Supermemory

Supermemory is a memory layer for AI applications with 23k+ GitHub stars — a Pinecone alternative with hybrid vector search, ontology-aware graph, and sub-100ms retrieval for agents and RAG.

Key features

Search capabilities

• Hybrid vector search combining dense and sparse retrieval
• Ontology-aware graph for semantic relationships
• Sub-100ms retrieval latency
• Metadata filtering and faceted search
• Multi-modal support (text, images)

Memory features

• Automatic memory extraction and indexing
• Temporal awareness for time-sensitive queries
• User and session-based memory isolation
• Memory deduplication and consolidation
• Conversation history and context management

Developer experience

• Simple REST API and SDKs
• Integration with OpenAI, LangChain, and LlamaIndex
• Real-time indexing and search
• Webhook support for event-driven workflows
• Dashboard for memory management

Production features

• Scalable infrastructure with automatic sharding
• Replication and high availability
• Usage analytics and monitoring
• Role-based access control
• Self-hosted and cloud options

At a glance

License	MIT
Stack	TypeScript, Rust, PostgreSQL
Self-hosted	Yes — Supermemory OSS
Cloud	supermemory.ai (managed)
Latency	Sub-100ms

Self-hosting

npm install supermemory

Supermemory can be embedded in your application or deployed as a service. The cloud version provides fully managed infrastructure.