Resolves F16 in docker-build-optimization-roadmap v5.
Root cause:
publish-outdated-packages.sh uses a pack-extract-repack pattern:
1. pnpm pack (rewrites workspace:* in tarball)
2. extract
3. npm pack (re-tar from extracted content)
4. npm publish
Step 3 is the bug. npm pack does not recognize the pnpm-specific
workspace: protocol — it treats workspace:* as a literal version
string and passes it through to the final tarball. Result: any
consumer doing 'pnpm install' inside Docker (where there is no
workspace context) fails with ERR_PNPM_WORKSPACE_PKG_NOT_FOUND.
Documented in roadmap §0 F16 + §3 Phase A-pre.
Fix (publish-outdated-packages.sh):
- Insert a workspace:* rewriter between publishConfig strip and
npm pack. Reads source package.json for each @bytelyst/* target,
resolves workspace:* / workspace:^ / workspace:~ to ^x.y.z.
- Add defense-in-depth: grep the post-rewrite package.json for any
surviving 'workspace:' literal. If found, refuse to publish.
Republished 10 affected packages with workspace:* → resolved semver:
@bytelyst/auth 0.1.5 → 0.1.6
@bytelyst/diagnostics-client 0.1.6 → 0.1.7
@bytelyst/events 0.1.5 → 0.1.6
@bytelyst/extraction 0.1.5 → 0.1.6
@bytelyst/fastify-auth 0.1.5 → 0.1.6
@bytelyst/fastify-core 0.1.5 → 0.1.6
@bytelyst/feedback-client 0.1.6 → 0.1.7
@bytelyst/field-encrypt 0.1.6 → 0.1.7
@bytelyst/react-auth 0.1.6 → 0.1.7
@bytelyst/sync 0.1.5 → 0.1.6
Verification: all 10 packages now scan with 0 workspace:* refs in
their published package.json (per registry curl scan).
Unblocks: A0-V verification on learning_ai_clock (currently blocked
at learning_ai_clock@0be887288).
Published 60 @bytelyst/* packages to local Gitea npm registry.
create-app skipped (private: true — internal scaffolding tool).
Token regenerated with full write:package scope to fix E401.
BUG 1: Azure locale derivation produced 'en-EN' (invalid) for 2-letter codes.
→ Added toAzureLocale() with 28-language mapping table (en→en-US, pt→pt-BR, etc.)
→ Exported for testing; falls back to code-CODE for unmapped languages.
BUG 2: model field from request schema was silently dropped after provider refactor.
→ Added optional model field to TranscriptionInput interface.
→ OpenAI provider now uses input.model override (falls back to config.model).
→ Route passes model through to provider.transcribe().
GAP 4: SUPPORTED_AUDIO_TYPES was defined but never validated against.
→ Route now rejects unsupported content-types with a clear error message.
→ Allows application/octet-stream (Azure Blob SAS URLs often return this).
GAP 5: Client JSDoc still said 'via OpenAI Whisper API' — now 'via configured STT provider'.
GAP 8: Azure WAV content-type hardcoded samplerate=16000 — now generic audio/wav.
Tests: 42 transcription tests (was 35), 178 total passing.
→ toAzureLocale: 4 tests (locale mapping, passthrough, fallback, case-insensitive)
→ setSTT: 1 test (singleton override)
→ model passthrough: 2 tests (mock ignores, input accepts)
Root cause: tinypool worker teardown calls kill() which returns EPERM
in the act_runner host environment on Node.js v25.2.1. Tests pass but
the vitest process crashes during cleanup, causing CI failure.
Fix: --pool forks CLI flag on every package/service test script, plus
pool: 'forks' in all vitest.config.ts files. This uses child_process.fork()
worker management which handles termination cleanly.
60 package.json files updated, 10 vitest.config.ts files updated.
- Created .npmrc with @bytelyst scoped registry pointing to local Gitea
- Added publishConfig.registry to all 57 @bytelyst/* package.json files
- Created scripts/harden-publish-config.sh for future re-runs
- Prevents accidental publish to npmjs.org or corporate JFrog registry
