memory-qdrant

You can store and recall local semantic memory using a vector-based store like Qdrant or an in-memory mode. It uses Transformers.js to generate embeddings locally, enabling context-aware recall without external services.

No, you do not need to configure Qdrant for basic local semantic memory. The system supports a zero-configuration in-memory mode for offline operation, while persistent storage with Qdrant remains optional for scaling contexts.

Semantic search over stored memories works by generating local text embeddings via Transformers.js. These embeddings are stored in a vector database, allowing agents to retrieve relevant historical context based on query similarity.

Yes, you can automatically capture conversations for context-aware recall. The system supports optional auto-capture of conversations, enabling proactive AI behavior by proactively retrieving relevant memory context.

You can manage the memory lifecycle and delete stored embeddings using the memory_forget tool for safe deletion. In memory mode, the system also supports optional LRU eviction to automatically manage storage capacity.

The best way to scale semantic memory from personal to team contexts is by transitioning from in-memory mode to persistent storage with Qdrant. This scales vector storage while maintaining local embedding generation and categorization.