Commit Graph
4 Commits
Author SHA1 Message Date
alexsong-oai 75db67bbc6 Enforce current provider requirements for the model catalog (#46917)
## Why

The app server retains its startup model provider, but managed provider requirements can change while it is running. Catalog reads and background refreshes must check those requirements even when models are cached.

## What changed

- Gate `model/list` and background refreshes through a shared catalog that validates the startup provider against current requirements.
- Return JSON-RPC error `-32600` with restart guidance when the provider no longer complies, and skip background requests to that endpoint. Requirement load failures also block catalog access.
- Log only the error kind for blocked refreshes to avoid exposing credentials in parser diagnostics.

## Testing

Add coverage for cached catalog access after provider selection or definition changes, invalid requirements, cloud configuration load failures, and recovery when requirements allow the retained provider again. Extend refresh tests to cover startup provider overrides, blocked endpoint requests, and credential-safe logging.

GitOrigin-RevId: 03127bba46f1f59f2a8b1dc37391eab75f9be23c
2026-09-21 01:19:40 +00:00
spasumar-oai 4ef1d4b89b Increase the app-server model refresh interval (#40498)
## What changed

Change the background model refresh interval from 3 minutes to 4 minutes 30
seconds.

GitOrigin-RevId: dcefe471c91dc94a04ddc74595c3ce00d9061e55
2026-08-24 23:29:10 +00:00
Michael Bolin 5892c7b69d model-provider: route model discovery through HTTP client factory (#31361) 2026-07-09 02:32:10 +00:00
jif 5935a90619 app-server: keep the model cache warm (#28699)
## Why

The app server is long-lived, but its shared model cache otherwise
refreshes only when a caller needs it. Once the five-minute cache
expires, starting a thread or calling `model/list` can wait for
`/models` on the request path.

Refresh the cache in the background before it expires so foreground
callers normally use fresh local state.

## What changed

- Start an app-server worker that refreshes models immediately and then
every three minutes using the existing models-manager API.
- Hold only a weak reference to the models manager between refreshes, so
the worker does not extend its lifetime.
- Stop scheduling refreshes when the app-server lifecycle handle is shut
down or dropped. A refresh already in progress is allowed to finish.
- Adjust affected app-server test fixtures to distinguish the background
`/models` probe from the connection they are testing.

The existing models-manager cache, refresh strategies, auth handling,
ETag behavior, and concurrency semantics are unchanged.

## Testing

-
`models_refresh_worker::tests::refreshes_immediately_periodically_and_stops_when_dropped`
-
`suite::v2::remote_control::listen_off_honors_persisted_remote_control_enable`
-
`suite::v2::attestation::attestation_generate_round_trip_adds_header_to_responses_websocket_handshake`
2026-06-17 16:18:39 +02:00