Browser tests on a server each, the Free cell names the model serving, finished runs slimmed (#461, #364, #192) #468

Merged
cmoriarty merged 1 commit from batch-2026-11-09b into main 2026-11-09 15:32:48 -05:00
Owner

Three issues in one change (openspec/changes/batch-2026-11-09b), the owner's batch of 2026-11-09.

braid#461, browser tests. Every test starts and stops its own server and database (ui/e2e/fixtures.ts); the run-tabs journey asserts on its two runs alone, and a serial pair shows the second test on a worker finds an empty console. Checks on the laptop: the suite 20 times with one worker in shuffled file orders, all green (53 s each); console.spec.ts:22 50 of 50 with four workers. docs/testing.md: 49 tests, 20 s with four workers.

braid#364, the Free cell. The top bar names the model serving now from the slot holders: hetzner/Qwen3.8-27B +1 more, each model's step count in the hover; waiting for a model, none serving; Paused until … kept for spent quota. No percentage on the bar; the shares stay on Inference. The backend cell's hover says its count spans every run. Unit tests in free.test.ts, a browser test on free_serving and a new free_paused fixture.

braid#192, slim finished runs. The finished-run tidy also deletes the environments .osf/deps.json records (node_modules, .venv), only inside the run and past unsafe_to_delete. A re-opened run whose recorded environments are missing is held while the scheduler runs deps install --changed in the working copy with the step's own environment; --changed counts a missing environment as changed. Sim: slimmed-run-resumes stops a quick-fix run on hello-venv, resumes it after: slimmed, and merges with backend/.venv back and script.dependencies run once. Its osfd log: tidied … dependencies 32955, installing backend/.venv again before its steps run, environments installed again in 2 s.

./tools/test.sh green in 46 s; openspec validate --strict passes.

🤖 Generated with Claude Code

Three issues in one change (`openspec/changes/batch-2026-11-09b`), the owner's batch of 2026-11-09. **braid#461, browser tests.** Every test starts and stops its own server and database (`ui/e2e/fixtures.ts`); the run-tabs journey asserts on its two runs alone, and a serial pair shows the second test on a worker finds an empty console. Checks on the laptop: the suite 20 times with one worker in shuffled file orders, all green (53 s each); `console.spec.ts:22` 50 of 50 with four workers. `docs/testing.md`: 49 tests, 20 s with four workers. **braid#364, the Free cell.** The top bar names the model serving now from the slot holders: `hetzner/Qwen3.8-27B +1 more`, each model's step count in the hover; `waiting for a model`, `none serving`; `Paused until …` kept for spent quota. No percentage on the bar; the shares stay on Inference. The backend cell's hover says its count spans every run. Unit tests in `free.test.ts`, a browser test on `free_serving` and a new `free_paused` fixture. **braid#192, slim finished runs.** The finished-run tidy also deletes the environments `.osf/deps.json` records (`node_modules`, `.venv`), only inside the run and past `unsafe_to_delete`. A re-opened run whose recorded environments are missing is held while the scheduler runs `deps install --changed` in the working copy with the step's own environment; `--changed` counts a missing environment as changed. Sim: `slimmed-run-resumes` stops a quick-fix run on hello-venv, resumes it `after: slimmed`, and merges with `backend/.venv` back and `script.dependencies` run once. Its osfd log: `tidied … dependencies 32955`, `installing backend/.venv again before its steps run`, `environments installed again in 2 s`. `./tools/test.sh` green in 46 s; `openspec validate --strict` passes. 🤖 Generated with [Claude Code](https://claude.com/claude-code)
Every browser test starts and stops its own server and database, and the run-tabs journey asserts
on its two runs alone; a serial pair shows the second test on a worker finds an empty console.
docs/testing.md carries the figures: 49 tests, 20 s with four workers, 53 s with one.

The top bar's Free cell reads the slot holders and names the model serving now, hetzner/Qwen3.8-27B
+1 more, with each model's step count in the hover; waiting for a model, none serving, and Paused
until … for spent quota. No percentage on the bar. The backend cell's count says it spans every run.

The finished-run tidy also deletes the environments .osf/deps.json records (node_modules, .venv),
inside the run only. A re-opened run whose recorded environments are missing is held while the
scheduler runs the dependencies step's --changed pass in the working copy with the step's own
environment; the pass counts a missing environment as changed. The sim's resume takes
after: slimmed, and slimmed-run-resumes stops a quick-fix run, waits for the slim and resumes it
to a merge with its environment back.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
cmoriarty deleted branch batch-2026-11-09b 2026-11-09 15:32:48 -05:00
Sign in to join this conversation.
No reviewers
No labels
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
cmoriarty/braid!468
No description provided.