| Filename | Latest commit message | Latest commit date |
|---|---|---|
|
Some checks failed
no-vanity-metrics / Scan committed Markdown for vanity metrics (push) Successful in 3s
no-vanity-metrics / WAI public text names no comparison (push) Successful in 3s
no-vanity-metrics / WAI public text names no company or product (push) Successful in 3s
standards workspaces build + test / eoc-rs (push) Successful in 1m51s
standards workspaces build + test / jcp-rs (push) Successful in 1m22s
standards workspaces build + test / proof-rs (push) Successful in 1m16s
standards workspaces build + test / sandbox-rs (push) Successful in 2m12s
standards workspaces build + test / smart-byte-rs (push) Successful in 3m37s
standards workspaces build + test / jouleclaw-rs (push) Has been cancelled
standards workspaces build + test / aac-rs (push) Has been cancelled
standards workspaces build + test / arl-rs (push) Has been cancelled
standards workspaces build + test / joule-code-rs (push) Has been cancelled
standards workspaces build + test / joule-ui-rs (push) Has been cancelled
standards workspaces build + test / joulecontract-rs (push) Has been cancelled
standards workspaces build + test / jouledesk-rs (push) Has been cancelled
standards workspaces build + test / joulehook-rs (push) Has been cancelled
standards workspaces build + test / map-rs (push) Has been cancelled
standards workspaces build + test / mesh-rs (push) Has been cancelled
standards workspaces build + test / openpay (push) Has been cancelled
standards workspaces build + test / sense-rs (push) Has been cancelled
standards workspaces build + test / macOS energy backend (jcp-energy) (push) Has been cancelled
What review found in the two-model change, before release:
- Replacing E2B with E4B removed E2B even when E4B then could not run
(out of memory), leaving no helper in any language. The model that
worked is now kept beside the new one until the new one has run here;
if it cannot, the one that worked is put back and the new one's parts
go ("Gemma 4 E4B could not run on this device, so Gemma 4 E2B is kept").
If the page dies during the check, the one that worked is in use when
it opens again, and Settings can set the new one up without downloading.
- install() takes the model Settings described, so Continue on a partial
download, or a language switch while Settings reads, cannot download
another model than the size the person agreed to.
- The "larger model" view shows what is left of a stopped download, with
Continue and "Remove what was downloaded"; its space check counts only
what is left. A stopped download of a model not in use is shown and can
be removed, and a finished install deletes every part it does not use.
- Settings clears what it showed when the language changes.
- Where the browser does not say how much memory the device has (Safari),
a model needing more than 4 GB says so before download; too little
memory names the model the language needs.
- A model the catalogue no longer lists is offered as an update.
- run.mjs needs MODEL= when a site has two models; build_ai.py refuses a
published build missing a model a ready language uses, and checks
--source ids.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
||
| .. | ||
| items.json | ||
| judge.py | ||
| merge_languages.py | ||
| results-2026-10-01-languages.json | ||
| results-2026-10-01.json | ||
| run.mjs | ||