Memory (Qdrant) — OpenClaw Plugin

Qdrant-backed long-term memory with HNSW indexing for sub-50ms latency.

Installation

openclaw plugins install @wowsuchbot/memory-qdrant

Configuration

Add to your openclaw.json:

plugins:
  entries:
    memory-qdrant:
      enabled: true
      config:
        qdrant:
          url: http://localhost:6333
          collectionName: memories
          onDisk: true
          hnswConfig:
            m: 16
            efConstruct: 100
        embedding:
          provider: openai
          model: text-embedding-3-small
          apiKey: ${OPENAI_API_KEY}
        autoCapture: true
        autoRecall: true
  slots:
    memory: memory-qdrant

Features

• HNSW Indexing: Fast approximate nearest neighbor search with tunable parameters
• Memory Mapping: Store vectors on disk for sub-50ms latency (fits in OS cache)
• Auto-Recall: Inject relevant memories into context before agent starts
• Auto-Capture: Automatically store important information from conversations
• Duplicate Detection: Skip storing memories with 95%+ similarity
• Rule-Based Filtering: Capture preferences, facts, decisions, entities
• Three Tools: memory_recall, memory_store, memory_forget

CLI Commands

openclaw qtm list              # List memories
openclaw qtm search "query"     # Search memories
openclaw qtm stats             # Show statistics

Migration

From LanceDB to Qdrant:

node migrate.ts <lanceDbPath> <qdrantUrl> <qdrantApiKey> <openaiApiKey>

Performance Test

node test-performance.ts <qdrantUrl> <openaiApiKey>

Requirements

• Qdrant Server: Running at configured URL
• OpenAI API Key: Set in OPENAI_API_KEY environment variable

Version

2026.3.13

License

MIT

---

Repository: https://github.com/wowsuchbot/memory-qdrant