test(registry): qualify CLI and browser catalog parity - #284
lloydrichards wants to merge 1 commit into
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
|
0ec8b1a to
cc4b496
Compare
cc4b496 to
2bd670a
Compare
2bd670a to
3b9650f
Compare
There was a problem hiding this comment.
Important
The new qualification tests can pass without exercising real CLI generation or a file-changing catalog update, leaving two acceptance criteria from #275 unverified.
Reviewed changes at 3b9650f: CLI/browser parity harness, compatible-update and session tests, session ID allocation, and registry user/agent documentation.
- Parity harness: A Bun script loads a controlled catalog and compares its service-level preview with the browser worker for Bun, Node, and Deno.
- Update checks: CLI schema output and browser catalog metadata are checked across compatible revisions.
- Guidance and sessions: Documentation explains registry/cache behavior, and builder session IDs become unique across hook instances.
GPT Sol | 𝕏
3b9650f to
481bbd7
Compare
There was a problem hiding this comment.
✅ No new issues found.
Reviewed changes since the prior Pullfrog review: the latest fixes exercise CLI file generation and verify catalog revisions in generated output.
- Ran real CLI commands: Compared browser previews with files produced by controlled CLI
createand incrementaladdfor the selected runtimes. - Verified revised contributions: Checked that an existing browser session keeps its files while a later session and CLI command observe changed file contents.
- Aligned generated configuration: Updated CLI configuration formatting for parity with browser output.
GPT Sol | 𝕏
There was a problem hiding this comment.
✅ No new issues found.
Reviewed changes in the full PR diff: the controlled CLI/browser parity harness, compatible catalog revisions, builder session IDs, and registry guidance.
- CLI/browser parity: Runs CLI
createfor Bun, Node, and Deno and incrementaladd, comparing generated files and Blueprint with worker previews. - Catalog updates: Checks that new CLI commands and browser sessions receive changed file contents while an existing browser session remains stable.
- Guidance and session handling: Allocates unique builder session IDs and documents registry loading, cache fallback, and schema metadata.
The scoped CLI tests and Chromium browser tests pass. Deployed registry HTTP qualification remains a separate release gate tracked by #275.
GPT Sol | 𝕏

Goals/Scope
Qualify the five registry implementation PRs against #275. Review the controlled CLI/browser comparison and the remaining release evidence.
Description
Controlled catalog responses drive actual CLI
createcommands and browser previews for Bun, Node, and Deno. Tests compare generated files and the resulting Blueprint. An incremental CLIaddtest compares the newly added target's files with browser output. A changed generated file appears in the next command and browser session while the current session stays stable. User guidance, agent guidance, and generated CLI reference text explain cache behavior and$schemametadata.How to Test
Run
bun format,bun lint,bun run type-check,bun run okf:check, scoped CLI and docs tests, and production CLI and docs builds. CI also runs generated-project E2E. The parent issue record separates local, CI, and deployed evidence.Comments
Tracks #275. Keep this PR draft until #272's public URLs return JSON and pass conditional-request, CORS, missing-path, and no-cache timing checks. They currently return HTTP 404; the preview requires Vercel SSO. Do not close #275 from local tests.