mirror of
https://github.com/BerriAI/litellm.git
synced 2026-10-09 03:18:44 +00:00
Files and batches routes take their model from a header, query param or a model-encoded resource id, which the auth layer never sees, so any key could name any deployment and act on that provider account with its server-side key. Every caller-supplied model now goes through can_key_call_resolved_model before deployment credentials are resolved, covering file create/retrieve/content/ delete/list, batch create/retrieve/list/cancel, and vector store files. |
||
|---|---|---|
| .. | ||
| __init__.py | ||
| test_endpoints.py | ||