chore(ci): merge dev branch

[yuneng-berri]

Relevant issues

Linear ticket

Pre-Submission checklist

Please complete all items before asking a LiteLLM maintainer to review your PR

• I have Added testing in the tests/test_litellm/ directory, Adding at least 1 test is a hard requirement - see details
• My PR passes all unit tests on make test-unit
• My PR's scope is as isolated as possible, it only solves 1 specific problem
• I have requested a Greptile review by commenting @greptileai and received a Confidence Score of at least 4/5 before requesting a maintainer review

Delays in PR merge?

If you're seeing a delay in your PR being merged, ping the LiteLLM Team on Slack (#pr-review).

CI (LiteLLM team)

CI status guideline:

• 50-55 passing tests: main is stable with minor issues.
• 45-49 passing tests: acceptable but needs attention
• <= 40 passing tests: unstable; be careful with your merges and assess the risk.

• Branch creation CI run
  Link:

• CI run for the last commit
  Link:

• Merge / cherry-pick CI run
  Links:

Screenshots / Proof of Fix

Type

🆕 New Feature
🐛 Bug Fix
🧹 Refactoring
📖 Documentation
🚄 Infrastructure
✅ Test

Changes

[greptile-apps]

Greptile Summary

This PR fixes two security-adjacent correctness issues in the proxy auth layer. It replaces unsafe "form" in content_type substring matching with proper RFC 7231 MIME type normalization, surfaces malformed form payloads as 400 (instead of silently caching an empty body that diverges from what the request handler reads), and propagates the ASGI scope's path into the synthetic HTTP Request built for WebSocket auth so that get_request_route never falls back to a Host-header-reconstructed URL.

• Content-type normalization (http_parsing_utils.py): new _normalize_media_type, _is_form_content_type, and _is_json_content_type helpers replace substring matching; application/json; charset=utf-8 is now correctly parsed as JSON, and non-canonical "form"-substring types no longer silently match as form data, closing a body-divergence vector.
• Form parse error hardening (_read_request_body): a failed request.form() now raises a ProxyException(code=400) so the auth-time pre-read and the downstream handler always see the same payload.
• WebSocket ASGI path fix (user_api_key_auth_websocket): scope["path"] and scope["root_path"] are copied from the real WebSocket ASGI scope into the synthetic Request, preventing get_request_route from reaching the Host-poisonable websocket.url fallback.

Confidence Score: 4/5

Safe to merge; both changes are targeted security hardening with good test coverage and no logic regressions found.

The auth and content-type changes are narrow and well-tested. The only open items are a misleading log message when a form-parse ProxyException is caught, and the fact that get_request_body → get_form_data path has no equivalent error handling for malformed form bodies (callers would see a 500 instead of a 400). Neither issue affects correctness of the security fix itself.

litellm/proxy/common_utils/http_parsing_utils.py — the get_request_body → get_form_data path lacks the malformed-form error handling that was added to _read_request_body.

Important Files Changed

Filename	Overview
litellm/proxy/common_utils/http_parsing_utils.py	Adds proper MIME type normalization helpers and fixes content-type substring matching; surfaces malformed form payloads as 400 instead of silently caching an empty body
litellm/proxy/auth/user_api_key_auth.py	Propagates ASGI scope path/root_path into the synthetic HTTP Request for WebSockets, preventing a Host-header path-poisoning in get_request_route
tests/proxy_unit_tests/test_user_api_key_auth.py	Adds focused mock-only test verifying ASGI path is carried into the synthetic request scope
tests/test_litellm/proxy/common_utils/test_http_parsing_utils.py	Adds comprehensive mock-only test coverage for the new content-type helpers and form-parse failure path

Comments Outside Diff (1)

1. litellm/proxy/common_utils/http_parsing_utils.py, line 131-134 (link)

   The outer exception handler catches ProxyException (raised from the form-parse failure path) but logs it with the message "Invalid JSON payload received", which is misleading for form errors and will confuse operators trying to diagnose why a multipart request was rejected.

_{Reviews (1): Last reviewed commit: "Merge remote-tracking branch 'origin/lit..." | Re-trigger Greptile}

[greptile-apps]

get_request_body routes form content-types to get_form_data, which calls await request.form() with no error handling. A malformed multipart body (e.g. missing boundary) will propagate as an unhandled exception rather than a clean 400. The same fix applied to _read_request_body (try/except → ProxyException(code=400)) was not applied here, so callers of get_request_body on malformed form requests will see a 500 instead of the expected 400.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!