You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Commit c31a4ce
Browse filesBrowse the repository at this point in the historyBrowse files
feat(provider)!: replace models.dev with a static model catalog (#40)
Built-in providers and models now come from a git-tracked catalog
(src/provider/catalog.json, one model per line in the Provider.Model
shape). Nothing is fetched, cached or snapshotted at build or run time.
The catalog covers the 18 providers with custom loaders. Any other
provider is defined under `provider` in aictrl.json. The dead `aictrl`
provider (a renamed upstream gateway that never resolved) is removed
with its loader, picker hints, small-model fallback and
gateway-only headers and transforms.
BREAKING CHANGE: providers outside the catalog no longer load from an
env var alone; define them in config. AICTRL_MODELS_URL,
AICTRL_MODELS_PATH, AICTRL_DISABLE_MODELS_FETCH and `models --refresh`
are removed.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Built-in providers and models live in the static catalog at `packages/cli/src/provider/catalog.json` (one model per line, in the `Provider.Model` shape). To add or update a built-in model, edit that file; providers that need custom loading also need an entry in `CUSTOM_LOADERS` in `packages/cli/src/provider/provider.ts`.
30
+
31
+
Any other OpenAI-compatible provider does not need a code change: users define it under `provider` in `aictrl.json`.
Copy file name to clipboardExpand all lines: EVENTS.md
+1-1Lines changed: 1 addition & 1 deletion
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -268,7 +268,7 @@ For compatibility with sessions written before usage provenance was persisted, a
268
268
**`context`** (context-window utilization):
269
269
270
270
-`used` (number) — tokens occupying the model's context window this turn: `input + cache.read + cache.write`.
271
-
-`limit` (number) — the model's total context-window size in tokens, sourced from the models.dev registry (`model.limit.context`).
271
+
-`limit` (number) — the model's total context-window size in tokens, sourced from the provider registry: the built-in model catalog plus `provider` config (`model.limit.context`).
272
272
-`ratio` (number) — `used / limit` (≥0; may exceed 1 if usage exceeds the model's registered limit). A value approaching or exceeding 1 signals context-exhaustion risk.
273
273
-`null` — emitted when the model's context limit is not known (e.g. unregistered custom endpoint), or usage is missing.
0 commit comments