Resources in this directory are used to debug LiteLLM and other AI providers.
Note that the requests are meant to be identical to what the Onyx application sends to litellm.
Double check that this is the case before using these scripts.

## Files

- **payload.json**: Contains a typical request from Onyx

- **test_litellm.py**: Imports Onyx's LiteLLM instance (with monkey patches) and outputs the raw stream events received back from LiteLLM as JSON. Does not use payload.json, but has a similar request body. Consider directly importing litellm to skip monkey patching.

- **directly_hit_azure_api.py**: Directly hits Azure OpenAI endpoints using payload.json. Bypasses LiteLLM for debugging purposes.