[Bugfix] diffusion end points allow model mismatch

[xiaohajiayou]

Purpose

Fix #2804

Return HTTP 400 when the request model does not match the served model for:

• /v1/images/generations
• /v1/images/edits
• video generation requests handled by the OpenAI-compatible API

Previously, these paths only logged a warning and silently used the server model, which is inconsistent with the documented error behavior for invalid parameters such as model mismatch.

• This was inconsistent with other OpenAI-serving paths in vllm-omni, where chat and speech already go through _check_model() instead of silently falling back to the server model:
  

  vllm-omni/vllm_omni/entrypoints/openai/serving_chat.py

  Lines 154 to 157 in d071f32

  	error_check_ret = await self._check_model(request)
  	if error_check_ret is not None:
  	logger.error("Error with model %s", error_check_ret)
  	return error_check_ret

  
  

  vllm-omni/vllm_omni/entrypoints/openai/serving_speech.py

  Lines 1628 to 1631 in d071f32

  	error_check_ret = await self._check_model(request)
  	if error_check_ret is not None:
  	logger.error("Error with model %s", error_check_ret)
  	return error_check_ret

  
  

  vllm-omni/vllm_omni/entrypoints/openai/serving_speech.py

  Lines 1712 to 1714 in d071f32

  	error_check_ret = await self._check_model(batch_request)
  	if error_check_ret is not None:
  	return error_check_ret

  .

• It was also inconsistent with the image API docs, which already classify model mismatch as an invalid-parameter 400 Bad Request case:
  

  vllm-omni/docs/serving/image_generation_api.md

  Lines 176 to 179 in d071f32

  	### 400 Bad Request
  	Invalid parameters (e.g., model mismatch):

  
  

  vllm-omni/docs/serving/image_edit_api.md

  Lines 162 to 165 in d071f32

  	### 400 Bad Request
  	Invalid parameters (e.g., model mismatch):

Changes

• Update /v1/images/generations to raise HTTPException(400) on model mismatch
• Update /v1/images/edits to raise HTTPException(400) on model mismatch
• Update video request parsing to raise HTTPException(400) on model mismatch
• Remove the previous warning-only fallback behavior for these endpoints

---

Essential Elements of an Effective PR Description Checklist

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan. Please provide the test scripts & test commands. Please state the reasons if your codes don't require additional test scripts. For test file guidelines, please check the test style doc
• The test results. Please paste the results comparison before and after, or the e2e results.
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model. Please run mkdocs serve to sync the documentation editions to ./docs.
• (Optional) Release notes update. If your change is user-facing, please update the release notes draft.

BEFORE SUBMITTING, PLEASE READ https://github.com/vllm-project/vllm-omni/blob/main/CONTRIBUTING.md (anything written below this line will be removed by GitHub Actions)

[chatgpt-codex-connector]

Codex usage limits have been reached for code reviews. Please check with the admins of this repo to increase the limits by adding credits.
Credits must be used to enable repository wide code reviews.

[lishunyang12]

Review Summary

This PR correctly changes three diffusion endpoints from logging a warning on model mismatch to raising HTTPException(400). The change aligns these endpoints with the existing behavior in serving_chat.py and serving_speech.py, and with what the API docs already promise.

Correctness

The logic is straightforward and correct across all three sites:

1. /v1/images/generations - Replaces logger.warning + silent fallback with raise HTTPException(400). The stale comment # Validate model field (warn if mismatch, don't error) is properly removed.
2. /v1/images/edits - Same pattern, same fix.
3. _parse_video_form - The raise happens inside a try block that already has except HTTPException: raise, so the new exception propagates correctly without being swallowed by the generic except Exception handler below it.

Error messages are clear and consistent across all three sites. Using HTTPStatus.BAD_REQUEST.value is consistent with the rest of the file.

Tests

Test coverage looks good: one test per endpoint (test_generate_images_rejects_model_mismatch, test_image_edit_rejects_model_mismatch, test_video_generation_rejects_model_mismatch), each asserting both the 400 status code and the presence of "model mismatch" in the error detail. The assertions use .lower() which makes them resilient to casing changes.

Minor Observations (non-blocking)

• The parentheses around the f-string in detail=(f"Model mismatch: ...") are unnecessary (single expression, not a tuple), but this is purely cosmetic and not worth a revision.

LGTM. Clean, minimal, well-tested fix that resolves a real behavioral inconsistency.

[Copilot]

Pull request overview

Fixes #2804 by making diffusion-oriented OpenAI-compatible endpoints reject requests whose model does not match the server’s served model, aligning behavior with existing chat/speech paths and the image API docs (400 on invalid params) instead of silently falling back.

Changes:

• /v1/images/generations: raise HTTP 400 on request.model mismatch vs served model.
• /v1/images/edits: raise HTTP 400 on model form field mismatch vs served model.
• Video form parsing for /v1/videos + /v1/videos/sync: raise HTTP 400 on request.model mismatch.
• Added regression tests asserting 400 + “model mismatch” detail for image + video endpoints.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
vllm_omni/entrypoints/openai/api_server.py	Enforces model mismatch validation (HTTP 400) for image generation/edit and video request parsing.
tests/entrypoints/openai_api/test_video_server.py	Adds test ensuring video generation rejects model mismatch with HTTP 400.
tests/entrypoints/openai_api/test_image_server.py	Adds tests ensuring image generation and image edits reject model mismatch with HTTP 400.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[SamitHuang]

conflicts to be fixed, otherwise LGTM

[xiaohajiayou]

@lishunyang12 @hsliuustc0106 Could you please take a look if we can merge this?