[AdaWorldAPI]

Summary

Follow-up to #1870, which merged before these two commits were pushed.

DR64 is global to local, not a size limit. OnnxGraph::build refused any graph beyond one 64 × 64 tile (TileOverflow).

• A global position p is bit p % 64 of tile p / 64. on, do_ and dep are dr3::Tiled relations (their non-empty DR64 tiles). dep composes tile by tile with the same relcomp::compose kernel, and schedule reads dep's pairs.
• The only refusal left is the u16 address width: Refusal::AddressOverflow, 65,536 positions per axis.
• Jina, measured: the full graph (7,620 nodes, 7,939 values) builds. Its tiled dep (391 tiles, 9,176 edges) equals the source edges, and it schedules with every edge running forward.

The codebook spans mask blocks. One WideFieldMask position is a u8, so the ONNX lens stopped at 256 fields. Registry::exposed is now a list of masks: field o is position o % 256 of block o / 256, up to 65,536 fields. Registry::expose shows or hides one field.

Tests

• New: graphs_beyond_one_tile_lay_over_tiles (a 200-node chain over four tiles, with a cross-tile edge), codebook_spans_mask_blocks (600 fields over three blocks), plus full-graph assertions in the Jina test.
• 25 lib, 40 tiny and 9 Jina tests pass (the Jina ones are ignored by default and need the fetch scripts). Clippy and fmt are clean.
• Disable runs, each red: compose restricted to the diagonal column tile; exposes reading block 0 only; set dropping cross-tile bits.

Not claimed

plan::resolve still refuses Jina at its first unregistered node (346 remain). Tiling removes the size refusal, not the coverage gap.

Plan §0.9; board entry and STATUS_BOARD row updated.

---

Generated by Claude Code

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

Run configuration

• Source mode: Auto
• Repository: Unavailable
• Branch: Unavailable
• File: No configuration file from this repository applied
• Commit: Unavailable
• Layers: Organization UI
• Review profile: CHILL
• Plan: Advanced
• Run ID: 65906e59-e28b-4520-873e-534f6423b18e

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

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.}

[AdaWorldAPI]

style / format failed on crates/lance-graph-contract/src/morton8x8.rs, which this PR doesn't touch. main has failed the same check since c04ffe92e (the D-LITHO G3 morton8x8 test): one assert_eq! is over the line width.

Ported the fix into this PR as 5295c0fdc (rustfmt only, 5+/1−). Every crate the style job checks now passes cargo fmt --check locally.

---

Generated by Claude Code

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-30T16:37:20.246189Z	5295c0f	Draft marked ready

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" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.