[dependabot]

Bumps juan294/sutura from 0.5.1 to 0.5.2.

Release notes

Sourced from juan294/sutura's releases.

v0.5.2

Sutura v0.5.2

Fixed

• Ignore signal names in Node test-result titles when judging failed-step completion. Real runner termination output still blocks diagnosis.
• Scan uv.lock source tables in linear time while preserving unsupported-dependency refusals.
• Read retained v1 daily fleet snapshots without inventing newer-schema fields.

Changed

• Add daily read-only judging-access collection and pinned public evidence.
• Refresh development dependencies, grouped CodeQL references and the GitHub App token action.
• Install local resource ownership guidance and retain the Devpost and overlay-investigation records.

Retired guidance: none.

This release does not repair consumer PTY flakiness or admit dependency-changing commits with mismatched reproduction-profile hashes. It has no new benchmark measurement. The Case Lab refresh deferral expired at 2026-11-05T13:02:35Z. On 2026-11-05, the repository owner approved a one-time exception to publish this existing v0.5.2 tag after that deadline, with the Case Lab refresh still deferred. The demo continues to label its v0.4.1 evidence.

Changelog

Sourced from juan294/sutura's changelog.

Changelog

[Unreleased]

[0.5.2] - 2026-11-05

Fixed

• Ignore signal names in Node test-result titles when checking failed-step completion. Actual runner termination output still blocks diagnosis, while passing tests named after SIGTERM or SIGKILL no longer make complete failure diagnostics appear incomplete.
• Scan uv.lock source tables in linear time while preserving refusal of unsupported dependency sources.
• Keep retained v1 daily fleet snapshots readable without inventing fields from newer schemas.

Changed

• Add daily read-only judging-access collection and a pinned public evidence manifest.
• Refresh development dependencies, grouped CodeQL references and the repository-scoped GitHub App token action.
• Install the local resource ownership policy and retain the Devpost refresh and overlay investigation records.

Retired guidance: none.

The signal-title fix does not repair consumer test flakiness or admit dependency-changing commits whose reproduction-profile hashes do not match. The Case Lab still labels its v0.4.1 evidence. The owner approved carrying its refresh deferral to v0.5.2 through 2026-11-05T13:02:35Z without extending the original expiry. This release has no new benchmark measurement.

[0.5.1] - 2026-10-29

Fixed

• Split large repository overlays into sequential archives capped at 64 MiB, preserving prepared dependencies and the complete source manifest. Refuse unsplittable files and enforce one shared deadline across chunk uploads and extraction.
• Preserve provider failure metadata and add the failed chunk's position and size to overlay errors.

Changed

• Updated Devpost for the published Jev release while preserving historical benchmark identities. Recorded one successful live 144 MiB Node overlay repair; this does not establish fleet recovery or a new benchmark score.
• Retained the owner-approved Case Lab refresh deferral through November 5 without extending its expiry; its exact previous and new tags remain covered during publication.

Retired guidance: none.

[0.5.0] - 2026-10-29

Added

• Explicit, default-off Jev ranking, grounding, classification and recovery roles for local healing and the GitHub Action, with a separate shared spend cap, pinned model, bounded typed choices and native fallbacks.
• sutura explain selects original evidence from completed case JSON without network by default; an explicitly capped semantic mode selects one best supplied excerpt.
• Finite Jev decision evaluation with frozen manifests, retained individual outcomes, resumable spend journals and explicit qualification limits. Experimental atomic audit questions retain the existing legacy veto behavior.
• Protected Node reproduction profiles, complete native failure identity and separately approved one-use final verification for longer trusted commands.

Fixed

• Prevent regex backtracking in native command parsing on whitespace-heavy input.
• Semantic evidence navigation and recovery selection no longer depend on independent model questions coordinating optional ranked choices. Required judgments retain the existing strict probability thresholds.
• Navigator evaluation scores actual selected evidence, excludes a declared no-result sentinel and rejects false retrieval credit from unselected probability alternatives.

... (truncated)

Commits

• c953b4b release: v0.5.2
• See full diff in compare view

Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting @dependabot rebase.

---

Dependabot commands and options

You can trigger Dependabot actions by commenting on this PR:

• @dependabot rebase will rebase this PR
• @dependabot recreate will recreate this PR, overwriting any edits that have been made to it
• @dependabot show <dependency name> ignore conditions will show all of the ignore conditions of the specified dependency
• @dependabot ignore this major version will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)
• @dependabot ignore this minor version will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)
• @dependabot ignore this dependency will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)

[strix-security]

Strix is installed on this repository, but we couldn't run this PR security review because this workspace's trial has ended. Add a card to resume code reviews here.

So far, Strix has reviewed 50 pull requests, surfaced 7 security issues (1 critical/high) and blocked 1 risky merge across this workspace.