fix(plans)+test: #500 review — OCR rides Full not Compressed; OCR-RT gate exact

[AdaWorldAPI]

Addresses the two open review threads on the merged #500.

• codex P2 (mine): I'd written "post-POC OCR rides Compressed" — but Compressed lacks Energy+Plasticity, so the schema-gated LayoutBlock::to_node_row would silently drop confidence (→Energy) and repair-provenance (→Plasticity). Corrected: OCR rides Full (the only existing preset with the codec residues AND the hot lifecycle columns). The OCR-SCHEMA contract test now asserts Compressed lacks Energy/Plasticity (regression guard) so this can't silently regress.
• CodeRabbit Major: the OCR-RT reversibility gate was ≥ 99 % — a lossy residue→rank map is not "reversible", so the gate is tightened to 100 % exact; any single miss fails it, and tolerance is moved to a separate quality probe.

ocr-canonical-soa-integration-v1.md + ocr-probes-v1.md + the contract test updated. contract lib green; fmt clean.

https://claude.ai/code/session_01D2WSmezQBNC3bUdHuGfGmo

---

Generated by Claude Code

Summary by CodeRabbit

Release Notes

• Documentation

  • Refined OCR integration specifications regarding preset configuration requirements.
  • Updated OCR testing criteria to enforce stricter validation standards.

• Tests

  • Enhanced test assertions for improved validation rigor and consistency.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 88a898bd-6bcb-4ee9-9330-fabd44ab4278

📥 Commits

Reviewing files that changed from the base of the PR and between e192266 and d65dcb7.

📒 Files selected for processing (3)

• .claude/plans/ocr-canonical-soa-integration-v1.md
• .claude/plans/ocr-probes-v1.md
• crates/lance-graph-contract/src/ocr.rs

---

📝 Walkthrough

Walkthrough

The PR corrects the OCR integration plan to specify ValueSchema::Full (not Compressed) as the preset OCR rides, tightens the OCR-RT probe's pass criterion from ≥99% to exact 100% reversibility, and strengthens the ocr_schema_fit_rides_existing_preset_no_new_variant contract test with explicit assertions that Compressed lacks Energy and Plasticity tenants.

Changes

OCR Preset Correction and Probe Tightening

Layer / File(s)	Summary
Plan and probe spec: Full preset + 100% RT gate .claude/plans/ocr-canonical-soa-integration-v1.md, .claude/plans/ocr-probes-v1.md	The integration plan replaces Compressed with Full as the OCR ValueSchema preset, restates the no-new-enum-variant contract rule, and describes the opt-in route via operator or OGAR class. The probe spec removes the ≥99% tolerance threshold and requires exact 100% residue→rank→word round-tripping; any miss now fails the gate.
Contract test: assert Compressed lacks Energy/Plasticity, OCR rides Full crates/lance-graph-contract/src/ocr.rs	Rewrites the test comment to document the Full-not-Compressed rationale, adds explicit assertions that ValueSchema::Compressed does not contain ValueTenant::Energy or ValueTenant::Plasticity, and updates the explanatory comment to confirm OCR therefore rides Full for the needed hot lifecycle columns.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~8 minutes

Possibly related PRs

• AdaWorldAPI/lance-graph#500: Rebaselines OCR plans and records the same ocr_schema_fit_rides_existing_preset_no_new_variant contract-test intent that this PR's assertions directly strengthen.
• AdaWorldAPI/lance-graph#498: Also touches crates/lance-graph-contract/src/ocr.rs to enforce OCR's ValueSchema preset and tenant expectations around Full vs Compressed.

Poem

🐇 Hop, hop — no more Compressed disguise,
The Full preset is where OCR flies!
Energy and Plasticity must reside,
One hundred percent round-trips, no threshold to hide.
The rabbit checked each tenant with care —
No new enum variants allowed, I declare! 🌿

🚥 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 accurately reflects the main changes: OCR schema preset correction (Full vs Compressed) and tightening of the OCR-RT reversibility gate criteria, with clear reference to the review issue #500.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.
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.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

---

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 and usage tips.}

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: d65dcb758c

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Align stale Compressed references with Full requirement

This correction now says post-POC OCR must ride Full because Compressed lacks Energy/Plasticity, but the same plan still tells implementers in §2 and the D-OCR-51 deliverable to use Full POC / Compressed. If the D-OCR-51 work follows those checklist lines, it can reintroduce the exact data-loss path this change is guarding against by selecting Compressed and dropping confidence/repair provenance, so the remaining summaries should be updated to Full or an operator-minted preset.

Useful? React with 👍 / 👎.