The actual starting point
The old repository mixed early RAG scaffolding with a later webhook/concurrency demonstration. Surviving original history is:
| Commit | Actual historical purpose |
|---|---|
431b89e | Initial project |
01e0a92 | Render binding/missing-key repair |
acb9d80 | Rewrite toward a webhook dispatcher UI |
52b6258 | SSE and unit-test work |
The later reviewed working tree had already repaired bounded admission, worker accounting, shutdown and slow-stream behavior, while making clear that task delivery was simulated. It was preserved as checkpoint ce80475 before this cleanup. That checkpoint records pre-existing work, not newly invented development.
Cleanup performed on 16 October
The application/library/local directory were renamed Batchlark. queue-lab remains a compatibility link. Unused RAG modules were archived privately and removed from the active module along with the OpenAI dependency. The generated executable is ignored. Local binding defaults to loopback; container binding is explicit.
The cleanup source is 3337ef8 . This is a retained engineering study, not a paid SaaS, production dispatcher or new hosted service. The old public repository URL remains a legacy identity pending the replacement decision.
Successful checks
go test -race ./...
go vet ./...
go build -o /tmp/batchlark .
The local review used Go 1.26.5. The active served module has no third-party dependency. Docker was unavailable for executing the retained recipe; no image or cloud deployment is claimed.
Why stop expanding it
A bounded pool is a useful mechanism to understand, but by itself offers little practical differentiation. Existing Asynq/Asynqmon, River, Temporal and Modal job processing already cover real queue/workflow use cases. Adding a name or simulated dashboard does not establish a product.
An extra observatory expansion was started, then parked after the user challenged the purpose. The prior working baseline was restored before cleanup. The unused expansion is not presented as shipped behavior. A science-oriented Go replacement remains a proposal; no biology/AI feature is claimed here.
Git recovery and a truthful new record
A private full working-tree/Git backup and verified full-history bundle preserve the old record. A local proposal/batchlark-clean-history branch contains a one-commit root with the same reviewed study tree. Its timestamp is the actual creation time; nothing was backdated. It was not force-pushed over published history.
If a genuinely new science application is selected, a fresh repository with a real initial implementation is the clearest starting point. The old study can remain archived instead of carrying unrelated RAG/webhook history into a new product. If this same repository is retained, choose the exact retained baseline and update affected source links before a reviewed history replacement. Deleting ancestry alone does not rename a repository or erase copies held elsewhere.
Actual startup failure found after the initial cleanup
A real local launch exposed a Go ServeMux conflict between the broad GET / root and method-unqualified /api/v1/dispatch. Existing handler tests had never constructed the full router. The correction uses exact GET /{$} root matching and a shared router constructor. A regression test constructs that actual router and checks home, health, metrics, dispatch, wrong method and missing path.
Race tests, vet and the build then passed; an actual local server admitted and completed five tasks. This is why compilation and handler tests alone were insufficient evidence of a working application. The corrected source is the startup-fix revision . The proposed science replacement still requires approval; this fix does not implement it.
Actual Chromium checks also passed at 320/390/768/1440: renamed UI, a submitted five-task batch, completed progress and no page errors/overflow. Source CI 38887493535
passed for bed2cb3. No new hosted Go service or replacement application is implied.