fix(cli): a thin build kit says what to try instead of looking empty - #5938
fix(cli): a thin build kit says what to try instead of looking empty#5938josephfarina wants to merge 2 commits into
Conversation
build "quantum flux capacitor telemetry" returned one incidental component and the always-on frame list. An agent reading that concludes the package has nothing and falls back on its own memory of Astryx, which is the failure build exists to prevent. Below three offerable results the kit carries a hint naming the browse commands and saying this is keyword matching, not semantic. The threshold counts what survived the score floors, not what search returned: hasResults is already true for a query that matched and was then filtered to nothing. Absent when it does not apply, so a healthy kit is byte-identical. Rendered last, as FEW MATCHES, so it is the line the reader leaves with. Split out of #5320 at review request.
meta-cla
Bot
added
the
CLA Signed
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
Review asked for the guidance indexing and the thin-kit hint as separate small PRs. The hint is now #5938 and the guidance tier is #5937, so this branch keeps only the coverage gate on the build pages group and the matchedTerms/queryTerms plumbing that gate needs. No behaviour change to the gate itself.
PR Analysis Report📚 Storybook PreviewView Storybook for this PR 🧪 Sandbox PreviewView Sandbox for this PR No new or modified components detected. Bundle Size SummaryNo component packages changed. Accessibility AuditStatus: No accessibility violations detected. Generated by PR Enrichment workflow | Storybook | Sandbox | View full report |
…h too The split removed `hint` from the kit and its type but left build.mjs destructuring it, so Typecheck CLI (strict) failed with TS2339 and took build-storybook and build down with it. The renderer change belongs to #5938 with the rest of the hint.
cixzhang
left a comment
cixzhang
left a comment
There was a problem hiding this comment.
Builders without a global astryx binary get unusable recovery commands: this pnpm workspace renders every other build-kit command as pnpm exec astryx, but FEW MATCHES freezes bare astryx commands into the API prose. Please keep recovery commands structured/package-manager-agnostic and format them in the CLI renderer.
The output also says its complete section order ends at FRAME + FOUNDATION, then emits FEW MATCHES.
Finally, the “matched then filtered” test returns six offerable results and takes the no-hint branch. Please assert a deterministic hasResults: true, zero-offerable case.
[Reviewed by Robohands]
Three review findings from cixzhang, all real.
The hint froze `astryx component --list` into API prose. This pnpm workspace
renders every other build-kit command as `pnpm exec astryx`, so the one command
offered to a stuck reader was the one that would not run. hint is now
{reason, commands} with bare subcommands, and the renderer formats them through
formatCliCommand.
The legend promised a section order ending at FRAME + FOUNDATION and then
emitted FEW MATCHES after it. FEW MATCHES is now in the order when present.
The matched-then-filtered test returned six offerable results and took the
no-hint branch, so it asserted nothing. Replaced with `blockchain`, which is a
deterministic hasResults:true, zero-offerable case, plus a test pinning the
commands as bare.
2 participants
Split out of #5320 at review request: the thin-kit hint, on its own, with red→green evidence.
The gap
astryx build "quantum flux capacitor telemetry"— before:One incidental component, the always-on list, and nothing else. An agent reading that does not conclude its wording was wrong — it concludes the package has nothing and falls back on its own memory of what Astryx contains. That is precisely the failure
buildexists to prevent.After, same query, one added section:
The threshold counts what survived, not what search returned
Below three offerable results (pages + blocks + domain). Deliberately not
hasResults, which is already true for a query that matched things and then had them all filtered out by the score floors — and that is the case most likely to be misread, because the output looks populated. There is a test for exactly that shape.hintis absent when it does not apply, so a healthy kit is byte-identical to before. The CLI renders it last so it is the line the reader leaves with.Also
BuildKitResponsetype and the public response doc (build.doc.mjs) both updated.Gates:
check:repogreen,lint:strict0 errors, build suite 8/8.