test: add /v1/messages to supported_endpoints schema enum

[tin-berri]

Relevant issues

CI is currently red on every backend PR against litellm_internal_staging; #32727 shows the same failure on an unrelated change

• feat: add Meta Model API provider and muse-spark-1.1 (day-0) #32701 added the meta/muse-spark-1.1 entry to model_prices_and_context_window.json with "/v1/messages" in its supported_endpoints
• test_aaamodel_prices_and_context_window_json_is_valid validates every entry of that json against a hardcoded endpoint allow-list inside tests/test_litellm/test_utils.py; the list predates /v1/messages, so jsonschema rejects the new entry
• the misc / Run tests job runs against the merge of the PR branch with current staging, so the muse-spark entry is present in every backend PR's run and the job fails on the same assertion regardless of what the PR itself changes
• adding "/v1/messages" to the allow-list makes the muse-spark entry validate again; the list has no consumers outside this one test, so the one-line change fixes the job for every PR without touching product behavior

Linear ticket

Pre-Submission checklist

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

• I have added meaningful tests
• My PR passes all CI/CD checks (e.g., lint, format, unit tests)
• My PR's scope is as isolated as possible; it only solves 1 specific problem
• I have received a Greptile Confidence Score of at least 4/5 before requesting a maintainer review (Greptile reviews automatically once the PR is opened; only comment @greptileai to re-request a review after pushing changes)

Screenshots / Proof of Fix

The defect is inside the test suite itself (a stale schema allow-list), so there is no runtime surface to curl; the proof is the exact check CI runs, captured at the pre-fix and post-fix commits

Before, at 2a12707372 (current litellm_internal_staging tip, which contains the muse-spark entry from #32701):

git checkout 2a12707372 -- tests/test_litellm/test_utils.py
python -m pytest "tests/test_litellm/test_utils.py::test_aaamodel_prices_and_context_window_json_is_valid" -q

E           jsonschema.exceptions.ValidationError: '/v1/messages' is not one of ['/v1/responses', '/v1/embeddings', '/v1/chat/completions', '/v1/completions', '/v1/images/generations', '/v1/realtime', '/v1/realtime/transcription_sessions', '/v1/images/variations', '/v1/images/edits', '/v1/batch', '/v1/audio/transcriptions', '/v1/audio/speech', '/v1/ocr', '/vertex_ai/live', '/v1/realtime/transcription_sessions']
E           On instance['meta/muse-spark-1.1']['supported_endpoints'][2]:
E               '/v1/messages'
1 failed in 0.51s

After, at 5409045644 (this PR):

python -m pytest "tests/test_litellm/test_utils.py::test_aaamodel_prices_and_context_window_json_is_valid" -q

1 passed in 0.36s

The full file also passes at 5409045644: python -m pytest tests/test_litellm/test_utils.py -q gives 227 passed in 0.66s

Type

✅ Test

Changes

#32701 added the meta/muse-spark-1.1 entry to model_prices_and_context_window.json declaring /v1/messages in supported_endpoints, without extending the endpoint allow-list that test_aaamodel_prices_and_context_window_json_is_valid validates the json against. The test therefore fails on litellm_internal_staging itself, and since CI runs against the merge of each PR branch with current staging, every backend PR fails this job regardless of its own diff

Extending the enum (rather than editing the muse-spark entry) is the behavior-preserving fix. /v1/messages is a real served route (@router.post("/v1/messages") in litellm/proxy/anthropic_endpoints/endpoints.py) and the value is load-bearing at runtime: _deployment_passes_through_anthropic_messages in litellm/llms/anthropic/experimental_pass_through/messages/handler.py gates Anthropic-spec passthrough on exactly this entry, so stripping it from the model entry would disable the feature #32701 shipped. The test enum has no consumers outside this test, so adding the member changes no product behavior

The same enum listed /v1/realtime/transcription_sessions twice; this PR drops the duplicate. Net diff is one line added, one removed

[greptile-apps]

Greptile Summary

This PR fixes a CI-breaking test failure caused by a missing enum entry in the JSON schema validator used by test_aaamodel_prices_and_context_window_json_is_valid. The prior PR (#32701) added meta/muse-spark-1.1 with /v1/messages in supported_endpoints, but the allow-list in this test was never updated to include it, causing every subsequent backend PR to fail CI.

• Adds /v1/messages to the supported_endpoints enum in the schema validator — /v1/messages is a real served route and a load-bearing value for Anthropic passthrough logic, so expanding the enum is the correct fix rather than removing it from the model entry.
• Removes a duplicate /v1/realtime/transcription_sessions entry that was already present in the list.

Confidence Score: 5/5

Safe to merge — the change is a minimal, targeted test fix with no production code touched.

The only change is adding a real, existing endpoint to a validation enum and removing a duplicate entry. It correctly unblocks CI by bringing the test allow-list in sync with the model data, and does not weaken any assertion or mask a regression.

No files require special attention.

Important Files Changed

Filename	Overview
tests/test_litellm/test_utils.py	Extends the supported_endpoints schema enum to include /v1/messages (a real served route) and removes a duplicate /v1/realtime/transcription_sessions entry — unblocks CI without weakening validation

_{Reviews (1): Last reviewed commit: "test: add /v1/messages to supported_endp..." | Re-trigger Greptile}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[codspeed-hq]

Merging this PR will not alter performance

✅ 31 untouched benchmarks

---

_{Comparing litellm_supported_endpoints_v1_messages (5409045) with litellm_internal_staging (2a12707)}