memory-lancedb-pro: what it is, what problem it solves & why it's gaining traction

What it solves

Most AI agents suffer from "amnesia," forgetting user preferences, past decisions, and project context as soon as a new session begins. memory-lancedb-pro provides a production-grade long-term memory system for OpenClaw agents, allowing them to learn from conversations and recall relevant information across different sessions, agents, and time periods without requiring manual tagging.

How it works

The plugin uses LanceDB as a vector store to create a semantic index of memories. It operates through several key mechanisms:

• Auto-Capture & Extraction: It automatically extracts facts, preferences, and entities from conversations using an LLM-powered 6-category classification system (profiles, preferences, entities, events, cases, and patterns).
• Hybrid Retrieval: To find the right memory, it combines vector search (semantic similarity) with BM25 full-text search (keyword matching), then refines results using cross-encoder reranking.
• Memory Lifecycle: It employs a Weibull decay model to ensure important and frequently accessed memories persist while noise naturally fades away, moving memories between Peripheral, Working, and Core tiers.
• Context Injection: Relevant memories are automatically injected into the agent's prompt before a reply is generated.
• Multi-Scope Isolation: Memories can be isolated by user, agent, project, or global scopes to ensure data privacy and organization.

Who it’s for

Developers using the OpenClaw agent framework who want to give their AI agents persistent, personalized long-term memory and the ability to recall past interactions and project-specific context.

Highlights

• Hybrid Search: Combines semantic vector search and BM25 keyword search for higher recall accuracy.
• Smart Extraction: Automatically categorizes and deduplicates memories using LLMs.
• Intelligent Forgetting: Uses a decay engine to manage memory relevance over time.
• Cross-Encoder Reranking: Integrates with providers like Jina and SiliconFlow to improve retrieval precision.
• Canonical Corpus Integration: Uses local Markdown files as the source of truth while using LanceDB for semantic indexing.
• Production Toolkit: Includes a CLI for exporting, importing, and migrating memories.