Skip to content

Commit

Permalink
Merge pull request #233 from lifeomic/FLDM-474/agents-client
Browse files Browse the repository at this point in the history 
Add Agents client for invoking basic agent
  • Loading branch information
@epeters3
epeters3 authored Nov 18, 2024
2 parents 0a9cd36 + 4a52da4 commit 306d697
Show file tree
Hide file tree
Showing 3 changed files with 131 additions and 54 deletions.
166 changes: 113 additions & 53 deletions CHANGELOG.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -3,10 +3,18 @@
All notable changes to this project will be documented in this file.

The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
and this project adheres to
[Semantic Versioning](https://semver.org/spec/v2.0.0.html).

_(NOTE: All examples use fictitious data or freely available data sets.)_

## [0.36.0] - 2024-11-15

### Added

- Added a new `phc.services.Agents` API client with an `invoke_basic` method for
invoking a basic multi-modal LLM.

## [0.35.0] - 2024-05-29

### Added
Expand All @@ -23,7 +31,8 @@ _(NOTE: All examples use fictitious data or freely available data sets.)_

### Added

- A New function `es_sql()` to `phc.services.Fhir` class to execute OpenSearch SQL.
- A New function `es_sql()` to `phc.services.Fhir` class to execute OpenSearch
SQL.

Example:

Expand All @@ -42,7 +51,8 @@ df = Fhir(session).es_sql(

### Fixed

- The `phc.services.PatientML` API client will now ignore extra values, instead of throwing an error if one is found when parsing an entity.
- The `phc.services.PatientML` API client will now ignore extra values, instead
of throwing an error if one is found when parsing an entity.

## [0.33.0] - 2023-07-28

Expand All @@ -58,21 +68,25 @@ df = Fhir(session).es_sql(

### Added

- The `phc.services.PatientML` API client is now fully typed. Request bodies, response bodies, and query parameters now allow for static type checking, and are runtime type-checked as well.
- The `phc.services.PatientML` API client is now fully typed. Request bodies,
response bodies, and query parameters now allow for static type checking, and
are runtime type-checked as well.

## [0.31.0] - 2023-07-18

### Fixed

- Pinned the major versions of the package's dependencies, so untested breaking changes are excluded from the package's installation
- Pinned the major versions of the package's dependencies, so untested breaking
changes are excluded from the package's installation
- Removed dev dependencies that were already specified as package dependencies
- Pinned the exact version of the dev dependencies, to improve CI/CD determinism

## [0.29.0] - 2023-06-22

### Added

- Added new `phc.services.PatientML` API client which can be used to call the LifeOmic Patient ML Service API.
- Added new `phc.services.PatientML` API client which can be used to call the
LifeOmic Patient ML Service API.

## [0.23.2] - 2021-09-27

Expand All @@ -83,7 +97,8 @@ df = Fhir(session).es_sql(

### Added

- Added support for single/multiple values with the `term` or `terms` (New!) parameters
- Added support for single/multiple values with the `term` or `terms` (New!)
parameters
- Added configurable term limit for sending multiple queries (param `max_terms`)
- Added `phc.DataLake` for sending SQL queries to the data lake

Expand All @@ -108,16 +123,21 @@ phc.DataLake.get_data_frame(

### Fixed

- `phc.Project` now operates properly when one of the accounts only has limited access
- `phc.Observation`, `phc.Condition`, and `phc.Procedures`'s `get_codes` method now uses the new summary APIs to return better results.
- `phc.Project` now operates properly when one of the accounts only has limited
access
- `phc.Observation`, `phc.Condition`, and `phc.Procedures`'s `get_codes` method
now uses the new summary APIs to return better results.

### Added

- Summary APIs ([PR #150](https://github.com/lifeomic/phc-sdk-py/pull/150))
- `phc.SummaryClinicalCounts` - Retrieve all clinical counts (across tables like observation, conditions, procedures, and medications)
- `phc.SummaryOmicsCounts` - Retrieve summaries across genomic data (counts of clinvar_significance, gene_variant, sequence, test, etc)
- `phc.SummaryClinicalCounts` - Retrieve all clinical counts (across tables
like observation, conditions, procedures, and medications)
- `phc.SummaryOmicsCounts` - Retrieve summaries across genomic data (counts of
clinvar_significance, gene_variant, sequence, test, etc)
- `phc.SummaryCounts` - Retrieve all summaries (across omics and clinical)
- `phc.SummaryItemCounts` - Retrieve counts for a specific table (e.g. `condition`, `procedure`)
- `phc.SummaryItemCounts` - Retrieve counts for a specific table (e.g.
`condition`, `procedure`)

Example

Expand All @@ -137,13 +157,15 @@ phc.SummaryClinicalCounts.get_data_frame(match="fuzzy", system=["snomed.info", "

### Fixed

- Make token optional when using a custom adapter that doesn't support refreshing the token
- Make token optional when using a custom adapter that doesn't support
refreshing the token

## [0.22.1] - 2021-04-27

### Added

- Added ability to use custom adapter for sending/receiving underlying data (e.g. for tests)
- Added ability to use custom adapter for sending/receiving underlying data
(e.g. for tests)

## [0.22.0] - 2021-03-24

Expand All @@ -153,16 +175,22 @@ phc.SummaryClinicalCounts.get_data_frame(match="fuzzy", system=["snomed.info", "

### Added

Added lots of Ocr functionality and a Composition module in the `easy` namespace.
Added lots of Ocr functionality and a Composition module in the `easy`
namespace.

- `phc.Ocr.Config` - Create and update PrecisionOCR config within a project
- `phc.Ocr.Document` - Retrieve PrecisionOCR documents
- `phc.Ocr.DocumentComposition` - Retrieve metadata by page for PrecisionOCR documents
- `phc.Ocr.Block` - Retrieve the text and layout metadata from a PrecisionOCR document
- `phc.Ocr.Suggestion` - Retrieve all permutations of PrecisionOCR medical suggestions
- `phc.Composition` - Base FHIR class for retrieving Composition resources from the FHIR Search Service

Added the ability to create, read, update, and delete using the FHIR DSTU3 API by appending `.DSTU3` to any easy module that supports it.
- `phc.Ocr.DocumentComposition` - Retrieve metadata by page for PrecisionOCR
documents
- `phc.Ocr.Block` - Retrieve the text and layout metadata from a PrecisionOCR
document
- `phc.Ocr.Suggestion` - Retrieve all permutations of PrecisionOCR medical
suggestions
- `phc.Composition` - Base FHIR class for retrieving Composition resources from
the FHIR Search Service

Added the ability to create, read, update, and delete using the FHIR DSTU3 API
by appending `.DSTU3` to any easy module that supports it.

```python
phc.Patient.DSTU3.create(...)
Expand All @@ -189,14 +217,16 @@ phc.Patient.DSTU3.delete(...)
- `tools.get` - Gets the default verson or a specific version of a tool
- `tools.add_version` - Adds a verson to an existing tool
- `tools.delete` - Deletes the tool or a specific version of a tool
- `tools.get_list` - Returns tools from the registry and allows for optional filters
- `tools.get_list` - Returns tools from the registry and allows for optional
filters

- `Workflows` - A service to manager workflows

- `workflows.run` - Runs a workflow using a provided tool from the registry
- `workflows.get` - Gets a workflow run
- `workflows.get_list` - Returns all workflows for a project
- `workflows.describe` - Returns a list of the inputs and types the workflow requires to run a tool
- `workflows.describe` - Returns a list of the inputs and types the workflow
requires to run a tool

- Added filtering by id for all `phc.easy` modules

Expand All @@ -212,8 +242,10 @@ phc.Observation.get_data_frame(ids=["<value1>", "<value2>"])

### Fixed

- Genomics modules (`phc.easy`) now handle an out of range date via a warning (and auto-conversion to `NaT`)
- Setting and retrieving projects now works properly again (Previously, projects were inaccurate or sometimes missing.)
- Genomics modules (`phc.easy`) now handle an out of range date via a warning
(and auto-conversion to `NaT`)
- Setting and retrieving projects now works properly again (Previously, projects
were inaccurate or sometimes missing.)

## [0.20.0] - 2020-11-19

Expand Down Expand Up @@ -347,7 +379,9 @@ phc.Condition.get_data_frame(code=["25910003", "30156004"], system="http://snome

### Changed

- Overhauled `get_codes` to make results more accurate and allow searching by display (See [#93](https://github.com/lifeomic/phc-sdk-py/pull/93) for full discussion)
- Overhauled `get_codes` to make results more accurate and allow searching by
display (See [#93](https://github.com/lifeomic/phc-sdk-py/pull/93) for full
discussion)

```python
# 1. Get display values and number of records they occur in
Expand Down Expand Up @@ -395,23 +429,28 @@ phc.Observation.get_codes("status", sample_size=10)

### Added

- Paging requests with `all_results=True` now automatically retries to the server with an exponentially smaller batch size on error (`pow(limit, 0.85)`). We can't tell what the error is, but we can retry with a smaller page size.
- Paging requests with `all_results=True` now automatically retries to the
server with an exponentially smaller batch size on error (`pow(limit, 0.85)`).
We can't tell what the error is, but we can retry with a smaller page size.
- Added `page_size` to the easy modules for a custom batch size
- Added `max_pages` to the easy modules for capping the number of pages returned
- Added pretty print to FHIR Search Service queries when passing `log=True`
- Warn and convert out of range date times (e.g. `0217-01-01`) to `NaT`

### Fixed

- Properly parse date columns with positive time zones into the local time and time zone
- Resolved a `KeyError` issue with `coding` where the `valueCodeableConcept` didn't have a system or url
- Passing `patient_id` / `patient_ids` with a `must` FHIR Search Service query now works as expected
- Properly parse date columns with positive time zones into the local time and
time zone
- Resolved a `KeyError` issue with `coding` where the `valueCodeableConcept`
didn't have a system or url
- Passing `patient_id` / `patient_ids` with a `must` FHIR Search Service query
now works as expected

### Changed

[BREAKING] The expanded columns have changed to more reflect the location of
the value. All systems and URLs are separated by `__` and prefixed with either
`url` or `system`. Here is an example:
[BREAKING] The expanded columns have changed to more reflect the location of the
value. All systems and URLs are separated by `__` and prefixed with either `url`
or `system`. Here is an example:

```python
input_dict = [
Expand Down Expand Up @@ -478,11 +517,14 @@ assert generic_codeable_to_dict(input_dict) == {
- ReferralRequest
- Sequence
- Specimen
- Add abstract `Item` class for entities that don't relate to a patient (e.g. Organization and Practitioner)
- Add abstract `Item` class for entities that don't relate to a patient (e.g.
Organization and Practitioner)

### Changed

All date columns now return two columns--one for the local time (with time zone removed) and one for the time zone offset in hours. Consider the `onsetDateTime` column from BRCA's `Condition` table:
All date columns now return two columns--one for the local time (with time zone
removed) and one for the time zone offset in hours. Consider the `onsetDateTime`
column from BRCA's `Condition` table:

```
onsetDateTime.tz onsetDateTime.local
Expand All @@ -506,8 +548,10 @@ Includes more work on the easy modules (imported via `import phc.easy as phc`).
- Added generic methods on `phc.easy.Query`
- `get_count_by_field`
- `get_codes`
- `execute_composite_aggregations` (used by `get_count_by_field` and `get_codes`)
- Added `phc.easy.PatientItem.get_count_by_patient` (Observation, Procedure, Specimen, etc.)
- `execute_composite_aggregations` (used by `get_count_by_field` and
`get_codes`)
- Added `phc.easy.PatientItem.get_count_by_patient` (Observation, Procedure,
Specimen, etc.)

```python
# Example: Get number of procedures by patient
Expand All @@ -520,7 +564,8 @@ phc.Procedure.get_count_by_patient()
# b41f8107-85e1-42c3-b36e-400085799ab5 176
```

- Added `phc.easy.PatientItem.get_count_by_field` (Observation, Procedure, Specimen, etc.)
- Added `phc.easy.PatientItem.get_count_by_field` (Observation, Procedure,
Specimen, etc.)

```python
# Example: Get count of unique procedure display codes
Expand All @@ -533,7 +578,8 @@ phc.Procedure.get_count_by_field("code.coding.display")
# 3 simple mastectomy 200
```

- Added `phc.easy.PatientItem.get_codes` (Observation, Procedure, Specimen, etc.)
- Added `phc.easy.PatientItem.get_codes` (Observation, Procedure, Specimen,
etc.)

```python
# Example: Get observation codes for specific patients
Expand All @@ -554,9 +600,12 @@ phc.Observation.get_codes(patient_ids=[
### Changed

- Passing `log` to any PatientItem entities now logs the FSS query being run
- For aggregations, `phc.Query.execute_fhir_dsl` now returns a `FhirAggregation` if an aggregation is specified in the query
- `phc.Query.execute_fhir_dsl_with_options` now caches aggregation queries in JSON format
- Specifying `patient_id` and/or `patient_ids` is now properly supported with a custom FHIR query.
- For aggregations, `phc.Query.execute_fhir_dsl` now returns a `FhirAggregation`
if an aggregation is specified in the query
- `phc.Query.execute_fhir_dsl_with_options` now caches aggregation queries in
JSON format
- Specifying `patient_id` and/or `patient_ids` is now properly supported with a
custom FHIR query.

```python
# Example: Get observations tagged with loinc for a specific patient
Expand Down Expand Up @@ -592,7 +641,8 @@ phc.Observation.get_data_frame(patient_id="<id>", query_overrides={
- Add `Query` for scrolling through FHIR Search Service (FSS) data
- Add `Frame` for expanding columns that contain FHIR data and parsing dates
- Add `APICache` for auto-caching results from easy modules
- Add `CSVWriter` for intelligently writing batches O(1) without having memory grow
- Add `CSVWriter` for intelligently writing batches O(1) without having memory
grow
- Includes `Project`, `Patient`, `Observation`, `Procedure`, `Condition`,
`Goal`, and `Specimen`

Expand Down Expand Up @@ -634,7 +684,8 @@ phc.Observation.get_data_frame(patient_id="<id>", query_overrides={

### Added

- Added the `trust_env` parameter to all service classes to enable http proxy support.
- Added the `trust_env` parameter to all service classes to enable http proxy
support.

## [0.10.0] - 2020-03-10

Expand Down Expand Up @@ -664,37 +715,46 @@ phc.Observation.get_data_frame(patient_id="<id>", query_overrides={

### Changed

- In `Analytics.load_data_lake_result_to_dataframe` increased the amount of time it takes to wait for a results file.
- In `Analytics.load_data_lake_result_to_dataframe` increased the amount of time
it takes to wait for a results file.

## [0.8.0] - 2019-11-25

### Added

- Added `Analytics.list_data_lake_schemas` to fetch the schemas of each data lake table.
- Added `Analytics.get_data_lake_schema` to fetch the schema of a single data lake table.
- Added `Analytics.execute_data_lake_query_to_dataframe` to execute a data lake query and load the results to a Pandas dataframe.
- Added `Analytics.load_data_lake_result_to_dataframe` to load the results of a previously executed data lake query to a Pandas dataframe.
- Added `Analytics.list_data_lake_schemas` to fetch the schemas of each data
lake table.
- Added `Analytics.get_data_lake_schema` to fetch the schema of a single data
lake table.
- Added `Analytics.execute_data_lake_query_to_dataframe` to execute a data lake
query and load the results to a Pandas dataframe.
- Added `Analytics.load_data_lake_result_to_dataframe` to load the results of a
previously executed data lake query to a Pandas dataframe.
- Added `Files.exists` to check if a file exists.

## [0.7.1] - 2019-11-21

### Fixed

- Fixed issue with `Files.download` to create target directories if they do not exist.
- Fixed issue with `Files.download` to create target directories if they do not
exist.

## [0.7.0] - 2019-11-20

### Added

- Added optional `pandas` setup install
- Added `ApiResponse.get_as_dataframe` to return a response item as a Pandas DataFrame.
- Added `ApiResponse.get_as_dataframe` to return a response item as a Pandas
DataFrame.

## [0.6.0] - 2019-11-01

### Added

- Added the `phc.services.Files` submodule that provides actions for files in PHC projects.
- Added the `phc.services.Cohorts` submodule that provides actions for files in PHC cohorts.
- Added the `phc.services.Files` submodule that provides actions for files in
PHC projects.
- Added the `phc.services.Cohorts` submodule that provides actions for files in
PHC cohorts.

[0.23.2]: https://github.com/lifeomic/phc-sdk-py/compare/v0.23.1...v0.23.2
[0.23.1]: https://github.com/lifeomic/phc-sdk-py/compare/v0.23.0...v0.23.1
Expand Down
17 changes: 17 additions & 0 deletions phc/services/agents.py
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,17 @@
from typing import Dict, List, Union
from phc.base_client import BaseClient


class Agents(BaseClient):
"""
Provides access to the PHC agents API, which allows you to call LLM-based agents.
"""

def invoke_basic(self, body: Union[str, List[Dict]]):
"""
Invokes a basic agent, which supports either a basic prompt string or a list of
messages which can include images as data urls.
"""
return self._api_call(
"/v1/agents/basic/invoke", json={"input": body}, http_verb="POST"
)
Loading

0 comments on commit 306d697

Please sign in to comment.