fix: forward VLM convert runtime generation settings

[geoHeil]

Forward the VLM convert stage generation settings from model_spec when constructing VlmEngineInput, and add debug timing around preprocessing and batch inference.

This change:

• adds temperature and extra_generation_config to VlmModelSpec
• threads temperature, max_new_tokens, stop_strings, and extra_generation_config through both VlmConvertModel.__call__() and process_images()
• logs rasterization/resize timing and batch inference timing in the VLM convert stage
• adds regression tests covering both VLM convert entry points

Issue resolved by this Pull Request:
Resolves #3321

Checklist:

• Documentation has been updated, if necessary.
• Examples have been added, if necessary.
• Tests have been added, if necessary.

[mergify]

Merge Protections

Your pull request matches the following merge protections and will not be merged until they are valid.

🟢 Enforce conventional commit

Wonderful, this rule succeeded.

Make sure that we follow https://www.conventionalcommits.org/en/v1.0.0/

• title ~= ^(fix|feat|docs|style|refactor|perf|test|build|ci|chore|revert)(?:\(.+\))?(!)?:

[github-actions]

✅ DCO Check Passed

Thanks @geoHeil, all your commits are properly signed off. 🎉

[Copilot]

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

This PR forwards VLM convert-stage generation settings from VlmModelSpec into VlmEngineInput, adds debug timing logs around preprocessing and inference, and introduces regression tests to ensure both VLM convert entry points respect the configured generation settings.

Changes:

• Extend VlmModelSpec with temperature and extra_generation_config, and forward generation settings when constructing VlmEngineInput.
• Add debug timing for image preparation (rasterize/resize) and batch inference in the VLM convert stage.
• Add regression tests for VLM convert generation settings and API engine parameter precedence behavior.

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
tests/test_vlm_convert_model.py	Adds regression tests asserting generation settings are forwarded for both process_images() and __call__() paths.
tests/test_api_vlm_engine.py	Adds tests validating API param precedence between model defaults, per-request settings, and explicit user params.
docling/models/stages/vlm_convert/vlm_convert_model.py	Centralizes VlmEngineInput construction and adds debug timing logs for preprocessing and inference.
docling/models/inference_engines/vlm/api_openai_compatible_engine.py	Changes API parameter precedence/merging and introduces separate storage for model vs user params.
docling/datamodel/stage_model_specs.py	Adds temperature and extra_generation_config fields to VlmModelSpec.

---

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

[codecov]

Codecov Report

❌ Patch coverage is 96.72131% with 2 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
...ing/models/stages/vlm_convert/vlm_convert_model.py	94.44%	2 Missing ⚠️

📢 Thoughts on this report? Let us know!