Add claude support for llm alerts

[andrewm4894]

This pull request introduces several significant changes to the anomaly detection system, including the addition of support for Anthropic's LLM, refactoring existing OpenAI-related code, and updating environment variables. The most important changes are summarized below:

Support for Anthropic LLM:

• Added new environment variables for Anthropic's LLM and updated the .example.env file to include these variables. (.example.env)
• Implemented a new module anthropic.py to handle anomaly detection using Anthropic's Messages API. (anomstack/llm/anthropic.py)

Refactoring and Code Improvements:

• Refactored the detect_anomalies function to dispatch to the appropriate LLM platform (OpenAI or Anthropic) based on the environment variable ANOMSTACK_LLM_PLATFORM. (anomstack/llm/detect.py)
• Moved the anomaly detection logic for OpenAI to a new module openai.py and renamed the function to detect_anomalies_openai. (anomstack/llm/openai.py) [1] [2] [3] [4] [5]

Model and Schema Definitions:

• Created new Pydantic models Anomaly and DetectAnomaliesResponse to standardize the response schema for detected anomalies. (anomstack/llm/models.py)

Minor Improvements:

• Updated the prompt instructions to consider the entire dataset when identifying anomalies. (metrics/defaults/python/prompt.py)
• Added the anthropic library to the requirements.compile file. (requirements.compile)

[github-actions]

Your pull request is automatically being deployed to Dagster Cloud.

Location	Status	Link	Updated
anomstack		View in Cloud	Dec 27, 2024 at 09:35 PM (UTC)