--- title: cache_manager | CodeWeaver Docs description: API reference for codeweaver.providers.embedding.cache_manager url: "https://docs.knitli.com/api/providers/embedding/cache_manager" type: static generatedAt: "2026-04-17T17:21:09.030Z" --- # cache_manager [Open in ChatGPT](https://chatgpt.com/?q=Read%20https%3A%2F%2Fdocs.knitli.com%2Fcodeweaver%2Fapi%2Fproviders%2Fembedding%2Fcache_manager%2F.%20I%20want%20to%20ask%20questions%20about%20it.)[Open in Claude](https://claude.ai/new?q=Read%20https%3A%2F%2Fdocs.knitli.com%2Fcodeweaver%2Fapi%2Fproviders%2Fembedding%2Fcache_manager%2F.%20I%20want%20to%20ask%20questions%20about%20it.)[View in Markdown](/codeweaver/api/providers/embedding/cache_manager.md) [Share on LinkedIn](https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fdocs.knitli.com%2Fcodeweaver%2Fapi%2Fproviders%2Fembedding%2Fcache_manager%2F)[Share on X](https://x.com/intent/tweet?url=https%3A%2F%2Fdocs.knitli.com%2Fcodeweaver%2Fapi%2Fproviders%2Fembedding%2Fcache_manager%2F&text=cache_manager)[Share on Threads](https://threads.net/intent/post?url=https%3A%2F%2Fdocs.knitli.com%2Fcodeweaver%2Fapi%2Fproviders%2Fembedding%2Fcache_manager%2F&text=cache_manager)[Share on Bluesky](https://bsky.app/intent/compose?text=cache_manager%20https%3A%2F%2Fdocs.knitli.com%2Fcodeweaver%2Fapi%2Fproviders%2Fembedding%2Fcache_manager%2F)[Share on Facebook](https://www.facebook.com/sharer/sharer.php?u=https%3A%2F%2Fdocs.knitli.com%2Fcodeweaver%2Fapi%2Fproviders%2Fembedding%2Fcache_manager%2F)[Share on Reddit](https://reddit.com/submit?url=https%3A%2F%2Fdocs.knitli.com%2Fcodeweaver%2Fapi%2Fproviders%2Fembedding%2Fcache_manager%2F&title=cache_manager)[Share on Hacker News](https://news.ycombinator.com/submitlink?u=https%3A%2F%2Fdocs.knitli.com%2Fcodeweaver%2Fapi%2Fproviders%2Fembedding%2Fcache_manager%2F&t=cache_manager)[Share on Email](mailto:?subject=cache_manager&body=https%3A%2F%2Fdocs.knitli.com%2Fcodeweaver%2Fapi%2Fproviders%2Fembedding%2Fcache_manager%2F)[Share on WhatsApp](https://wa.me/?text=cache_manager%20https%3A%2F%2Fdocs.knitli.com%2Fcodeweaver%2Fapi%2Fproviders%2Fembedding%2Fcache_manager%2F)[Share on Telegram](https://t.me/share/url?url=https%3A%2F%2Fdocs.knitli.com%2Fcodeweaver%2Fapi%2Fproviders%2Fembedding%2Fcache_manager%2F&text=cache_manager) # `codeweaver.providers.embedding.cache_manager` [Section titled “codeweaver.providers.embedding.cache_manager”](#codeweaverprovidersembeddingcache_manager) Centralized cache manager for embedding providers with namespace isolation. This module provides the EmbeddingCacheManager class which replaces per-instance stores with a centralized, namespace-isolated caching system. This enables true cross-instance deduplication while maintaining async safety. Architecture: - Namespace isolation: {provider_id}.{embedding_kind} prevents collisions - Async-safe locking: Uses asyncio.Lock instead of threading.Lock - Centralized storage: Replaces ClassVar registries with singleton pattern - DI integration: Singleton via FastAPI lifespan and dependency injection Example: >>> cache_manager = EmbeddingCacheManager(registry=get_embedding_registry()) >>> namespace = cache_manager._get_namespace(“voyage-code-2”, “dense”) >>> chunks, hash_map = await cache_manager.deduplicate(chunks, namespace, batch_id) ## Class: `EmbeddingCacheManager` [Section titled “Class: EmbeddingCacheManager”](#class-embeddingcachemanager) Centralized cache manager with namespace isolation for embedding providers. Replaces per-instance _store and _hash_store ClassVar registries with a singleton pattern that provides: - True cross-instance deduplication - Namespace isolation (dense vs sparse, different providers) - Async-safe locking per namespace - Statistics tracking per namespace The cache manager is designed to be injected as a singleton dependency into all embedding provider instances via FastAPI’s lifespan and DI system. Attributes: registry: Global embedding registry for cross-provider coordination _batch_stores: Namespace-isolated UUID stores for batches _hash_stores: Namespace-isolated Blake hash stores for deduplication _locks: Async locks per namespace for thread-safe operations _stats: Statistics per namespace (hits, misses, unique chunks) ### Method: `clear_namespace` [Section titled “Method: clear_namespace”](#method-clear_namespace) ** ``` clear_namespace() ``` Clear all data for a specific namespace. Useful for testing or resetting provider state. Args: namespace: Namespace to clear Warning: This is a destructive operation - all cached data will be lost ### Method: `deduplicate` [Section titled “Method: deduplicate”](#method-deduplicate) ** ``` deduplicate() ``` Deduplicate chunks using namespace-isolated hash store. This method identifies chunks that have already been processed in this namespace and returns only the unique chunks that need embedding. Args: chunks: List of code chunks to deduplicate namespace: Namespace key (e.g., “voyage-code-2.dense”) batch_id: UUID for this batch (used for tracking) Returns: Tuple of (unique_chunks, hash_mapping) where: - unique_chunks: Chunks not seen before in this namespace - hash_mapping: Dict mapping chunk index to its Blake hash Thread Safety: Uses async lock per namespace to ensure thread-safe deduplication ### Method: `get_batch` [Section titled “Method: get_batch”](#method-get_batch) ** ``` get_batch() ``` Get batch by ID from namespace-isolated store. Args: batch_id: UUID of the batch namespace: Namespace key Returns: List of chunks if batch exists, None otherwise ### Method: `get_stats` [Section titled “Method: get_stats”](#method-get_stats) ** ``` get_stats() ``` Get deduplication statistics. Args: namespace: Specific namespace, or None for all namespaces Returns: Statistics dict for namespace(s) ### Method: `register_embeddings` [Section titled “Method: register_embeddings”](#method-register_embeddings) ** ``` register_embeddings() ``` Register embeddings in global registry. Handles updating existing entries or creating new ones in the registry. This is NOT namespace-isolated - the registry maintains embeddings for all chunks regardless of which provider created them. Args: chunk_id: UUID of the chunk embedding_info: Batch information with embeddings chunk: The CodeChunk object to store ### Method: `store_batch` [Section titled “Method: store_batch”](#method-store_batch) ** ``` store_batch() ``` Store batch for potential reprocessing. Stores the final chunk list (after deduplication) for this batch. This enables reprocessing or retrieval of batches later. Args: chunks: Final list of chunks to store batch_id: UUID for this batch namespace: Namespace key Thread Safety: Uses async lock per namespace to ensure thread-safe storage