fix(dual-solve): read grounding items from the context-pack dict

[Yurii214]

ground_prompt gated its items on isinstance(pack, ContextPack), but build_context_pack returns a model_dump'd dict in production (its signature is -> dict[str, Any]). so the guard was always false, items was always [], and dual-solve grounding always returned "the knowledge base has nothing on this topic yet." — even when the kb had matching claims. the code engines never received any kb context; the whole grounding feature was silently dead.

this reads items out of the dict (tolerating a ContextPack object too), mirroring the shape handling already in synthesize.py.

the existing test_ground_prompt_renders_items monkeypatched build_context_pack to return a real ContextPack, so it exercised the dead branch and never saw the production dict shape — which is why the bug was invisible to the suite. added test_ground_prompt_renders_real_context_pack, which seeds a real claim and drives the unmocked pipeline.

repro before: a kb with a claim matching the query still yields "nothing". after: - [c1] auth uses jwt tokens.

verified: pytest tests/test_dual_solve.py, mypy src, ruff check src tests all green.

Summary by CodeRabbit

• Bug Fixes

  • Improved knowledge-base grounding to support multiple context data formats.
  • Grounding entries are now rendered correctly regardless of their underlying representation.
  • Prevented valid context from incorrectly falling back to a “nothing found” response.

• Tests

  • Added regression coverage for grounding prompts built from real context data.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro Plus

Run ID: 7ad75331-8277-4e96-9bb2-806670caf1c7

📥 Commits

Reviewing files that changed from the base of the PR and between bfb6b0a and debb239.

📒 Files selected for processing (2)

• src/vouch/dual_solve.py
• tests/test_dual_solve.py

---

Walkthrough

ground_prompt now accepts dictionary- and object-shaped context packs and items, with a regression test covering a real KBStore claim rendered into the grounded prompt.

Changes

Context pack grounding

Layer / File(s)	Summary
Ground prompt context compatibility src/vouch/dual_solve.py, tests/test_dual_solve.py	ground_prompt handles dictionary or object context packs and items, while the regression test verifies rendering from a seeded real claim.

Estimated code review effort: 2 (Simple) | ~10 minutes

Suggested reviewers: plind-junior

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly matches the main change: grounding items are now read from the dict-form context pack.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[plind-junior]

/auto-merge