llm: omit image collection names in prompt placeholders

[matthewYang08]

Summary

This PR addresses #521 by omitting collection names when rendering image collections into prompt placeholders.

What changed

• Updated StructuredContent.render_with_images() behavior:
  • Omit field/key names only for image collections (lists/tuples/dicts of images).
  • Keep semantic keys for mixed-content dicts.
  • Keep non-collection image field labels (e.g. page_view) unchanged.
• Added regression tests for:
  • Plain image list rendering without collection name
  • Dict image collection rendering without keys
  • Tuple image collection behavior
  • Mixed dict behavior preserving semantic keys
  • Prompt extraction expectation for nested with_images

Validation

• pytest tests/unit/pipelex/core/stuffs/test_structured_content_render_with_images.py
• pytest tests/integration/pipelex/pipes/llm_prompt_inputs/test_prompt_image_extraction.py --disable-inference
• ruff check on changed files
• pyright on changed files

Refs #521

---

Summary by cubic

Updates prompt rendering to omit names for image-only collections (lists, tuples, dicts) and keep labels for single-image fields, reducing noise and aligning with image extraction (addresses #521). Also extends nested image discovery to detect dict-based image collections.

• Bug Fixes
  • Updated StructuredContent.render_with_images() to hide names for image-only collections; preserve names for mixed dicts and non-collection fields.
  • Improved image-only detection across nested sequences, dicts, and ListContent via _is_image_collection_value() and _is_image_only_value().
  • Extended nested image discovery to include dict generics in search_for_nested_image_fields()/check_generic_container_for_images(); added tests for dict containers, tuples, nested lists, mixed dicts, and label preservation in prompt extraction.

^{Written for commit c8a3732. Summary will update on new commits. Review in cubic}

[github-actions]

All contributors have signed the CLA ✍️ ✅
_{Posted by the CLA Assistant Lite bot.}

[matthewYang08]

I have read the CLA Document and I hereby sign the CLA

[greptile-apps]

Greptile Summary

This PR changes how image collections are rendered in prompt placeholders. The main changes are:

• Omits field names for image-only lists, tuples, and dicts.
• Keeps labels for single image fields such as page_view.
• Keeps semantic keys for mixed dicts with image and non-image values.
• Adds regression coverage for collection rendering and nested prompt extraction.

Confidence Score: 4/5

This is close, but the dict image collection path should be fixed before merging.

• Dict image collections can render in StructuredContent but can still be rejected by prompt validation.

• Existing image-field discovery does not appear to recognize dict containers.

• List, tuple, mixed dict, and single-image behavior look covered by the changed tests.

• pipelex/core/stuffs/image_field_search.py should be updated alongside the new dict rendering behavior.

Important Files Changed

Filename	Overview
pipelex/core/stuffs/structured_content.py	Adds image-only collection detection and suppresses labels for those rendered collections.
tests/unit/pipelex/core/stuffs/test_structured_content_render_with_images.py	Adds unit coverage for image-only lists, dicts, tuples, nested lists, and mixed dicts.
tests/integration/pipelex/pipes/llm_prompt_inputs/test_prompt_image_extraction.py	Adds coverage that single nested image fields still keep their label.

Prompt To Fix All With AI

Fix the following 1 code review issue. Work through them one at a time, proposing concise fixes.

---

### Issue 1 of 1
pipelex/core/stuffs/structured_content.py:203-205
**Dict fields stay undiscoverable** The renderer now treats `dict[str, ImageContent]` as an image collection, but the image-field discovery path used by prompt validation still only recognizes list and tuple containers. A structured concept whose only images are in a dict field can render correctly here, but `{{ gallery | with_images }}` is rejected earlier as having no nested images, so this new dict support is unreachable in normal prompt flows.

_{Reviews (1): Last reviewed commit: "llm: omit image collection names in prom..." | Re-trigger Greptile}

[matthewYang08]

感谢你的投稿，我们非常感激。像许多开源项目一样，我们要求您签署贡献者许可协议，才能接受您的贡献。你只需发布一则拉取请求评论，格式与以下相同即可签署CLA。

我已阅读CLA文件，特此签署CLA

您可以通过在此拉取请求中评论recheck来重新触发该机器人。由CLA助理轻量机器人发布。

recheck

[greptile-apps]

Dict fields stay undiscoverable The renderer now treats dict[str, ImageContent] as an image collection, but the image-field discovery path used by prompt validation still only recognizes list and tuple containers. A structured concept whose only images are in a dict field can render correctly here, but {{ gallery | with_images }} is rejected earlier as having no nested images, so this new dict support is unreachable in normal prompt flows.

Prompt To Fix With AI

This is a comment left during a code review.
Path: pipelex/core/stuffs/structured_content.py
Line: 203-205

Comment:
**Dict fields stay undiscoverable** The renderer now treats `dict[str, ImageContent]` as an image collection, but the image-field discovery path used by prompt validation still only recognizes list and tuple containers. A structured concept whose only images are in a dict field can render correctly here, but `{{ gallery | with_images }}` is rejected earlier as having no nested images, so this new dict support is unreachable in normal prompt flows.

How can I resolve this? If you propose a fix, please make it concise.

[cubic-dev-ai]

No issues found across 3 files

_{Re-trigger cubic}

[thomashebrard]

Closed because it was pointing on main: see #917