Skip to content

template: catalog-valid model names — 'claude-fable-5-high' is not a model id - #83

Merged
JPDuchesne merged 1 commit into
mainfrom
jpd/model-names-catalog-valid
Sep 13, 2026
Merged

JPDuchesne merged 1 commit into
mainfrom
jpd/model-names-catalog-valid

Conversation

@JPDuchesne

Copy link
Copy Markdown
Contributor

Third live-fire finding of the plans#36 ceremony: with workspace (#81) and credentials (#82) fixed, the agent spawn now fully works — and the run died on model 'claude-fable-5-high' is not in the agent's catalog. The template's default (which cellbound-3d copied) names a model that does not exist: the -high thinking suffix is not part of CLI model names. Fix: claude-fable-5 as the template default, and the per-command example now shows the exact-modelId bracket syntax (claude-fable-5[thinking=true,effort=high]) that resolve_model already matches, which is how a thinking-effort variant is actually pinned. docs/architecture.md example aligned. Companion config fix in cellbound-3d.

Made with Cursor

…model id

The template's default (copied into cellbound-3d) named a model that does
not exist; the '-high' thinking suffix is not part of CLI model names.
First live /build with a fully working agent spawn died on it. Plain
names come from the catalog; a thinking-effort variant is pinned via the
exact modelId bracket syntax, which resolve_model already matches.

Co-authored-by: Cursor <cursoragent@cursor.com>
@codecov

codecov Bot commented Sep 13, 2026

Copy link
Copy Markdown

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

@JPDuchesne
JPDuchesne merged commit 31565ff into main Sep 13, 2026
4 checks passed
@JPDuchesne
JPDuchesne deleted the jpd/model-names-catalog-valid branch September 13, 2026 03:59
@JPDuchesne

Copy link
Copy Markdown
Contributor Author

Correction for the record: claude-fable-5-high was and still is a valid agent --list-models / --model id — past runs used it green. What changed is ai-flow's ACP migration: session/set_model matches the ACP session catalog, which uses plain names + parameterized modelIds — a different naming domain from --list-models. The fix here is still right for the ACP path; the mismatch (job log and template comment point at the wrong catalog) is now tracked in the follow-up issue.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant