OpenAI Integration for AgentM Python

[debuggerone]

Overview

This issue focuses on integrating OpenAI's API with the Python version of the AgentM library. This includes setting up the API client, configuring models, handling tokenization, and managing model-specific functionalities like completions and structured outputs.

Tasks

• [x] API Client Setup

  • [x] Use an existing OpenAI Python client to manage API interactions.
  • [x] Migrate OpenAI-specific configurations from openai.ts to Python, and store them in .settings/config.py.
  • [x] Implement environment variable management for sensitive information like API keys.

• [x] Model Handling

  • [x] Implement Python functions to handle OpenAI model completions, structured outputs, and error handling.
  • [ ] Ensure robust error handling, including retry logic and wait mechanisms for handling rate limits (e.g., 429 status codes).
  • [ ] Implement logging and monitoring for API usage and errors.

• [ ] Tokenizer Support

  • [ ] Migrate the tokenizer configuration logic from openai.ts to Python.
  • [ ] Ensure compatibility with Python tokenizer libraries and validate token counts for different models.
  • [ ] Test tokenization for various inputs to ensure accuracy and performance.

Subtasks Overview

1. API Client Setup:

   • Set up and configure the OpenAI Python client, including environment management for API keys and secrets.
   • Store configuration details in a .settings/config.py file for easy access and modification.

2. Model Handling:

   • Develop Python functions for managing OpenAI model interactions, including completions and error handling.
   • Implement logic to handle API rate limits and errors effectively.

3. Tokenizer Support:

   • Port tokenizer configurations and functions to Python, ensuring they work seamlessly with Python libraries.
   • Validate that token counts are accurate and aligned with OpenAI’s token limits for various models.