The Free Inference page says what's running, and every limit in numbers (#370) #372

Merged
cmoriarty merged 2 commits from claude/free-inference-page into main 2026-11-03 15:07:04 -05:00
Owner

The Free Inference page, redesigned and then reworked from the owner's review (#370). The owner has seen it and approved this merge.

From the top:

  • Now serving: the models that last answered a run's steps and whether Braid has their keys, then what the run has done as a whole, then its steps. The figures are never over a window of time: steps, run time, requests, tokens in and out, models, and waits and refusals when there are any.
  • Charts: moved up under that, and each shows only if it has data in the chosen window. Only on Free.
  • Free quota: a row per tier, collapsed, with a dot per model and how many are ready. Each opens to its chain.
  • Models: a grid with used/limit over a bar for each limit, the model's last hour in columns, and its key. The words are only in the headers.

Backend:

  • GET /api/slots gains free_runs.
  • GET /api/free/quota gains each model's key, provider_name and recent.
  • The ledger's Use keeps tokens out, duration and whether a request was served, so the last hour survives a restart.

Checked:

  • ./tools/test.sh passes, including a new browser journey on a --free test server: the real service, with no router behind it.
  • Every Free sim passes.
  • Looked at in a browser at desktop and phone width, on seeded data. It'll be checked on production's data once it deploys, since production's API has none of the new fields until then.

🤖 Generated with Claude Code

The Free Inference page, redesigned and then reworked from the owner's review (#370). The owner has seen it and approved this merge. **From the top:** - **Now serving:** the models that last answered a run's steps and whether Braid has their keys, then what the run has done as a whole, then its steps. The figures are never over a window of time: steps, run time, requests, tokens in and out, models, and waits and refusals when there are any. - **Charts:** moved up under that, and each shows only if it has data in the chosen window. Only on Free. - **Free quota:** a row per tier, collapsed, with a dot per model and how many are ready. Each opens to its chain. - **Models:** a grid with `used/limit` over a bar for each limit, the model's last hour in columns, and its key. The words are only in the headers. **Backend:** - `GET /api/slots` gains `free_runs`. - `GET /api/free/quota` gains each model's `key`, `provider_name` and `recent`. - The ledger's `Use` keeps tokens out, duration and whether a request was served, so the last hour survives a restart. **Checked:** - `./tools/test.sh` passes, including a new browser journey on a `--free` test server: the real service, with no router behind it. - Every Free sim passes. - Looked at in a browser at desktop and phone width, on seeded data. It'll be checked on production's data once it deploys, since production's API has none of the new fields until then. 🤖 Generated with [Claude Code](https://claude.com/claude-code)
The page was mostly sentences. It said nothing of which model was running, and its one clear figure
was a percentage left, of nothing it named. Now, from the top:

- The models that last answered a run's steps, whether Braid has their keys, what the run has done as
  a whole (steps, run time, requests, tokens, models, and any waits and refusals), and its steps.
  The figures come from the new `free_runs` on GET /api/slots.
- The charts, each only if it has data in the chosen window.
- A row per tier, collapsed, with a dot per model and how many are ready. It opens to its chain.
- A table of every model as a matrix: `used/limit` over a bar for each limit, its last hour, and its
  key, with the words in the headers.

The page's data comes from the ledger: `Use` keeps tokens out, duration and whether a request was
served, from `free_call` at boot and live after. `--free` gives its browser journey a server of its
own, the real service with no router behind it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
chore(openspec): archive free-inference-page
All checks were successful
test / suite (pull_request) Successful in 42s
9b836491a8
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
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!372
No description provided.