Module capability_cache 

Process-local capability cache keyed by provider-qualified model id. Populated when a live Codex catalog is parsed; consulted by validation before falling back to the static table. No credentials stored.

Functions

generation

Current monotonic catalog generation. Every successful cache mutation advances it, allowing UI snapshots to reject stale capability decisions.

get

Look up by provider-qualified id (e.g. "openai-codex/gpt-5.6-sol").

has_provider

True if the cache has any entries for the given provider prefix.

insert

Insert or overwrite an entry keyed by model.runtime_id().

populate

Bulk-populate from an incremental parsed catalog slice.

replace_provider

Atomically replace one provider’s complete catalog snapshot.