mirror of
https://github.com/BerriAI/litellm.git
synced 2026-09-20 00:11:50 +00:00
4b9c289a72
* fix(a2a): forward caller identity headers on message/send and message/stream _forwarding_headers() stamped X-LiteLLM-User-Id/-Team-Id from the authenticated caller, but was only wired into the tasks/* and tasks/resubscribe branches. The primary message/send and message/stream conversational path forwarded agent_extra_headers unchanged, so a downstream agent never learned which end user was calling it except on secondary task-management calls. This broke per-user MCP scoping and per-customer FinOps budget enforcement for any agent invoked through the normal conversational flow. Resolves LIT-7342 * fix(a2a): snapshot key-bound caller identity before pre-call processing user_header_mappings lets add_litellm_data_to_request rewrite user_api_key_dict.user_id from a client header, so the identity stamped onto X-LiteLLM-User-Id is now captured before that step runs. Header tests updated to expect the caller identity on message/send. Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> * test(a2a): drop redundant docstrings from message identity tests Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> * test(a2a): type the message method test helpers Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> * test(a2a): make message method test helpers immutable Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --------- Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> |
||
|---|---|---|
| .. | ||
| auth | ||
| a2a_endpoints.py | ||
| a2a_routing.py | ||
| agent_registry.py | ||
| agent_search.py | ||
| databricks_oauth.py | ||
| endpoints.py | ||
| model_list_helpers.py | ||
| utils.py | ||