Struct RegistryStore 

pub struct RegistryStore { /* private fields */ }

SQLite-backed registry store.

Thread-safe via internal Mutex<Connection> — registry operations are short-lived and infrequent so contention is negligible.

Implementations

impl RegistryStore

pub fn open(data_dir: &Path) -> Result<Self>

Opens (or creates) the registry database at <data_dir>/registry.db.

On first run (empty database), seeds the cache from the embedded registry snapshot so the provider catalog is immediately available without a network fetch.

If an existing database is unreadable (corrupt file, truncated WAL/SHM, or SQLite disk I/O errors on open), the broken files are removed and a fresh cache is recreated from the embedded snapshot. Callers that need the remote catalog should follow up with sync_registry.

pub fn meta_get(&self, key: &str) -> Result<Option<String>>

Returns the value of a metadata key, or None.

pub fn meta_set(&self, key: &str, value: &str) -> Result<()>

Sets a metadata key-value pair (upsert).

pub fn is_empty(&self) -> Result<bool>

Returns true if the providers table is empty.

pub fn provider_count(&self) -> Result<usize>

Returns the number of providers in the cache.

pub fn model_count(&self) -> Result<usize>

Returns the total number of models across all providers.

pub fn provider_sha256(&self, provider_id: &str) -> Result<Option<String>>

Returns the stored SHA-256 hash for a provider, or None.

pub fn provider_ids(&self) -> Result<HashSet<String>>

Returns the set of provider ids currently in the cache.

pub fn load_provider_models( &self, provider_id: &str, ) -> Result<HashMap<String, RegistryModel>>

Loads existing models for a provider from the cache, keyed by model name. Used by aggregator sync to preserve metadata (context_window, etc) for models that the API returns without rich metadata.

pub fn delete_providers_not_in(&self, keep: &HashSet<&str>) -> Result<()>

Deletes providers that are not in the given set of ids. Used during sync to remove providers that were deleted from the remote registry.

pub fn upsert_provider(&self, provider: &RegistryProvider) -> Result<()>

Upserts a provider and its models from a RegistryProvider.

pub fn upsert_transcription_provider( &self, provider: &RegistryTranscriptionProvider, sha256: Option<&str>, ) -> Result<()>

Upserts a remote transcription provider (stored as JSON for simplicity).

pub fn transcription_provider_sha256(&self, id: &str) -> Result<Option<String>>

SHA-256 of a cached transcription provider, if known.

pub fn load_transcription_providers( &self, ) -> Result<Vec<RegistryTranscriptionProvider>>

Loads all cached transcription providers.

pub fn delete_transcription_providers_not_in( &self, keep: &HashSet<&str>, ) -> Result<()>

Deletes transcription providers not in keep.

pub fn upsert_canonical_model( &self, id: &str, model: &CanonicalModel, sha256: Option<&str>, ) -> Result<()>

Upserts a canonical model JSON blob with its integrity hash.

pub fn canonical_model_sha256(&self, id: &str) -> Result<Option<String>>

Returns the SHA-256 of a cached canonical model, if any.

pub fn load_canonical_model_catalog(&self) -> Result<ModelCatalog>

Loads the full canonical model catalog from the cache.

pub fn delete_canonical_models_not_in(&self, keep: &HashSet<&str>) -> Result<()>

Deletes canonical models not present in keep.

pub fn canonical_model_count(&self) -> Result<usize>

Number of cached canonical models.

pub fn transcription_provider_count(&self) -> Result<usize>

Number of cached transcription providers.

pub fn seed_transcription_from_embedded_if_empty(&self) -> Result<()>

Seeds the transcription cache from the embedded snapshot when empty.

pub fn upsert_provider_union_models( &self, provider: &RegistryProvider, sha256: Option<&str>, ) -> Result<()>

Upserts a provider while preserving any local models that are not in provider.models.

Used by embedded/remote registry updates so a smaller catalog snapshot cannot wipe models discovered via sync models / provider /models. Incoming models win on name conflicts (refresh metadata); local-only names are kept.

pub fn rehydrate_provider_models_from_catalog(&self) -> Result<usize>

Re-apply canonical model catalog metadata onto every cached provider model.

Fixes stale rows written by API sync sibling inheritance (e.g. grok-4.5 wrongly holding grok-4.3’s 1M context / low+high efforts) without wiping API-discovered model names or changing provider sha256 markers.

Returns the number of model rows touched by the UPDATE.

pub fn upsert_provider_with_sha256( &self, provider: &RegistryProvider, sha256: Option<&str>, ) -> Result<()>

Upserts a provider with its SHA-256 hash for diff-based sync.

pub fn load_all_providers(&self) -> Result<Vec<ProviderConfig>>

Loads all providers from the cache as ProviderConfig values compatible with the existing catalog.

pub fn replace_all(&self, providers: &[RegistryProvider]) -> Result<()>

Replaces the entire cache with the given providers (full refresh).

pub fn save_manifest_meta(&self, manifest: &RegistryManifest) -> Result<()>

Saves the manifest metadata.

pub fn manifest_version(&self) -> Result<Option<u32>>

Returns the stored manifest version, if any.

pub fn manifest_updated_at(&self) -> Result<Option<String>>

Returns the stored manifest updated_at, if any.

pub fn upsert_capabilities( &self, model_id: &str, provider_id: &str, capabilities: &[(String, String)], ) -> Result<()>

Upserts capabilities for a model. Replaces all existing capabilities for that model.

pub fn load_capabilities(&self, model_id: &str) -> Result<Vec<ModelCapability>>

Loads all capabilities for a model.

pub fn upsert_pricing( &self, model_id: &str, provider_id: &str, input_price: Option<f64>, output_price: Option<f64>, ) -> Result<()>

Upserts pricing for a model.

pub fn load_pricing(&self, model_id: &str) -> Result<Option<ModelPricing>>

Loads pricing for a model.

pub fn upsert_profile(&self, profile: &Profile) -> Result<()>

Upserts a profile definition.

pub fn upsert_model_profile( &self, model_id: &str, provider_id: &str, profile_id: &str, score: f64, ) -> Result<()>

Upserts a model-profile association.

pub fn query_models_by_profile( &self, profile_id: &str, ) -> Result<Vec<RankedModel>>

Queries for models matching a profile, ranked by score and price.

Returns models that satisfy the profile’s constraints (min context, max price, tool support) ordered by score descending, then input price ascending.

pub fn seed_default_profiles(&self) -> Result<()>

Seeds the default built-in profile definitions.

pub fn delete_provider_metadata(&self, provider_id: &str) -> Result<()>

Deletes all capabilities, pricing, and model-profile entries for a provider.