fix(azure): omit model from deployment image gen and image edit bodies

[Sameerlite]

Summary

Azure OpenAI selects the model via /openai/deployments/{deployment}/... in the URL. Including model in the JSON body (image generation) or multipart form fields (image edits) sends the deployment resource name, which Azure rejects for gpt-image-2 (invalid_value / model does not exist). This aligns LiteLLM with direct Azure REST calls.

Changes

• Image generation (azure.py): POST JSON for deployment-scoped images/generations URLs omits model; /providers/... (e.g. FLUX) unchanged.
• Image edit (azure/image_edit/transformation.py): multipart data omits model when the resolved URL is deployment-scoped images/edits.

Tests

• tests/test_litellm/llms/azure/image_generation/test_azure_image_generation_init.py — wire payload assertions + helper coverage.
• tests/test_litellm/llms/azure/image_edit/test_azure_image_edit_transformation.py — new tests for form-data stripping and transform.

Fixes #26539
Fixes LIT-2757

---

Note

Low Risk
Low risk: change is narrowly scoped to Azure deployment-style image endpoints and is covered by new/updated tests, but it slightly alters wire payloads for those routes.

Overview
Azure deployment-scoped image calls now strip model from the outbound payload: JSON bodies for .../openai/deployments/.../images/generations and multipart form fields for .../openai/deployments/.../images/edits no longer include model, relying on the deployment in the URL instead.

This adds a provider hook (finalize_image_edit_multipart_data) so image-edit providers can adjust form fields after URL resolution, updates request/fixture expectations, and adds targeted tests to assert the on-the-wire payload for both deployment and non-deployment (/providers/...) routes.

^{Reviewed by Cursor Bugbot for commit c53c71a. Bugbot is set up for automated code reviews on this repo. Configure here.}

[greptile-apps]

Greptile Summary

This PR fixes Azure OpenAI deployment-scoped image generation and image edit requests by omitting model from the JSON body and multipart form fields respectively, since Azure routes by deployment name in the URL and rejects model in the payload (notably for gpt-image-2). The implementation adds a finalize_image_edit_request_data hook to the base image-edit config (no-op by default), overridden in AzureImageEditConfig to strip model when the resolved URL matches the deployment pattern, and applies a parallel helper for image generation in azure.py. Tests are updated to assert the corrected wire payload and now mock at the HTTPHandler.post level for stronger integration coverage.

Confidence Score: 5/5

Safe to merge — the change is narrowly scoped to Azure deployment-style image endpoints, aligns with Azure's documented API contract, and is covered by updated tests that assert the wire payload directly.

No P0 or P1 issues found. The fix is correct: azure_deployment_image_generation_json_body and azure_deployment_image_edit_form_data both create new dicts (no mutation of the original), the URL detection strings are specific enough to avoid false positives on non-deployment URLs, and finalize_image_edit_request_data is called after get_complete_url so it always receives the fully-resolved URL. Test updates correctly replace the old high-level mock with lower-level HTTPHandler.post mocking to actually exercise the filter in the integration path.

No files require special attention.

Important Files Changed

Filename	Overview
litellm/llms/azure/image_generation/http_utils.py	New helper that strips model from the JSON body for deployment-scoped Azure image generation URLs; clean, no logic issues.
litellm/llms/azure/image_edit/transformation.py	Adds azure_deployment_image_edit_form_data static method and finalize_image_edit_request_data override to strip model from multipart fields for deployment-scoped edit URLs.
litellm/llms/base_llm/image_edit/transformation.py	Adds no-op finalize_image_edit_request_data hook to the base class, allowing provider subclasses to post-process form data after URL resolution.
litellm/llms/custom_httpx/llm_http_handler.py	Calls finalize_image_edit_request_data(data, api_base) in both sync and async image-edit handlers after URL resolution and before logging/sending; correctly placed.
litellm/llms/azure/azure.py	Applies azure_deployment_image_generation_json_body to filter the JSON body before both sync and async image generation POSTs.
tests/test_litellm/llms/azure/image_generation/test_azure_image_generation_init.py	Updates integration tests to mock at HTTPHandler.post level and assert model is absent from the wire JSON; adds two unit tests for the new helper.
tests/test_litellm/llms/azure/image_edit/test_azure_image_edit_transformation.py	New unit tests for azure_deployment_image_edit_form_data and finalize_image_edit_request_data; pure mock, no network calls.
tests/image_gen_tests/test_image_edits.py	Updates test_azure_image_edit_litellm_sdk assertion to expect model absent from multipart form-data, aligning with the new wire behavior.
tests/image_gen_tests/request_payloads/azure_gpt_image_1.json	Removes model from the fixture file to match the updated expected wire payload.

_{Reviews (4): Last reviewed commit: "refactor(azure): move image gen JSON hel..." | Re-trigger Greptile}

[Sameerlite]

@greptile-apps re review

[Sameerlite]

bugbot run

[cursor]

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit 41ac026. Configure here.}

[Sameerlite]

bugbot run

[Sameerlite]

@greptile-apps re review

[cursor]

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit c53c71a. Configure here.}

[mateo-berri]

Some non-blocking feedback but otherwise LGTM! Thanks

Also, fix the CI tests

[codecov]

Codecov Report

❌ Patch coverage is 84.21053% with 3 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
litellm/llms/custom_httpx/llm_http_handler.py	0.00%	2 Missing ⚠️
litellm/llms/base_llm/image_edit/transformation.py	50.00%	1 Missing ⚠️

📢 Thoughts on this report? Let us know!

[Sameerlite]

@greptile-apps re review