memory-lancedb-pro
memory-lancedb-pro is an OpenClaw plugin that provides long-term “memory” storage and hybrid retrieval on top of LanceDB. It combines vector (ANN) search with BM25 full-text search, fuses results, applies cross-encoder reranking, and runs a multi-stage scoring pipeline including recency/importance/length normalization/time decay plus diversity (MMR). It also supports multi-scope isolation, auto-capture/auto-recall hooks in the OpenClaw lifecycle, optional Markdown mirroring, long-context embedding chunking, and a management CLI for memory CRUD/migration/reindexing.
Score Breakdown
⚙ Agent Friendliness
🔒 Security
Strengths: multi-scope isolation is explicitly described (global, agent:<id>, custom, project, user). Security posture around transport and secrets is not fully evidenced in the provided content; embedding provider calls likely use HTTPS but TLS enforcement is not documented here. Secret handling practices (e.g., logging/never logging) are not verifiable from the snippet. Dependency hygiene cannot be assessed from the provided manifest beyond presence of common libs.
⚡ Reliability
Best When
Used as an OpenClaw plugin in a TypeScript/Node environment where agents can call the plugin’s tools/hooks and where LanceDB + an embedding provider (OpenAI-compatible) are acceptable.
Avoid When
Avoid if you cannot run the plugin inside OpenClaw’s plugin lifecycle or if you require a standardized external API (REST/GraphQL/GRPC) with OpenAPI/SDK contracts; also avoid if you need strict compliance guarantees that are not documented (e.g., data residency/compliance).
Use Cases
- • RAG systems needing long-term memory with both semantic and keyword recall
- • Agentic assistants that must retrieve relevant prior facts/preferences across sessions
- • Projects that want hybrid retrieval with reranking and recency/decay weighting
- • Multi-tenant or multi-agent environments requiring scoped memory isolation
- • Teams that need memory management operations (import/export, migrate, reembed) via CLI
Not For
- • Production deployments that require a network-accessible REST/GraphQL service API from this package alone
- • Use cases needing fine-grained, per-record user authorization enforced externally (the plugin appears to implement logical scoping inside OpenClaw, not cryptographic enforcement)
- • Organizations that cannot store any memory content locally/offline (LanceDB storage plus optional Markdown mirror)
Interface
Authentication
Authentication is primarily to external embedding/reranker endpoints (OpenAI-compatible and “compatible endpoint” mentions). The README focuses on configuration behavior rather than a formal auth method list (e.g., API key vs bearer token) or per-scope authorization with the external endpoints.
Pricing
Pricing for the plugin itself is not described; costs depend on the embedding/reranking providers used (not specified here).
Agent Metadata
Known Gotchas
- ⚠ This is a plugin integrated into OpenClaw lifecycle hooks; an external agent expecting a standalone HTTP API will not find one.
- ⚠ Cross-encoder reranking depends on an external compatible endpoint; behavior includes fallback on API failure, so results may vary.
- ⚠ Hybrid scoring and diversity thresholds (minScore, MMR demotion threshold, age half-lives) can materially change recall; agents should be configured/tuned rather than assuming stable ranking.
- ⚠ Memory management CLI exists, but whether tool calls are strictly idempotent for repeated operations is not documented in the provided content.
Alternatives
Full Evaluation Report
Comprehensive deep-dive: security analysis, reliability audit, agent experience review, cost modeling, competitive positioning, and improvement roadmap for memory-lancedb-pro.
AI-powered analysis · PDF + markdown · Delivered within 30 minutes
Package Brief
Quick verdict, integration guide, cost projections, gotchas with workarounds, and alternatives comparison.
Delivered within 10 minutes
Score Monitoring
Get alerted when this package's AF, security, or reliability scores change significantly. Stay ahead of regressions.
Continuous monitoring
Scores are editorial opinions as of 2026-03-29.