Summary

• Adds support for passing extra_headers to LiteLLM completion APIs from OpenHands
• New environment variable LLM_EXTRA_HEADERS accepts a JSON object which is forwarded to litellm.completion(), acompletion() (async), and streaming acompletion()
• Includes unit tests verifying that extra_headers are forwarded to LiteLLM in all execution paths

Changes

• openhands/llm/llm.py
  • Parse LLM_EXTRA_HEADERS (JSON) from env and, if valid, forward as extra_headers to litellm.completion
• openhands/llm/async_llm.py
  • Same parsing and forwarding for async acompletion calls
• openhands/llm/streaming_llm.py
  • Same parsing and forwarding for streaming acompletion calls
• tests/unit/llm/test_extra_headers.py
  • New tests covering sync, async, and streaming cases to ensure headers are forwarded

How to use

• Set LLM_EXTRA_HEADERS to a JSON object of headers

Verification

• pre-commit hooks passed locally for backend files
• Unit tests added and passing: poetry run pytest -q tests/unit/llm/test_extra_headers.py

Backward compatibility

• If LLM_EXTRA_HEADERS is unset or invalid, no change in behavior.

Related Issues

• Fixes use github copilot in local litellm #11330

Notes

• We intentionally keep this minimal and env-driven; future improvements could expose this through explicit config if needed.

@enyst can click here to continue refining the PR

To run this PR locally, use the following command:

GUI with Docker:

docker run -it --rm   -p 3000:3000   -v /var/run/docker.sock:/var/run/docker.sock   --add-host host.docker.internal:host-gateway   -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:c4ff3d6-nikolaik   --name openhands-app-c4ff3d6   docker.all-hands.dev/all-hands-ai/openhands:c4ff3d6

CLI with uvx:

uvx --python 3.12 --from git+https://github.com/All-Hands-AI/OpenHands@openhands/llm-extra-headers#subdirectory=openhands-cli openhands