镜像站点 · 本页由第三方 GitHub 只读镜像提供,非 GitHub 官方站点,不接受任何登录或凭据输入。前往 github.com
Skip to content

chore(models): sync Command Code catalog to 1.88.3 - #199

Merged
pierreraby merged 1 commit into
mainfrom
chore/catalog-1.88.3
Nov 4, 2026
Merged

pierreraby merged 1 commit into
mainfrom
chore/catalog-1.88.3

Conversation

@pierreraby

Copy link
Copy Markdown
Collaborator

Summary

Synchronize the generated capability snapshot and its documented version with the published command-code@1.88.3 package. This is a focused catalog PR on current main, separate from the entrypoint typecheck work in #198.

  • inclusionai/ling-3.1-flash-1019: reasoning, low/medium/high efforts, 262,144-token output limit; remains text-only.
  • stealth/crouton-alpha: image input, reasoning, low/medium/high/xhigh/max efforts, 943,718-token output limit.
  • Preserve the existing context-window cap on output limits. Pricing, endpoint routing, transport fallback, and Go/GOAT behavior are unchanged. No manual effort override was removed.
  • Add two focused regression cases and an Unreleased changelog entry.

The 1.88.3 publication caused the catalog check on #198 to fail while its main CI remained green. This PR does not modify or merge #198, and does not include its type-system changes.

Upstream evidence

Inspected the published package's bundled model reference and CLI model objects without running the CLI or its lifecycle scripts. The archived tarball SHA-512 matches the registry integrity value. The generated diff contains only the version update and the listed additions: images 73→74, reasoning 83→85, effort metadata 67→69, output limits 10→12.

Validation

  • The two new model tests fail against the old 1.88.2 snapshot, then pass after synchronization. They check input modalities, reasoning, exact efforts, published output limits, unchanged default endpoint, and output clamping against a smaller context window.
  • npm test PASS, with real Pi 1.5.0 and OMP 18.16.0 required against local mock APIs (preexisting temporary Bun 1.4.4).
  • Catalog comparison against command-code@latest: current 1.88.3.
  • Public pricing comparison: PASS, 96 models; pricing source unchanged.
  • Project typecheck, formatting, git diff --check: PASS. Active LSP probe of all four changed files: no diagnostics.

Static-validation limit

Main's project typecheck excludes index.ts. A separate expanded probe against real Pi 1.5.0 declarations reports the same 17 inherited diagnostics before and after this catalog change, with byte-identical diagnostic logs. This PR does not claim an exhaustive clean entrypoint typecheck; #198 handles that separately. No diagnostic baseline or suppression is added to CI.

Scope / stop

No independent agent review or packed-artifact smoke is claimed for this catalog PR. No production install/config/auth changes, real login, paid application inference, merge, release, or publication to npm. #159 is already integrated via #180 and is not replayed; #164 remains deferred.

Advertise Crouton image input and the published reasoning efforts and output limits for Crouton and Ling 1019. Preserve pricing, endpoint routing and context-window clamping.

Add focused red/green capability regressions and update the documented snapshot version. Keep entrypoint static validation separate in #198.
@github-actions

github-actions Bot commented Nov 4, 2026

Copy link
Copy Markdown
Contributor

Dependency Review

✅ No vulnerabilities or license issues or OpenSSF Scorecard issues found.

Scanned Files

None

@github-actions

github-actions Bot commented Nov 4, 2026

Copy link
Copy Markdown
Contributor

Memory benchmark

Compared base 52cbc36 with PR head a99eb84 on the same GitHub-hosted darwin runner. Lower values are better.

Metric Base PR PR − Base Change
Stable RSS 107.0 MiB ± 0.7 MiB 108.2 MiB ± 0.8 MiB +0.9 MiB ± 1.1 MiB +0.8%
Physical footprint 61.0 MiB ± 0.6 MiB 61.2 MiB ± 0.1 MiB +0.4 MiB ± 0.6 MiB +0.6%
Physical peak 72.7 MiB ± 1.0 MiB 73.7 MiB ± 0.6 MiB +0.6 MiB ± 0.8 MiB +0.8%

Interpretation: No metric shows a clear Base-to-PR difference beyond its measured run-to-run variation.

Extension overhead above pi baseline

Metric Base overhead PR overhead Difference
Stable RSS +6.0 MiB ± 0.7 MiB +7.2 MiB ± 1.3 MiB +0.9 MiB ± 1.1 MiB
Physical footprint +3.0 MiB ± 0.6 MiB +3.4 MiB ± 0.2 MiB +0.4 MiB ± 0.6 MiB
Physical peak +3.9 MiB ± 1.6 MiB +4.9 MiB ± 1.2 MiB +0.6 MiB ± 0.8 MiB

Values are medians of 6 alternating, paired runs. The value after ± is the median absolute deviation (MAD). Each process was sampled 12 times after a 2500 ms warm-up.

Environment: pi 0.84.4, Bun 1.4.4, darwin arm64. RSS comes from ps; physical footprint and peak come from macOS footprint.

This is a comparative signal, not a pass/fail threshold. GitHub-hosted runner noise can affect absolute values.

@pierreraby
pierreraby merged commit ead11f1 into main Nov 4, 2026
15 checks passed
@pierreraby
pierreraby deleted the chore/catalog-1.88.3 branch November 4, 2026 22:53
pierreraby added a commit that referenced this pull request Nov 4, 2026
Integrate the validated catalog 1.88.3 update from main after #199. Preserve both Unreleased entries, capability regressions and registration metadata tests.

Revalidate the actual-Pi entrypoint contracts, full required Pi/OMP mock suites, catalog, pricing and formatting before publishing this branch update.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant