- Introduction
- Features
- Requirements
- Installation
- Getting Started
- Understanding how Enola-AI works
- 6.1. Track Activity
- 6.2. Steps in Enola-AI
- 6.3. Feedback Evaluation
- 6.4. Extracting Information
- Documentation: Sending Data to Enola-AI
- Summary
- Contributing
- License
- Contact
Enola AI is an advanced GenAI platform designed to validate and monitor the robustness of artificial intelligence models in highly regulated industries such as finance, healthcare, and education. Our solution ensures that AI implementations comply with strict regulatory standards through continuous assessments, seamless integrations, and real-time monitoring.
This documentation provides a comprehensive guide on how to use the Enola-AI Python library to interact with the platform. You will learn how to install the library, understand the types of steps used in tracking, send different types of data, and utilize various features to enhance your AI model's performance and compliance.
- Multilevel Evaluation: Collect feedback from users, automated assessments, and reviews from internal experts.
- Real-Time Monitoring: Continuous monitoring capabilities to detect deviations in AI model behavior.
- Seamless Integration: Compatible with existing infrastructures such as ERP systems, CRM platforms, and data analytics tools.
- Customizable Configuration: Adapt the evaluation methodology according to the specific needs of the client.
- Security and Compliance: Advanced security measures and compliance with data protection regulations.
- Python 3.7+
- Enola API Token
Before installing the Enola-AI Python library, ensure that you have Python 3.7 or higher installed.
Install the SDK via pip
In your Command Prompt (Windows) or Terminal (Linux/macOS) type:
pip install enolaBy doing a pip install, you have the Enola-AI Python library and its dependencies automatically installed on your device.
Note: Advanced users may prefer to use a Virtual Environment to manage dependencies and isolate the Enola-AI library from other Python packages.
To start using the Enola-AI Python library, follow the steps below to initialize tracking in your application.
To connect to Enola and initialize tracking, you will need:
- A token provided by Enola-AI. This token is essential for authentication and authorization purposes.
- A Python script. You can start by creating an empty Python file with a
.pyextension (e.g.,enola_sample.py,enola_script.py).
Step 1: Load the Enola API Token
You can load the token from a .env file (recommended):
This method is recommended due to better security for token management.
-
Go to the same directory where your Python file is located.
-
Create a file named
.envin the same directory as your Python script. -
Open the file with a text editor and add this line:
ENOLA_TOKEN=your_api_token -
Replace
your_api_tokenwith your Enola API token. -
Ensure that your Python file and the
.envfile are in the same directory so that the token can be loaded.
Alternatively, you can set it directly in your script:
This method is easier and fine for testing purposes, but it is not recommended because the token is exposed in the script.
token = 'your_api_token'You can also load the token using Environment Variables:
Another method is by setting your Enola API token as an environment variable. This method (recommended for advanced users) offers more flexibility but requires more complex configurations, depending on your operating system.
Step 2: Import the Necessary Libraries
Assuming you are loading the token from the .env file, in your Python script, start by importing the necessary libraries:
# Import necessary libraries
from enola.tracking import Tracking # Enola Tracking module
from dotenv import load_dotenv # .env Loader
import os # For environment variable accessStep 3: Define User Input
# Define the user input message
user_input = "Hello, what can you do?" # Input message from the userStep 4: Load the Enola API Token
# Load .env file
load_dotenv()
# Set up your token
token = os.getenv('ENOLA_TOKEN')Step 5: Initialize the Tracking Agent
# Initialize the tracking agent
monitor = Tracking(
token=token, # Your Enola API token
name="My Enola Project", # Name of your tracking session
is_test=True, # Set to True if this is a test session
app_id="my_app_id_01", # Application ID
user_id="user_123", # User ID
session_id="session_456", # Session ID
channel_id="console", # Channel ID (e.g., 'console', 'web', 'mobile')
ip="192.168.1.1", # IP address of the client
message_input=user_input # Input message from the user
)Step 6: Create a New Step
# Create a step
step_chat = monitor.new_step("User LLM Question")Step 7: Add Extra Information
Add any additional information relevant to the step, such as the user's question.
# Add user's question
step_chat.add_extra_info("UserQuestion", user_input)Step 8: Process the User Input with the Language Model
Simulate the model generating a response to the user's question.
# Simulated model response
model_response = "I'm here to assist you in finding the help or information you need."Note: You can replace the simulated response with an actual model response (e.g., GPT-4, Ollama, BERT). You can check our user guide to build a chatbot using Ollama by visiting our section Building an Ollama Chatbot.
Step 9: Add the Model's Response to the Step
# Add model's response
step_chat.add_extra_info("ModelResponse", model_response)Step 10: Close the LLM Step
Indicate that the step has completed successfully and include token usage and costs.
# Close the LLM Step with close_step_token
monitor.close_step_token(
step=step_chat,
successfull=True,
message_output=model_response,
token_input_num=12, # Number of input tokens (estimated)
token_output_num=15, # Number of output tokens (estimated)
token_total_cost=0.0025, # Total cost (example)
token_input_cost=0.001, # Cost for input tokens
token_output_cost=0.0015 # Cost for output tokens
)Step 11: Execute the Tracking
Send the tracking data to the Enola-AI server.
# Execute the tracking and send the data to Enola-AI server
monitor.execute(
successfull=True,
message_output=model_response,
num_iteratons=1
)Here's the complete code incorporating all the steps:
# Import necessary libraries
from enola.tracking import Tracking
from dotenv import load_dotenv
import os
# Define the user input message
user_input = "Hello, what can you do?" # Input message from the user
# Load .env file and set up your token
load_dotenv()
token = os.getenv('ENOLA_TOKEN')
# Initialize the tracking agent
monitor = Tracking(
token=token,
name="My Enola Project", # Name of your tracking session
is_test=True, # Set to True if this is a test session
app_id="my_app_id_01", # Application ID
user_id="user_123", # User ID
session_id="session_456", # Session ID
channel_id="console", # Channel ID (e.g., 'console', 'web', 'mobile')
ip="192.168.1.1", # IP address of the client
message_input=user_input # Input message from the user
)
# Create a step
step_chat = monitor.new_step("User LLM Question")
# Add user's question
step_chat.add_extra_info("UserQuestion", user_input)
# Simulated model response
model_response = "I'm here to assist you in finding the help or information you need."
# Add model's response
step_chat.add_extra_info("ModelResponse", model_response)
# Close the LLM Step
monitor.close_step_token(
step=step_chat,
successfull=True,
message_output=model_response,
token_input_num=12,
token_output_num=15,
token_total_cost=0.0025,
token_input_cost=0.001,
token_output_cost=0.0015
)
# Execute the tracking and send the data to Enola-AI server
monitor.execute(
successfull=True,
message_output=model_response,
num_iteratons=1
)After initializing the tracking agent and executing it, you should get a console output like this:
2024-10-30 09:43:29,909 WELCOME to Enola...
2024-10-30 09:43:29,909 authorized...
2024-10-30 09:43:29,909 STARTED!!!
My Enola Project: sending to server...
My Enola Project: finish OK!
This means you have successfully connected to Enola-AI and sent the data to the servers.
To better understand how Enola-AI works and what has to offer, it is important that you understand its different features. This section is going to be divided in 4 subcategories:
When using Enola-AI, you can add Track Activity, allowing you to track any interactions from the system and users. By tracking the interactions in your system with Enola-AI, you can effectively monitor, validate and evaluate your models.
In Enola-AI, the concept of steps is fundamental for tracking the execution flow of your AI agents. Each step represents a significant action or event in your agent's processing pipeline.
There are two main types of steps:
-
Generic Steps: Used for general-purpose tracking of actions that are not specific to language models, such as data retrieval, preprocessing, or any custom logic.
-
LLM Steps: Specifically designed for tracking interactions with Language Models (e.g., GPT-4, Ollama, BERT), where token usage and costs are relevant.
Understanding the difference between these step types is crucial for accurate tracking and cost analysis.
Enola-AI provides a feedback system that allows users to evaluate AI agent executions. Feedback can be submitted either through the Enola-AI platform or programmatically via code. This helps in assessing the performance of your AI agents and gathering user insights for improvement.
Enola-AI allows you to retrieve the Input and Output data from previous executions for analysis or further manual and automatic processing.
In this section, you will find complete documentation about sending data to the Enola-AI servers, this guide includes step-by-step instructions and code examples, along with explanations of the system's functionalities.
For the complete documentation, you can visit our guide Sending Data to Enola-AI covering the following sections:
- This documentation provides a basic guide on using the Enola-AI Python library to initialize tracking and send data.
- For detailed documentation about sending data with Enola-AI, you can visit our Sending Data to Enola-AI guide.
- For a comprehensive guide on how to build a Chatbot using Ollama with Enola-AI Tracking, you can visit our section Building an Ollama Chatbot.
- If you have questions, you are invited to visit our Frequently Asked Questions section.
- If you want to check all the complete code examples, you can visit our Complete Code Examples section.
Contributions are welcome! Please open an issue or submit a pull request for any improvements or corrections.
When contributing, please ensure to:
- Follow the existing coding style.
- Write clear commit messages.
- Update documentation as necessary.
- Ensure that any code changes are covered by tests.
This project is licensed under the BSD 3-Clause License.
For any inquiries or support, please contact us at [email protected].