Dev/add two stage vlm

[PeterStaar-IBM]

Idea: provide a new approach to run a two stage approach: first the layout-model on page-images and then use the predicted clusters as input (through an updated prompt) for the VLM. This approach was first explored Dolphin (be that with the same VLM).

The hope here is to improve the overall recall for VLM's.

Can be tested via the CLI,

uv run docling --pipeline vlm --vlm-model vlm2stage <path-to-doc>

TBD: how do we properly build a prompt that the VLM can use!

Checklist:

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

[github-actions]

✅ DCO Check Passed

Thanks @PeterStaar-IBM, all your commits are properly signed off. 🎉

[mergify]

Merge Protections

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

🔴 Enforce conventional commit

This rule is failing.

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)(?:\(.+\))?(!)?:

🔴 Require two reviewer for test updates

This rule is failing.

When test data is updated, we require two reviewers

• #approved-reviews-by >= 2

[codecov]

Codecov Report

❌ Patch coverage is 45.40816% with 107 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
docling/models/vlm_models_inline/mlx_model.py	13.15%	33 Missing ⚠️
...ng/models/vlm_models_inline/two_stage_vlm_model.py	38.46%	32 Missing ⚠️
.../models/vlm_models_inline/hf_transformers_model.py	20.68%	23 Missing ⚠️
docling/pipeline/vlm_pipeline.py	20.00%	12 Missing ⚠️
docling/models/base_model.py	77.77%	4 Missing ⚠️
docling/cli/main.py	0.00%	3 Missing ⚠️

📢 Thoughts on this report? Let us know!

[PeterStaar-IBM]

@dolfim-ibm @cau-git Let's start to review, functionally working, but needs now some overall thinking.

[PeterStaar-IBM]

close for now, superseded by #2084