| name | testAll |
| description | Run all tests across entire poly-repo org root in parallel. Trigger: /testAll, "run all tests", "test everything", "test all repos", "run tests everywhere". |
testAll
Run deno run -A scripts/test-all.ts from org root. Script finds all
workspaces with *_test.ts / *.test.ts files, runs deno test in each
in parallel, reports results as table.
Workflow
- Run the script:
deno run -A scripts/test-all.ts
- Script outputs a table:
| workspace | passed | failed | duration |
| -------------------- | ------- | ------- | -------- |
| hono-pds | 100 | 0 | 11s |
| deno-worker-sandbox | 24 | 0 | 5s |
| hono-jsr | 5 | 0 | 2s |
| -------------------- | ------- | ------- | -------- |
| total | 129 | 0 | |
-
If any workspace has failures: show the workspace name + count. Offer
to re-run that workspace individually for details.
-
If a workspace has zero test files: skip it (not listed in table).
How it works
- Walks org root (maxdepth 5) finding
*_test.ts / *.test.ts files
- Groups files by nearest parent containing
deno.json (workspace root)
- Runs
deno test --no-check with full permissions in each workspace
- All workspaces run in parallel via
Promise.all
- Parses
deno test stdout for a ok | N passed | N failed or
FAILED | N passed | N failed summary line
Environment setup done automatically
- codebase-rag-proxy embeddings:
retrieval-skalex and
hono-codebase-rag-proxy tests need a real OpenAI-compatible
/v1/embeddings endpoint. The script spawns
scripts/fake-embeddings-server.ts on localhost:18080 (deterministic
hash-based vectors, no real model) before running those two workspaces,
and injects EMBEDDING_URL=http://localhost:18080/v1 for
hono-codebase-rag-proxy (its default points at a LAN-only address).
Server is killed after the run.
- hono-compute-provider container tests: needs the macOS
container
backend running (container system start). The script checks
container system status and prints a warning (not a failure) if it's
down — those tests will show connection-timeout failures, not code bugs,
when the backend is unavailable. It does NOT auto-start the backend or
clean up leftover test containers — if container list shows a pile of
leaked pdr-* VMs from previous runs, stop+rm them by hand
(container stop <id> && container rm <id>) before rerunning, since a
large pile can cause IP/resource exhaustion that looks like test flakiness.
Script flags
<org-root> — override org root as first positional arg (default: cwd)
