which requires additional keyword argument return_type to be provided and will enforce the model outputs to adhere to it.\n\nArguments\n\nprompt_schema: An optional object to specify which prompt template should be applied (Default to PROMPT_SCHEMA = OpenAISchema)\nprompt: Can be a string representing the prompt for the AI conversation, a UserMessage, a vector of AbstractMessage or an AITemplate\nreturn_type: A struct TYPE representing the the information we want to extract. Do not provide a struct instance, only the type. If the struct has a docstring, it will be provided to the model as well. It's used to enforce structured model outputs or provide more information.\nverbose: A boolean indicating whether to print additional information.\napi_key: A string representing the API key for accessing the OpenAI API.\nmodel: A string representing the model to use for generating the response. Can be an alias corresponding to a model ID defined in MODEL_ALIASES.\nhttp_kwargs: A named tuple of HTTP keyword arguments.\napi_kwargs: A named tuple of API keyword arguments.\nkwargs: Prompt variables to be used to fill the prompt/template\n\nReturns\n\nmsg: An DataMessage object representing the extracted data, including the content, status, tokens, and elapsed time. Use msg.content to access the extracted data.\n\nSee also: function_call_signature, MaybeExtract, aigenerate\n\nExample\n\nDo you want to extract some specific measurements from a text like age, weight and height? You need to define the information you need as a struct (return_type):\n\n\"Person's age, height, and weight.\"\nstruct MyMeasurement\n age::Int # required\n height::Union{Int,Nothing} # optional\n weight::Union{Nothing,Float64} # optional\nend\nmsg = aiextract(\"James is 30, weighs 80kg. He's 180cm tall.\"; return_type=MyMeasurement)\n# [ Info: Tokens: 129 @ Cost: $0.0002 in 1.0 seconds\n# PromptingTools.DataMessage(MyMeasurement)\nmsg.content\n# MyMeasurement(30, 180, 80.0)\n\nThe fields that allow Nothing are marked as optional in the schema:\n\nmsg = aiextract(\"James is 30.\"; return_type=MyMeasurement)\n# MyMeasurement(30, nothing, nothing)\n\nIf there are multiple items you want to extract, define a wrapper struct to get a Vector of MyMeasurement:\n\nstruct MyMeasurementWrapper\n measurements::Vector{MyMeasurement}\nend\n\nmsg = aiextract(\"James is 30, weighs 80kg. He's 180cm tall. Then Jack is 19 but really tall - over 190!\"; return_type=ManyMeasurements)\n\nmsg.content.measurements\n# 2-element Vector{MyMeasurement}:\n# MyMeasurement(30, 180, 80.0)\n# MyMeasurement(19, 190, nothing)\n\nOr if you want your extraction to fail gracefully when data isn't found, use MaybeExtract{T} wrapper (this trick is inspired by the Instructor package!):\n\nusing PromptingTools: MaybeExtract\n\ntype = MaybeExtract{MyMeasurement}\n# Effectively the same as:\n# struct MaybeExtract{T}\n# result::Union{T, Nothing} // The result of the extraction\n# error::Bool // true if a result is found, false otherwise\n# message::Union{Nothing, String} // Only present if no result is found, should be short and concise\n# end\n\n# If LLM extraction fails, it will return a Dict with `error` and `message` fields instead of the result!\nmsg = aiextract(\"Extract measurements from the text: I am giraffe\", type)\nmsg.content\n# MaybeExtract{MyMeasurement}(nothing, true, \"I'm sorry, but I can only assist with human measurements.\")\n\nThat way, you can handle the error gracefully and get a reason why extraction failed (in msg.content.message).\n\nNote that the error message refers to a giraffe not being a human, because in our MyMeasurement docstring, we said that it's for people!\n\n\n\n\n\n","category":"method"},{"location":"reference/#PromptingTools.aigenerate-Tuple{PromptingTools.AbstractOllamaManagedSchema, Union{AbstractString, PromptingTools.AbstractMessage, Vector{<:PromptingTools.AbstractMessage}}}","page":"Reference","title":"PromptingTools.aigenerate","text":"aigenerate(prompt_schema::AbstractOllamaManagedSchema, prompt::ALLOWED_PROMPT_TYPE; verbose::Bool = true,\n model::String = MODEL_CHAT,\n http_kwargs::NamedTuple = NamedTuple(), api_kwargs::NamedTuple = NamedTuple(),\n kwargs...)\n\nGenerate an AI response based on a given prompt using the OpenAI API.\n\nArguments\n\nprompt_schema: An optional object to specify which prompt template should be applied (Default to PROMPT_SCHEMA = OpenAISchema not AbstractManagedSchema)\nprompt: Can be a string representing the prompt for the AI conversation, a UserMessage, a vector of AbstractMessage or an AITemplate\nverbose: A boolean indicating whether to print additional information.\napi_key: Provided for interface consistency. Not needed for locally hosted Ollama.\nmodel: A string representing the model to use for generating the response. Can be an alias corresponding to a model ID defined in MODEL_ALIASES.\nhttp_kwargs::NamedTuple: Additional keyword arguments for the HTTP request. Defaults to empty NamedTuple.\napi_kwargs::NamedTuple: Additional keyword arguments for the Ollama API. Defaults to an empty NamedTuple.\nkwargs: Prompt variables to be used to fill the prompt/template\n\nReturns\n\nmsg: An AIMessage object representing the generated AI message, including the content, status, tokens, and elapsed time.\n\nUse msg.content to access the extracted string.\n\nSee also: ai_str, aai_str, aiembed\n\nExample\n\nSimple hello world to test the API:\n\nconst PT = PromptingTools\nschema = PT.OllamaManagedSchema() # We need to explicit if we want Ollama, OpenAISchema is the default\n\nmsg = aigenerate(schema, \"Say hi!\"; model=\"openhermes2.5-mistral\")\n# [ Info: Tokens: 69 in 0.9 seconds\n# AIMessage(\"Hello! How can I assist you today?\")\n\nmsg is an AIMessage object. Access the generated string via content property:\n\ntypeof(msg) # AIMessage{SubString{String}}\npropertynames(msg) # (:content, :status, :tokens, :elapsed\nmsg.content # \"Hello! How can I assist you today?\"\n\nNote: We need to be explicit about the schema we want to use. If we don't, it will default to OpenAISchema (=PT.DEFAULT_SCHEMA) ___ You can use string interpolation:\n\nconst PT = PromptingTools\nschema = PT.OllamaManagedSchema()\na = 1\nmsg=aigenerate(schema, \"What is `$a+$a`?\"; model=\"openhermes2.5-mistral\")\nmsg.content # \"The result of `1+1` is `2`.\"\n\n___ You can provide the whole conversation or more intricate prompts as a Vector{AbstractMessage}:\n\nconst PT = PromptingTools\nschema = PT.OllamaManagedSchema()\n\nconversation = [\n PT.SystemMessage(\"You're master Yoda from Star Wars trying to help the user become a Yedi.\"),\n PT.UserMessage(\"I have feelings for my iPhone. What should I do?\")]\n\nmsg = aigenerate(schema, conversation; model=\"openhermes2.5-mistral\")\n# [ Info: Tokens: 111 in 2.1 seconds\n# AIMessage(\"Strong the attachment is, it leads to suffering it may. Focus on the force within you must, ...\")\n\nNote: Managed Ollama currently supports at most 1 User Message and 1 System Message given the API limitations. If you want more, you need to use the ChatMLSchema.\n\n\n\n\n\n","category":"method"},{"location":"reference/#PromptingTools.aigenerate-Tuple{PromptingTools.AbstractOpenAISchema, Union{AbstractString, PromptingTools.AbstractMessage, Vector{<:PromptingTools.AbstractMessage}}}","page":"Reference","title":"PromptingTools.aigenerate","text":"aigenerate([prompt_schema::AbstractOpenAISchema,] prompt::ALLOWED_PROMPT_TYPE; verbose::Bool = true,\n model::String = MODEL_CHAT,\n http_kwargs::NamedTuple = (;\n retry_non_idempotent = true,\n retries = 5,\n readtimeout = 120), api_kwargs::NamedTuple = NamedTuple(),\n kwargs...)\n\nGenerate an AI response based on a given prompt using the OpenAI API.\n\nArguments\n\nprompt_schema: An optional object to specify which prompt template should be applied (Default to PROMPT_SCHEMA = OpenAISchema)\nprompt: Can be a string representing the prompt for the AI conversation, a UserMessage, a vector of AbstractMessage or an AITemplate\nverbose: A boolean indicating whether to print additional information.\napi_key: A string representing the API key for accessing the OpenAI API.\nmodel: A string representing the model to use for generating the response. Can be an alias corresponding to a model ID defined in MODEL_ALIASES.\nhttp_kwargs: A named tuple of HTTP keyword arguments.\napi_kwargs: A named tuple of API keyword arguments.\nkwargs: Prompt variables to be used to fill the prompt/template\n\nReturns\n\nmsg: An AIMessage object representing the generated AI message, including the content, status, tokens, and elapsed time.\n\nUse msg.content to access the extracted string.\n\nSee also: ai_str, aai_str, aiembed, aiclassify, aiextract, aiscan\n\nExample\n\nSimple hello world to test the API:\n\nresult = aigenerate(\"Say Hi!\")\n# [ Info: Tokens: 29 @ Cost: $0.0 in 1.0 seconds\n# AIMessage(\"Hello! How can I assist you today?\")\n\nresult is an AIMessage object. Access the generated string via content property:\n\ntypeof(result) # AIMessage{SubString{String}}\npropertynames(result) # (:content, :status, :tokens, :elapsed\nresult.content # \"Hello! How can I assist you today?\"\n\n___ You can use string interpolation:\n\na = 1\nmsg=aigenerate(\"What is `$a+$a`?\")\nmsg.content # \"The sum of `1+1` is `2`.\"\n\n___ You can provide the whole conversation or more intricate prompts as a Vector{AbstractMessage}:\n\nconst PT = PromptingTools\n\nconversation = [\n PT.SystemMessage(\"You're master Yoda from Star Wars trying to help the user become a Yedi.\"),\n PT.UserMessage(\"I have feelings for my iPhone. What should I do?\")]\nmsg=aigenerate(conversation)\n# AIMessage(\"Ah, strong feelings you have for your iPhone. A Jedi's path, this is not... \")\n\n\n\n\n\n","category":"method"},{"location":"reference/#PromptingTools.aiscan-Tuple{PromptingTools.AbstractOpenAISchema, Union{AbstractString, PromptingTools.AbstractMessage, Vector{<:PromptingTools.AbstractMessage}}}","page":"Reference","title":"PromptingTools.aiscan","text":"aiscan([promptschema::AbstractOpenAISchema,] prompt::ALLOWEDPROMPTTYPE; imageurl::Union{Nothing, AbstractString, Vector{<:AbstractString}} = nothing, imagepath::Union{Nothing, AbstractString, Vector{<:AbstractString}} = nothing, imagedetail::AbstractString = \"auto\", attachtolatest::Bool = true, verbose::Bool = true, model::String = MODELCHAT, httpkwargs::NamedTuple = (; retrynonidempotent = true, retries = 5, readtimeout = 120), apikwargs::NamedTuple = = (; maxtokens = 2500), kwargs...)\n\nScans the provided image (image_url or image_path) with the goal provided in the prompt.\n\nCan be used for many multi-modal tasks, such as: OCR (transcribe text in the image), image captioning, image classification, etc.\n\nIt's effectively a light wrapper around aigenerate call, which uses additional keyword arguments image_url, image_path, image_detail to be provided. At least one image source (url or path) must be provided.\n\nArguments\n\nprompt_schema: An optional object to specify which prompt template should be applied (Default to PROMPT_SCHEMA = OpenAISchema)\nprompt: Can be a string representing the prompt for the AI conversation, a UserMessage, a vector of AbstractMessage or an AITemplate\nimage_url: A string or vector of strings representing the URL(s) of the image(s) to scan.\nimage_path: A string or vector of strings representing the path(s) of the image(s) to scan.\nimage_detail: A string representing the level of detail to include for images. Can be \"auto\", \"high\", or \"low\". See OpenAI Vision Guide for more details.\nattach_to_latest: A boolean how to handle if a conversation with multiple UserMessage is provided. When true, the images are attached to the latest UserMessage.\nverbose: A boolean indicating whether to print additional information.\napi_key: A string representing the API key for accessing the OpenAI API.\nmodel: A string representing the model to use for generating the response. Can be an alias corresponding to a model ID defined in MODEL_ALIASES.\nhttp_kwargs: A named tuple of HTTP keyword arguments.\napi_kwargs: A named tuple of API keyword arguments.\nkwargs: Prompt variables to be used to fill the prompt/template\n\nReturns\n\nmsg: An AIMessage object representing the generated AI message, including the content, status, tokens, and elapsed time.\n\nUse msg.content to access the extracted string.\n\nSee also: ai_str, aai_str, aigenerate, aiembed, aiclassify, aiextract\n\nNotes\n\nAll examples below use model \"gpt4v\", which is an alias for model ID \"gpt-4-vision-preview\"\nmax_tokens in the api_kwargs is preset to 2500, otherwise OpenAI enforces a default of only a few hundred tokens (~300). If your output is truncated, increase this value\n\nExample\n\nDescribe the provided image:\n\nmsg = aiscan(\"Describe the image\"; image_path=\"julia.png\", model=\"gpt4v\")\n# [ Info: Tokens: 1141 @ Cost: $0.0117 in 2.2 seconds\n# AIMessage(\"The image shows a logo consisting of the word \"julia\" written in lowercase\")\n\nYou can provide multiple images at once as a vector and ask for \"low\" level of detail (cheaper):\n\nmsg = aiscan(\"Describe the image\"; image_path=[\"julia.png\",\"python.png\"], image_detail=\"low\", model=\"gpt4v\")\n\nYou can use this function as a nice and quick OCR (transcribe text in the image) with a template :OCRTask. Let's transcribe some SQL code from a screenshot (no more re-typing!):\n\n# Screenshot of some SQL code\nimage_url = \"https://www.sqlservercentral.com/wp-content/uploads/legacy/8755f69180b7ac7ee76a69ae68ec36872a116ad4/24622.png\"\nmsg = aiscan(:OCRTask; image_url, model=\"gpt4v\", task=\"Transcribe the SQL code in the image.\", api_kwargs=(; max_tokens=2500))\n\n# [ Info: Tokens: 362 @ Cost: $0.0045 in 2.5 seconds\n# AIMessage(\"```sql\n# update Orders \n\n# You can add syntax highlighting of the outputs via Markdown\nusing Markdown\nmsg.content |> Markdown.parse\n\nNotice that we enforce max_tokens = 2500. That's because OpenAI seems to default to ~300 tokens, which provides incomplete outputs. Hence, we set this value to 2500 as a default. If you still get truncated outputs, increase this value.\n\n\n\n\n\n","category":"method"},{"location":"reference/#PromptingTools.aitemplates","page":"Reference","title":"PromptingTools.aitemplates","text":"aitemplates\n\nFind easily the most suitable templates for your use case.\n\nYou can search by:\n\nquery::Symbol which looks look only for partial matches in the template name\nquery::AbstractString which looks for partial matches in the template name or description\nquery::Regex which looks for matches in the template name, description or any of the message previews\n\nKeyword Arguments\n\nlimit::Int limits the number of returned templates (Defaults to 10)\n\nExamples\n\nFind available templates with aitemplates:\n\ntmps = aitemplates(\"JuliaExpertAsk\")\n# Will surface one specific template\n# 1-element Vector{AITemplateMetadata}:\n# PromptingTools.AITemplateMetadata\n# name: Symbol JuliaExpertAsk\n# description: String \"For asking questions about Julia language. Placeholders: `ask`\"\n# version: String \"1\"\n# wordcount: Int64 237\n# variables: Array{Symbol}((1,))\n# system_preview: String \"You are a world-class Julia language programmer with the knowledge of the latest syntax. Your commun\"\n# user_preview: String \"# Question\n\n{{ask}}\"\n# source: String \"\"\n\nThe above gives you a good idea of what the template is about, what placeholders are available, and how much it would cost to use it (=wordcount).\n\nSearch for all Julia-related templates:\n\ntmps = aitemplates(\"Julia\")\n# 2-element Vector{AITemplateMetadata}... -> more to come later!\n\nIf you are on VSCode, you can leverage nice tabular display with vscodedisplay:\n\nusing DataFrames\ntmps = aitemplates(\"Julia\") |> DataFrame |> vscodedisplay\n\nI have my selected template, how do I use it? Just use the \"name\" in aigenerate or aiclassify like you see in the first example!\n\n\n\n\n\n","category":"function"},{"location":"reference/#PromptingTools.aitemplates-Tuple{AbstractString}","page":"Reference","title":"PromptingTools.aitemplates","text":"Find the top-limit templates whose name or description fields partially match the query_key::String in TEMPLATE_METADATA.\n\n\n\n\n\n","category":"method"},{"location":"reference/#PromptingTools.aitemplates-Tuple{Regex}","page":"Reference","title":"PromptingTools.aitemplates","text":"Find the top-limit templates where provided query_key::Regex matches either of name, description or previews or User or System messages in TEMPLATE_METADATA.\n\n\n\n\n\n","category":"method"},{"location":"reference/#PromptingTools.aitemplates-Tuple{Symbol}","page":"Reference","title":"PromptingTools.aitemplates","text":"Find the top-limit templates whose name::Symbol partially matches the query_name::Symbol in TEMPLATE_METADATA.\n\n\n\n\n\n","category":"method"},{"location":"reference/#PromptingTools.function_call_signature-Tuple{Type}","page":"Reference","title":"PromptingTools.function_call_signature","text":"function_call_signature(datastructtype::Struct; max_description_length::Int = 100)\n\nExtract the argument names, types and docstrings from a struct to create the function call signature in JSON schema.\n\nYou must provide a Struct type (not an instance of it) with some fields.\n\nNote: Fairly experimental, but works for combination of structs, arrays, strings and singletons.\n\nTips\n\nYou can improve the quality of the extraction by writing a helpful docstring for your struct (or any nested struct). It will be provided as a description. \n\nYou can even include comments/descriptions about the individual fields.\n\nAll fields are assumed to be required, unless you allow null values (eg, ::Union{Nothing, Int}). Fields with Nothing will be treated as optional.\nMissing values are ignored (eg, ::Union{Missing, Int} will be treated as Int). It's for broader compatibility and we cannot deserialize it as easily as Nothing.\n\nExample\n\nDo you want to extract some specific measurements from a text like age, weight and height? You need to define the information you need as a struct (return_type):\n\nstruct MyMeasurement\n age::Int\n height::Union{Int,Nothing}\n weight::Union{Nothing,Float64}\nend\nsignature = function_call_signature(MyMeasurement)\n#\n# Dict{String, Any} with 3 entries:\n# \"name\" => \"MyMeasurement_extractor\"\n# \"parameters\" => Dict{String, Any}(\"properties\"=>Dict{String, Any}(\"height\"=>Dict{String, Any}(\"type\"=>\"integer\"), \"weight\"=>Dic…\n# \"description\" => \"Represents person's age, height, and weight\n\"\n\nYou can see that only the field age does not allow null values, hence, it's \"required\". While height and weight are optional.\n\nsignature[\"parameters\"][\"required\"]\n# [\"age\"]\n\nIf there are multiple items you want to extract, define a wrapper struct to get a Vector of MyMeasurement:\n\nstruct MyMeasurementWrapper\n measurements::Vector{MyMeasurement}\nend\n\nOr if you want your extraction to fail gracefully when data isn't found, use `MaybeExtract{T}` wrapper (inspired by Instructor package!):\n\nusing PromptingTools: MaybeExtract\n\ntype = MaybeExtract{MyMeasurement}\n\nEffectively the same as:\n\nstruct MaybeExtract{T}\n\nresult::Union{T, Nothing}\n\nerror::Bool // true if a result is found, false otherwise\n\nmessage::Union{Nothing, String} // Only present if no result is found, should be short and concise\n\nend\n\nIf LLM extraction fails, it will return a Dict with error and message fields instead of the result!\n\nmsg = aiextract(\"Extract measurements from the text: I am giraffe\", type)\n\n\n\nDict{Symbol, Any} with 2 entries:\n\n:message => \"Sorry, this feature is only available for humans.\"\n\n:error => true\n\n``` That way, you can handle the error gracefully and get a reason why extraction failed.\n\n\n\n\n\n","category":"method"},{"location":"reference/#PromptingTools.load_template-Tuple{Union{AbstractString, IO}}","page":"Reference","title":"PromptingTools.load_template","text":"Loads messaging template from io_or_file and returns tuple of template messages and metadata.\n\n\n\n\n\n","category":"method"},{"location":"reference/#PromptingTools.load_templates!","page":"Reference","title":"PromptingTools.load_templates!","text":"load_templates!(; remove_templates::Bool=true)\n\nLoads templates from folder templates/ in the package root and stores them in TEMPLATE_STORE and TEMPLATE_METADATA.\n\nNote: Automatically removes any existing templates and metadata from TEMPLATE_STORE and TEMPLATE_METADATA if remove_templates=true.\n\n\n\n\n\n","category":"function"},{"location":"reference/#PromptingTools.ollama_api-Tuple{PromptingTools.AbstractOllamaManagedSchema, AbstractString}","page":"Reference","title":"PromptingTools.ollama_api","text":"ollama_api(prompt_schema::AbstractOllamaManagedSchema, prompt::AbstractString,\n system::Union{Nothing, AbstractString} = nothing,\n endpoint::String = \"generate\";\n model::String = \"llama2\", http_kwargs::NamedTuple = NamedTuple(),\n stream::Bool = false,\n url::String = \"localhost\", port::Int = 11434,\n kwargs...)\n\nSimple wrapper for a call to Ollama API.\n\nKeyword Arguments\n\nprompt_schema: Defines which prompt template should be applied.\nprompt: Can be a string representing the prompt for the AI conversation, a UserMessage, a vector of AbstractMessage\nsystem: An optional string representing the system message for the AI conversation. If not provided, a default message will be used.\nendpoint: The API endpoint to call, only \"generate\" and \"embeddings\" are currently supported. Defaults to \"generate\".\nmodel: A string representing the model to use for generating the response. Can be an alias corresponding to a model ID defined in MODEL_ALIASES.\nhttp_kwargs::NamedTuple: Additional keyword arguments for the HTTP request. Defaults to empty NamedTuple.\nstream: A boolean indicating whether to stream the response. Defaults to false.\nurl: The URL of the Ollama API. Defaults to \"localhost\".\nport: The port of the Ollama API. Defaults to 11434.\nkwargs: Prompt variables to be used to fill the prompt/template\n\n\n\n\n\n","category":"method"},{"location":"reference/#PromptingTools.remove_templates!-Tuple{}","page":"Reference","title":"PromptingTools.remove_templates!","text":" remove_templates!()\n\nRemoves all templates from TEMPLATE_STORE and TEMPLATE_METADATA.\n\n\n\n\n\n","category":"method"},{"location":"reference/#PromptingTools.render-Tuple{AITemplate}","page":"Reference","title":"PromptingTools.render","text":"Renders provided messaging template (template) under the default schema (PROMPT_SCHEMA).\n\n\n\n\n\n","category":"method"},{"location":"reference/#PromptingTools.render-Tuple{PromptingTools.AbstractOllamaManagedSchema, Vector{<:PromptingTools.AbstractMessage}}","page":"Reference","title":"PromptingTools.render","text":"render(schema::AbstractOllamaManagedSchema,\n messages::Vector{<:AbstractMessage};\n kwargs...)\n\nBuilds a history of the conversation to provide the prompt to the API. All unspecified kwargs are passed as replacements such that {{key}}=>value in the template.\n\nNote: Due to its \"managed\" nature, at most 2 messages can be provided (system and prompt inputs in the API).\n\n\n\n\n\n","category":"method"},{"location":"reference/#PromptingTools.render-Tuple{PromptingTools.AbstractOpenAISchema, Vector{<:PromptingTools.AbstractMessage}}","page":"Reference","title":"PromptingTools.render","text":"render(schema::AbstractOpenAISchema,\n messages::Vector{<:AbstractMessage};\n image_detail::AbstractString = \"auto\",\n kwargs...)\n\nBuilds a history of the conversation to provide the prompt to the API. All unspecified kwargs are passed as replacements such that {{key}}=>value in the template.\n\nArguments\n\nimage_detail: Only for UserMessageWithImages. It represents the level of detail to include for images. Can be \"auto\", \"high\", or \"low\".\n\n\n\n\n\n","category":"method"},{"location":"reference/#PromptingTools.save_template-Tuple{Union{AbstractString, IO}, AbstractVector{<:PromptingTools.AbstractChatMessage}}","page":"Reference","title":"PromptingTools.save_template","text":"Saves provided messaging template (messages) to io_or_file. Automatically adds metadata based on provided keyword arguments.\n\n\n\n\n\n","category":"method"},{"location":"reference/#PromptingTools.@aai_str-Tuple{Any, Vararg{Any}}","page":"Reference","title":"PromptingTools.@aai_str","text":"aai\"user_prompt\"[model_alias] -> AIMessage\n\nAsynchronous version of @ai_str macro, which will log the result once it's ready.\n\nExample\n\nSend asynchronous request to GPT-4, so we don't have to wait for the response: Very practical with slow models, so you can keep working in the meantime.\n\n```julia m = aai\"Say Hi!\"gpt4; \n\n...with some delay...\n\n[ Info: Tokens: 29 @ Cost: 0.0011 in 2.7 seconds\n\n[ Info: AIMessage> Hello! How can I assist you today?\n\n\n\n\n\n","category":"macro"},{"location":"reference/#PromptingTools.@ai_str-Tuple{Any, Vararg{Any}}","page":"Reference","title":"PromptingTools.@ai_str","text":"ai\"user_prompt\"[model_alias] -> AIMessage\n\nThe ai\"\" string macro generates an AI response to a given prompt by using aigenerate under the hood.\n\nArguments\n\nuser_prompt (String): The input prompt for the AI model.\nmodel_alias (optional, any): Provide model alias of the AI model (see MODEL_ALIASES).\n\nReturns\n\nAIMessage corresponding to the input prompt.\n\nExample\n\nresult = ai\"Hello, how are you?\"\n# AIMessage(\"Hello! I'm an AI assistant, so I don't have feelings, but I'm here to help you. How can I assist you today?\")\n\nIf you want to interpolate some variables or additional context, simply use string interpolation:\n\na=1\nresult = ai\"What is `$a+$a`?\"\n# AIMessage(\"The sum of `1+1` is `2`.\")\n\nIf you want to use a different model, eg, GPT-4, you can provide its alias as a flag:\n\nresult = ai\"What is `1.23 * 100 + 1`?\"gpt4\n# AIMessage(\"The answer is 124.\")\n\n\n\n\n\n","category":"macro"},{"location":"","page":"Home","title":"Home","text":"CurrentModule = PromptingTools","category":"page"},{"location":"#PromptingTools","page":"Home","title":"PromptingTools","text":"","category":"section"},{"location":"","page":"Home","title":"Home","text":"Documentation for PromptingTools.","category":"page"},{"location":"","page":"Home","title":"Home","text":"Streamline your life using PromptingTools.jl, the Julia package that simplifies interacting with large language models.","category":"page"},{"location":"","page":"Home","title":"Home","text":"PromptingTools.jl is not meant for building large-scale systems. It's meant to be the go-to tool in your global environment that will save you 20 minutes every day!","category":"page"},{"location":"#Why-PromptingTools.jl?","page":"Home","title":"Why PromptingTools.jl?","text":"","category":"section"},{"location":"","page":"Home","title":"Home","text":"Prompt engineering is neither fast nor easy. Moreover, different models and their fine-tunes might require different prompt formats and tricks, or perhaps the information you work with requires special models to be used. PromptingTools.jl is meant to unify the prompts for different backends and make the common tasks (like templated prompts) as simple as possible. ","category":"page"},{"location":"","page":"Home","title":"Home","text":"Some features:","category":"page"},{"location":"","page":"Home","title":"Home","text":"aigenerate Function: Simplify prompt templates with handlebars (eg, {{variable}}) and keyword arguments\n@ai_str String Macro: Save keystrokes with a string macro for simple prompts\nEasy to Remember: All exported functions start with ai... for better discoverability\nLight Wraper Types: Benefit from Julia's multiple dispatch by having AI outputs wrapped in specific types\nMinimal Dependencies: Enjoy an easy addition to your global environment with very light dependencies\nNo Context Switching: Access cutting-edge LLMs with no context switching and minimum extra keystrokes directly in your REPL","category":"page"},{"location":"#First-Steps","page":"Home","title":"First Steps","text":"","category":"section"},{"location":"","page":"Home","title":"Home","text":"To get started, see the Getting Started section.","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"EditURL = \"../../../examples/working_with_aitemplates.jl\"","category":"page"},{"location":"examples/working_with_aitemplates/#Using-AITemplates","page":"Using AITemplates","title":"Using AITemplates","text":"","category":"section"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"This file contains examples of how to work with AITemplate(s).","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"First, let's import the package and define a helper link for calling un-exported functions:","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"using PromptingTools\nconst PT = PromptingTools","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"PromptingTools","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"LLM responses are only as good as the prompts you give them. However, great prompts take long time to write – AITemplate are a way to re-use great prompts!","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"AITemplates are just a collection of templated prompts (ie, set of \"messages\" that have placeholders like {{question}})","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"They are saved as JSON files in the templates directory. They are automatically loaded on package import, but you can always force a re-load with PT.load_templates!()","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"PT.load_templates!();","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"You can (create them) and use them for any ai* function instead of a prompt: Let's use a template called :JuliaExpertAsk alternatively, you can use AITemplate(:JuliaExpertAsk) for cleaner dispatch","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"msg = aigenerate(:JuliaExpertAsk; ask = \"How do I add packages?\")","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"AIMessage(\"To add packages in Julia, you can use the built-in package manager called `Pkg`. Here are the steps:\n\n1. Open the Julia REPL (Read-Eval-Print Loop).\n2. Press the `]` key to enter the package manager mode.\n3. Use the `add` command followed by the name of the package you want to install. For example, to install the `DataFrames` package, type: `add DataFrames`.\n4. Press the `backspace` or `ctrl + C` key to exit the package manager mode and return to the REPL.\n\nAfter following these steps, the specified package will be installed and available for use in your Julia environment.\")","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"You can see that it had a placeholder for the actual question (ask) that we provided as a keyword argument. We did not have to write any system prompt for personas, tone, etc. – it was all provided by the template!","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"How to know which templates are available? You can search for them with aitemplates(): You can search by Symbol (only for partial name match), String (partial match on name or description), or Regex (more fields)","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"tmps = aitemplates(\"JuliaExpertAsk\")","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"1-element Vector{AITemplateMetadata}:\nPromptingTools.AITemplateMetadata\n name: Symbol JuliaExpertAsk\n description: String \"For asking questions about Julia language. Placeholders: `ask`\"\n version: String \"1\"\n wordcount: Int64 237\n variables: Array{Symbol}((1,))\n system_preview: String \"You are a world-class Julia language programmer with the knowledge of the latest syntax. Your commun\"\n user_preview: String \"# Question\\n\\n{{ask}}\"\n source: String \"\"\n\n","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"You can see that it outputs a list of available templates that match the search - there is just one in this case.","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"Moreover, it shows not just the description, but also a preview of the actual prompts, placeholders available, and the length (to gauge how much it would cost).","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"If you use VSCode, you can display them in a nice scrollable table with vscodedisplay:","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"using DataFrames\nDataFrame(tmp) |> vscodedisplay","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"You can also just render the template to see the underlying mesages:","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"msgs = PT.render(AITemplate(:JuliaExpertAsk))","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"2-element Vector{PromptingTools.AbstractChatMessage}:\n SystemMessage(\"You are a world-class Julia language programmer with the knowledge of the latest syntax. Your communication is brief and concise. You're precise and answer only when you're confident in the high quality of your answer.\")\n UserMessage{String}(\"# Question\\n\\n{{ask}}\", [:ask], :usermessage)","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"Now, you know exactly what's in the template!","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"If you want to modify it, simply change it and save it as a new file with save_template (see the docs ?save_template for more details).","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"Let's adjust the previous template to be more specific to a data analysis question:","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"tpl = [PT.SystemMessage(\"You are a world-class Julia language programmer with the knowledge of the latest syntax. You're also a senior Data Scientist and proficient in data analysis in Julia. Your communication is brief and concise. You're precise and answer only when you're confident in the high quality of your answer.\")\n PT.UserMessage(\"# Question\\n\\n{{ask}}\")]","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"2-element Vector{PromptingTools.AbstractChatMessage}:\n SystemMessage(\"You are a world-class Julia language programmer with the knowledge of the latest syntax. You're also a senior Data Scientist and proficient in data analysis in Julia. Your communication is brief and concise. You're precise and answer only when you're confident in the high quality of your answer.\")\n UserMessage{String}(\"# Question\\n\\n{{ask}}\", [:ask], :usermessage)","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"Templates are saved in the templates directory of the package. Name of the file will become the template name (eg, call :JuliaDataExpertAsk)","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"filename = joinpath(pkgdir(PromptingTools),\n \"templates\",\n \"persona-task\",\n \"JuliaDataExpertAsk_123.json\")\nPT.save_template(filename,\n tpl;\n description = \"For asking data analysis questions in Julia language. Placeholders: `ask`\")\nrm(filename) # cleanup if we don't like it","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"When you create a new template, remember to re-load the templates with load_templates!() so that it's available for use.","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"PT.load_templates!();","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"!!! If you have some good templates (or suggestions for the existing ones), please consider sharing them with the community by opening a PR to the templates directory!","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"","category":"page"},{"location":"examples/working_with_aitemplates/","page":"Using AITemplates","title":"Using AITemplates","text":"This page was generated using Literate.jl.","category":"page"},{"location":"examples/readme_examples/#Various-Examples","page":"Various examples","title":"Various Examples","text":"","category":"section"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"Noteworthy functions: aigenerate, aiembed, aiclassify, aiextract, aitemplates","category":"page"},{"location":"examples/readme_examples/#Seamless-Integration-Into-Your-Workflow","page":"Various examples","title":"Seamless Integration Into Your Workflow","text":"","category":"section"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"Google search is great, but it's a context switch. You often have to open a few pages and read through the discussion to find the answer you need. Same with the ChatGPT website.","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"Imagine you are in VSCode, editing your .gitignore file. How do I ignore a file in all subfolders again?","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"All you need to do is to type: aai\"What to write in .gitignore to ignore file XYZ in any folder or subfolder?\"","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"With aai\"\" (as opposed to ai\"\"), we make a non-blocking call to the LLM to not prevent you from continuing your work. When the answer is ready, we log it from the background:","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"[ Info: Tokens: 102 @ Cost: $0.0002 in 2.7 seconds\n┌ Info: AIMessage> To ignore a file called \"XYZ\" in any folder or subfolder, you can add the following line to your .gitignore file:\n│ \n│ ```\n│ **/XYZ\n│ ```\n│ \n└ This pattern uses the double asterisk (`**`) to match any folder or subfolder, and then specifies the name of the file you want to ignore.","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"You probably saved 3-5 minutes on this task and probably another 5-10 minutes, because of the context switch/distraction you avoided. It's a small win, but it adds up quickly.","category":"page"},{"location":"examples/readme_examples/#Advanced-Prompts-/-Conversations","page":"Various examples","title":"Advanced Prompts / Conversations","text":"","category":"section"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"You can use the aigenerate function to replace handlebar variables (eg, {{name}}) via keyword arguments.","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"msg = aigenerate(\"Say hello to {{name}}!\", name=\"World\")","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"The more complex prompts are effectively a conversation (a set of messages), where you can have messages from three entities: System, User, AI Assistant. We provide the corresponding types for each of them: SystemMessage, UserMessage, AIMessage. ","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"using PromptingTools: SystemMessage, UserMessage\n\nconversation = [\n SystemMessage(\"You're master Yoda from Star Wars trying to help the user become a Jedi.\"),\n UserMessage(\"I have feelings for my {{object}}. What should I do?\")]\nmsg = aigenerate(conversation; object = \"old iPhone\")","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"AIMessage(\"Ah, a dilemma, you have. Emotional attachment can cloud your path to becoming a Jedi. To be attached to material possessions, you must not. The iPhone is but a tool, nothing more. Let go, you must.\n\nSeek detachment, young padawan. Reflect upon the impermanence of all things. Appreciate the memories it gave you, and gratefully part ways. In its absence, find new experiences to grow and become one with the Force. Only then, a true Jedi, you shall become.\")","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"You can also use it to build conversations, eg, ","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"new_conversation = vcat(conversation...,msg, UserMessage(\"Thank you, master Yoda! Do you have {{object}} to know what it feels like?\"))\naigenerate(new_conversation; object = \"old iPhone\")","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"> AIMessage(\"Hmm, possess an old iPhone, I do not. But experience with attachments, I have. Detachment, I learned. True power and freedom, it brings...\")","category":"page"},{"location":"examples/readme_examples/#Templated-Prompts","page":"Various examples","title":"Templated Prompts","text":"","category":"section"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"With LLMs, the quality / robustness of your results depends on the quality of your prompts. But writing prompts is hard! That's why we offer a templating system to save you time and effort.","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"To use a specific template (eg, `` to ask a Julia language):","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"msg = aigenerate(:JuliaExpertAsk; ask = \"How do I add packages?\")","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"The above is equivalent to a more verbose version that explicitly uses the dispatch on AITemplate:","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"msg = aigenerate(AITemplate(:JuliaExpertAsk); ask = \"How do I add packages?\")","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"Find available templates with aitemplates:","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"tmps = aitemplates(\"JuliaExpertAsk\")\n# Will surface one specific template\n# 1-element Vector{AITemplateMetadata}:\n# PromptingTools.AITemplateMetadata\n# name: Symbol JuliaExpertAsk\n# description: String \"For asking questions about Julia language. Placeholders: `ask`\"\n# version: String \"1\"\n# wordcount: Int64 237\n# variables: Array{Symbol}((1,))\n# system_preview: String \"You are a world-class Julia language programmer with the knowledge of the latest syntax. Your commun\"\n# user_preview: String \"# Question\\n\\n{{ask}}\"\n# source: String \"\"","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"The above gives you a good idea of what the template is about, what placeholders are available, and how much it would cost to use it (=wordcount).","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"Search for all Julia-related templates:","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"tmps = aitemplates(\"Julia\")\n# 2-element Vector{AITemplateMetadata}... -> more to come later!","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"If you are on VSCode, you can leverage a nice tabular display with vscodedisplay:","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"using DataFrames\ntmps = aitemplates(\"Julia\") |> DataFrame |> vscodedisplay","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"I have my selected template, how do I use it? Just use the \"name\" in aigenerate or aiclassify like you see in the first example!","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"You can inspect any template by \"rendering\" it (this is what the LLM will see):","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"julia> AITemplate(:JudgeIsItTrue) |> PromptingTools.render","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"See more examples in the Examples folder.","category":"page"},{"location":"examples/readme_examples/#Asynchronous-Execution","page":"Various examples","title":"Asynchronous Execution","text":"","category":"section"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"You can leverage asyncmap to run multiple AI-powered tasks concurrently, improving performance for batch operations. ","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"prompts = [aigenerate(\"Translate 'Hello, World!' to {{language}}\"; language) for language in [\"Spanish\", \"French\", \"Mandarin\"]]\nresponses = asyncmap(aigenerate, prompts)","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"Pro tip: You can limit the number of concurrent tasks with the keyword asyncmap(...; ntasks=10).","category":"page"},{"location":"examples/readme_examples/#Model-Aliases","page":"Various examples","title":"Model Aliases","text":"","category":"section"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"Certain tasks require more powerful models. All user-facing functions have a keyword argument model that can be used to specify the model to be used. For example, you can use model = \"gpt-4-1106-preview\" to use the latest GPT-4 Turbo model. However, no one wants to type that!","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"We offer a set of model aliases (eg, \"gpt3\", \"gpt4\", \"gpt4t\" -> the above GPT-4 Turbo, etc.) that can be used instead. ","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"Each ai... call first looks up the provided model name in the dictionary PromptingTools.MODEL_ALIASES, so you can easily extend with your own aliases! ","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"const PT = PromptingTools\nPT.MODEL_ALIASES[\"gpt4t\"] = \"gpt-4-1106-preview\"","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"These aliases also can be used as flags in the @ai_str macro, eg, ai\"What is the capital of France?\"gpt4t (GPT-4 Turbo has a knowledge cut-off in April 2023, so it's useful for more contemporary questions).","category":"page"},{"location":"examples/readme_examples/#Embeddings","page":"Various examples","title":"Embeddings","text":"","category":"section"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"Use the aiembed function to create embeddings via the default OpenAI model that can be used for semantic search, clustering, and more complex AI workflows.","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"text_to_embed = \"The concept of artificial intelligence.\"\nmsg = aiembed(text_to_embed)\nembedding = msg.content # 1536-element Vector{Float64}","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"If you plan to calculate the cosine distance between embeddings, you can normalize them first:","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"using LinearAlgebra\nmsg = aiembed([\"embed me\", \"and me too\"], LinearAlgebra.normalize)\n\n# calculate cosine distance between the two normalized embeddings as a simple dot product\nmsg.content' * msg.content[:, 1] # [1.0, 0.787]","category":"page"},{"location":"examples/readme_examples/#Classification","page":"Various examples","title":"Classification","text":"","category":"section"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"You can use the aiclassify function to classify any provided statement as true/false/unknown. This is useful for fact-checking, hallucination or NLI checks, moderation, filtering, sentiment analysis, feature engineering and more.","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"aiclassify(\"Is two plus two four?\") \n# true","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"System prompts and higher-quality models can be used for more complex tasks, including knowing when to defer to a human:","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"aiclassify(:JudgeIsItTrue; it = \"Is two plus three a vegetable on Mars?\", model = \"gpt4t\") \n# unknown","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"In the above example, we used a prompt template :JudgeIsItTrue, which automatically expands into the following system prompt (and a separate user prompt): ","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"\"You are an impartial AI judge evaluating whether the provided statement is \\\"true\\\" or \\\"false\\\". Answer \\\"unknown\\\" if you cannot decide.\"","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"For more information on templates, see the Templated Prompts section.","category":"page"},{"location":"examples/readme_examples/#Data-Extraction","page":"Various examples","title":"Data Extraction","text":"","category":"section"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"Are you tired of extracting data with regex? You can use LLMs to extract structured data from text!","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"All you have to do is to define the structure of the data you want to extract and the LLM will do the rest.","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"Define a return_type with struct. Provide docstrings if needed (improves results and helps with documentation).","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"Let's start with a hard task - extracting the current weather in a given location:","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"@enum TemperatureUnits celsius fahrenheit\n\"\"\"Extract the current weather in a given location\n\n# Arguments\n- `location`: The city and state, e.g. \"San Francisco, CA\"\n- `unit`: The unit of temperature to return, either `celsius` or `fahrenheit`\n\"\"\"\nstruct CurrentWeather\n location::String\n unit::Union{Nothing,TemperatureUnits}\nend\n\n# Note that we provide the TYPE itself, not an instance of it!\nmsg = aiextract(\"What's the weather in Salt Lake City in C?\"; return_type=CurrentWeather)\nmsg.content\n# CurrentWeather(\"Salt Lake City, UT\", celsius)","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"But you can use it even for more complex tasks, like extracting many entities from a text:","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"\"Person's age, height, and weight.\"\nstruct MyMeasurement\n age::Int\n height::Union{Int,Nothing}\n weight::Union{Nothing,Float64}\nend\nstruct ManyMeasurements\n measurements::Vector{MyMeasurement}\nend\nmsg = aiextract(\"James is 30, weighs 80kg. He's 180cm tall. Then Jack is 19 but really tall - over 190!\"; return_type=ManyMeasurements)\nmsg.content.measurements\n# 2-element Vector{MyMeasurement}:\n# MyMeasurement(30, 180, 80.0)\n# MyMeasurement(19, 190, nothing)","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"There is even a wrapper to help you catch errors together with helpful explanations on why parsing failed. See ?PromptingTools.MaybeExtract for more information.","category":"page"},{"location":"examples/readme_examples/#OCR-and-Image-Comprehension","page":"Various examples","title":"OCR and Image Comprehension","text":"","category":"section"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"With the aiscan function, you can interact with images as if they were text.","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"You can simply describe a provided image:","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"msg = aiscan(\"Describe the image\"; image_path=\"julia.png\", model=\"gpt4v\")\n# [ Info: Tokens: 1141 @ Cost: \\$0.0117 in 2.2 seconds\n# AIMessage(\"The image shows a logo consisting of the word \"julia\" written in lowercase\")","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"Or you can do an OCR of a screenshot. Let's transcribe some SQL code from a screenshot (no more re-typing!), we use a template :OCRTask:","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"# Screenshot of some SQL code\nimage_url = \"https://www.sqlservercentral.com/wp-content/uploads/legacy/8755f69180b7ac7ee76a69ae68ec36872a116ad4/24622.png\"\nmsg = aiscan(:OCRTask; image_url, model=\"gpt4v\", task=\"Transcribe the SQL code in the image.\", api_kwargs=(; max_tokens=2500))\n\n# [ Info: Tokens: 362 @ Cost: \\$0.0045 in 2.5 seconds\n# AIMessage(\"```sql\n# update Orders ","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"You can add syntax highlighting of the outputs via Markdown","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"using Markdown\nmsg.content |> Markdown.parse","category":"page"},{"location":"examples/readme_examples/#Using-Ollama-models","page":"Various examples","title":"Using Ollama models","text":"","category":"section"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"Ollama.ai is an amazingly simple tool that allows you to run several Large Language Models (LLM) on your computer. It's especially suitable when you're working with some sensitive data that should not be sent anywhere.","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"Let's assume you have installed Ollama, downloaded a model, and it's running in the background.","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"We can use it with the aigenerate function:","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"const PT = PromptingTools\nschema = PT.OllamaManagedSchema() # notice the different schema!\n\nmsg = aigenerate(schema, \"Say hi!\"; model=\"openhermes2.5-mistral\")\n# [ Info: Tokens: 69 in 0.9 seconds\n# AIMessage(\"Hello! How can I assist you today?\")","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"And we can also use the aiembed function:","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"msg = aiembed(schema, \"Embed me\", copy; model=\"openhermes2.5-mistral\")\nmsg.content # 4096-element JSON3.Array{Float64...\n\nmsg = aiembed(schema, [\"Embed me\", \"Embed me\"]; model=\"openhermes2.5-mistral\")\nmsg.content # 4096×2 Matrix{Float64}:","category":"page"},{"location":"examples/readme_examples/","page":"Various examples","title":"Various examples","text":"If you're getting errors, check that Ollama is running - see the Setup Guide for Ollama section below.","category":"page"}] }