🗃️ Serverless storage & retrieval
Object storage
Store text, embeddings, and their metadata with flexible schema support across 8 different index types.
Vector search
k-NN search with cosine, euclidean, dot product, max inner product metrics.
Full-text search
Multi-language analysis (English, Korean, Japanese, standard).
Hybrid search
Combines vector + lexical with score normalization (RRF, Min-Max, L2).
Multi-field vector search
Search across multiple vector fields simultaneously.
Sparse vector search
Efficient high-dimensional sparse vector storage and search.
🏢 Enterprise features
Backup & recovery
- Continuous backups: Automatic collection-level backups with 30 days retention by default.
- Point-in-time recovery (PITR): Restore a collection to any specific moment within retention period.
- Zero-copy collection clone: Instant collection cloning without data duplication.
Operation & management
- Serverless-native architecture: Zero infrastructure management required without paying for idle resources.
- Configurable rate limiting: Project-level controls for API usage.
- RESTful API: Simple HTTP-based interface with comprehensive SDKs.