Repository navigation
Offer complete three-rule table drafts with visible review guidance - #233
Merged
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Captioned tables with three horizontal rules can have complete source text but insufficient borders for verified cell structure. With
emit_table_proposals: true, qualifying regions now include a proposed table arrangement with exact source text, multiline fragments and an inert HTML preview. The text response shows the preview and a visible notice to check its grouping against the PDF.The draft uses whitespace corridors recurring across every complete body line. Inconsistent partitions, overlapping or wrapped rows, empty cells, straddling text and incomplete evidence produce no draft. Every retained item must appear exactly once. The candidate binds the PDF, region, algorithm, assignments and presentation; validation reconstructs the draft from retained geometry and, when supplied, the full layout.
Ordinary Markdown, gaps, document-map content and strict
verify_table_proposalacceptance stay unchanged. Draft cells remain explicitly unverified and use no inferred header roles or spans.Validation includes synthetic adversarial cases, actual MCP presentation and strict-rejection checks, and a fresh parse of a retained measured PDF. That page produces a deterministic five-row, four-column draft using all 38 text items, preserving its multiline label and values while excluding neighboring prose. The private PDF is not included here. Source/share files match, and independent review accepted the implementation after a packet-validation gap was corrected.
Mac Node 22.23.2 passed all 303 tests across 12 affected/adjoining suites, including conversion, packet validation, strict verification, MCP discovery, packaging coverage, document maps and extraction scorer/evidence tests. Fresh measured-page replay also passed at the final source. Generated oracle changes are exclusively five source-provenance values; oracle cases are unchanged. Captured tool-contract changes are the deliberate option description and corresponding digests.
This PR establishes a usable draft for review. It makes no general table-accuracy or benchmark claim.