[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M71S18PHMJMH9KK0SWM92GVG":3,"skill-same:01M71S18PHMJMH9KK0SWM92GVG":69},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":23,"publishedAt":43,"linkOnly":44,"markdownContent":45,"markdownBody":46,"frontMatter":47,"refName":48,"commitSha":49,"skillDirectory":50,"forksCount":51,"repositoryHtmlUrl":52,"repositoryHomepageUrl":6,"ownerHtmlUrl":53,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":54,"install":55,"resources":68},"01M71S18PHMJMH9KK0SWM92GVG","AgentDB Advanced Features",null,"Master advanced AgentDB features including QUIC synchronization, multi-database management, custom distance metrics, hybrid search, and distributed systems integration. Use when building distributed AI systems, multi-agent coordination, or advanced vector search applications.","manykarim\u002Frobotframework-agentguard","manykarim","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F61293164?v=4","01M71S15ZAXR15JHKM1N7BHT21","Many Kasiriha","01M71S160CWQXETSCYWXDVNYJ2",1,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Fmanykarim\u002Frobotframework-agentguard\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fagentdb-advanced\u002FSKILL.md",[18],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000010","后端与接口","SCENARIO","使用场景",[24,5,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42],"features","advanced","use","when","hybrid","custom distance metrics","custom","agent coordination","Hybrid Search","master","distributed","multi-agent","coordination","synchronization","applications","vector","search","metrics","database","2026-11-10 13:02:02",false,"---\nname: \"AgentDB Advanced Features\"\ndescription: \"Master advanced AgentDB features including QUIC synchronization, multi-database management, custom distance metrics, hybrid search, and distributed systems integration. Use when building distributed AI systems, multi-agent coordination, or advanced vector search applications.\"\n---\n\n# AgentDB Advanced Features\n\n## What This Skill Does\n\nCovers advanced AgentDB capabilities for distributed systems, multi-database coordination, custom distance metrics, hybrid search (vector + metadata), QUIC synchronization, and production deployment patterns. Enables building sophisticated AI systems with sub-millisecond cross-node communication and advanced search capabilities.\n\n**Performance**: \u003C1ms QUIC sync, hybrid search with filters, custom distance metrics.\n\n## Prerequisites\n\n- Node.js 18+\n- AgentDB v1.0.7+ (via agentic-flow)\n- Understanding of distributed systems (for QUIC sync)\n- Vector search fundamentals\n\n---\n\n## QUIC Synchronization\n\n### What is QUIC Sync?\n\nQUIC (Quick UDP Internet Connections) enables sub-millisecond latency synchronization between AgentDB instances across network boundaries with automatic retry, multiplexing, and encryption.\n\n**Benefits**:\n- \u003C1ms latency between nodes\n- Multiplexed streams (multiple operations simultaneously)\n- Built-in encryption (TLS 1.3)\n- Automatic retry and recovery\n- Event-based broadcasting\n\n### Enable QUIC Sync\n\n```typescript\nimport { createAgentDBAdapter } from 'agentic-flow\u002Freasoningbank';\n\n\u002F\u002F Initialize with QUIC synchronization\nconst adapter = await createAgentDBAdapter({\n  dbPath: '.agentdb\u002Fdistributed.db',\n  enableQUICSync: true,\n  syncPort: 4433,\n  syncPeers: [\n    '192.168.1.10:4433',\n    '192.168.1.11:4433',\n    '192.168.1.12:4433',\n  ],\n});\n\n\u002F\u002F Patterns automatically sync across all peers\nawait adapter.insertPattern({\n  \u002F\u002F ... pattern data\n});\n\n\u002F\u002F Available on all peers within ~1ms\n```\n\n### QUIC Configuration\n\n```typescript\nconst adapter = await createAgentDBAdapter({\n  enableQUICSync: true,\n  syncPort: 4433,              \u002F\u002F QUIC server port\n  syncPeers: ['host1:4433'],   \u002F\u002F Peer addresses\n  syncInterval: 1000,          \u002F\u002F Sync interval (ms)\n  syncBatchSize: 100,          \u002F\u002F Patterns per batch\n  maxRetries: 3,               \u002F\u002F Retry failed syncs\n  compression: true,           \u002F\u002F Enable compression\n});\n```\n\n### Multi-Node Deployment\n\n```bash\n# Node 1 (192.168.1.10)\nAGENTDB_QUIC_SYNC=true \\\nAGENTDB_QUIC_PORT=4433 \\\nAGENTDB_QUIC_PEERS=192.168.1.11:4433,192.168.1.12:4433 \\\nnode server.js\n\n# Node 2 (192.168.1.11)\nAGENTDB_QUIC_SYNC=true \\\nAGENTDB_QUIC_PORT=4433 \\\nAGENTDB_QUIC_PEERS=192.168.1.10:4433,192.168.1.12:4433 \\\nnode server.js\n\n# Node 3 (192.168.1.12)\nAGENTDB_QUIC_SYNC=true \\\nAGENTDB_QUIC_PORT=4433 \\\nAGENTDB_QUIC_PEERS=192.168.1.10:4433,192.168.1.11:4433 \\\nnode server.js\n```\n\n---\n\n## Distance Metrics\n\n### Cosine Similarity (Default)\n\nBest for normalized vectors, semantic similarity:\n\n```bash\n# CLI\nnpx agentdb@latest query .\u002Fvectors.db \"[0.1,0.2,...]\" -m cosine\n\n# API\nconst result = await adapter.retrieveWithReasoning(queryEmbedding, {\n  metric: 'cosine',\n  k: 10,\n});\n```\n\n**Use Cases**:\n- Text embeddings (BERT, GPT, etc.)\n- Semantic search\n- Document similarity\n- Most general-purpose applications\n\n**Formula**: `cos(θ) = (A · B) \u002F (||A|| × ||B||)`\n**Range**: [-1, 1] (1 = identical, -1 = opposite)\n\n### Euclidean Distance (L2)\n\nBest for spatial data, geometric similarity:\n\n```bash\n# CLI\nnpx agentdb@latest query .\u002Fvectors.db \"[0.1,0.2,...]\" -m euclidean\n\n# API\nconst result = await adapter.retrieveWithReasoning(queryEmbedding, {\n  metric: 'euclidean',\n  k: 10,\n});\n```\n\n**Use Cases**:\n- Image embeddings\n- Spatial data\n- Computer vision\n- When vector magnitude matters\n\n**Formula**: `d = √(Σ(ai - bi)²)`\n**Range**: [0, ∞] (0 = identical, ∞ = very different)\n\n### Dot Product\n\nBest for pre-normalized vectors, fast computation:\n\n```bash\n# CLI\nnpx agentdb@latest query .\u002Fvectors.db \"[0.1,0.2,...]\" -m dot\n\n# API\nconst result = await adapter.retrieveWithReasoning(queryEmbedding, {\n  metric: 'dot',\n  k: 10,\n});\n```\n\n**Use Cases**:\n- Pre-normalized embeddings\n- Fast similarity computation\n- When vectors are already unit-length\n\n**Formula**: `dot = Σ(ai × bi)`\n**Range**: [-∞, ∞] (higher = more similar)\n\n### Custom Distance Metrics\n\n```typescript\n\u002F\u002F Implement custom distance function\nfunction customDistance(vec1: number[], vec2: number[]): number {\n  \u002F\u002F Weighted Euclidean distance\n  const weights = [1.0, 2.0, 1.5, ...];\n  let sum = 0;\n  for (let i = 0; i \u003C vec1.length; i++) {\n    sum += weights[i] * Math.pow(vec1[i] - vec2[i], 2);\n  }\n  return Math.sqrt(sum);\n}\n\n\u002F\u002F Use in search (requires custom implementation)\n```\n\n---\n\n## Hybrid Search (Vector + Metadata)\n\n### Basic Hybrid Search\n\nCombine vector similarity with metadata filtering:\n\n```typescript\n\u002F\u002F Store documents with metadata\nawait adapter.insertPattern({\n  id: '',\n  type: 'document',\n  domain: 'research-papers',\n  pattern_data: JSON.stringify({\n    embedding: documentEmbedding,\n    text: documentText,\n    metadata: {\n      author: 'Jane Smith',\n      year: 2025,\n      category: 'machine-learning',\n      citations: 150,\n    }\n  }),\n  confidence: 1.0,\n  usage_count: 0,\n  success_count: 0,\n  created_at: Date.now(),\n  last_used: Date.now(),\n});\n\n\u002F\u002F Hybrid search: vector similarity + metadata filters\nconst result = await adapter.retrieveWithReasoning(queryEmbedding, {\n  domain: 'research-papers',\n  k: 20,\n  filters: {\n    year: { $gte: 2023 },          \u002F\u002F Published 2023 or later\n    category: 'machine-learning',   \u002F\u002F ML papers only\n    citations: { $gte: 50 },       \u002F\u002F Highly cited\n  },\n});\n```\n\n### Advanced Filtering\n\n```typescript\n\u002F\u002F Complex metadata queries\nconst result = await adapter.retrieveWithReasoning(queryEmbedding, {\n  domain: 'products',\n  k: 50,\n  filters: {\n    price: { $gte: 10, $lte: 100 },      \u002F\u002F Price range\n    category: { $in: ['electronics', 'gadgets'] },  \u002F\u002F Multiple categories\n    rating: { $gte: 4.0 },                \u002F\u002F High rated\n    inStock: true,                        \u002F\u002F Available\n    tags: { $contains: 'wireless' },      \u002F\u002F Has tag\n  },\n});\n```\n\n### Weighted Hybrid Search\n\nCombine vector and metadata scores:\n\n```typescript\nconst result = await adapter.retrieveWithReasoning(queryEmbedding, {\n  domain: 'content',\n  k: 20,\n  hybridWeights: {\n    vectorSimilarity: 0.7,  \u002F\u002F 70% weight on semantic similarity\n    metadataScore: 0.3,     \u002F\u002F 30% weight on metadata match\n  },\n  filters: {\n    category: 'technology',\n    recency: { $gte: Date.now() - 30 * 24 * 3600000 },  \u002F\u002F Last 30 days\n  },\n});\n```\n\n---\n\n## Multi-Database Management\n\n### Multiple Databases\n\n```typescript\n\u002F\u002F Separate databases for different domains\nconst knowledgeDB = await createAgentDBAdapter({\n  dbPath: '.agentdb\u002Fknowledge.db',\n});\n\nconst conversationDB = await createAgentDBAdapter({\n  dbPath: '.agentdb\u002Fconversations.db',\n});\n\nconst codeDB = await createAgentDBAdapter({\n  dbPath: '.agentdb\u002Fcode.db',\n});\n\n\u002F\u002F Use appropriate database for each task\nawait knowledgeDB.insertPattern({ \u002F* knowledge *\u002F });\nawait conversationDB.insertPattern({ \u002F* conversation *\u002F });\nawait codeDB.insertPattern({ \u002F* code *\u002F });\n```\n\n### Database Sharding\n\n```typescript\n\u002F\u002F Shard by domain for horizontal scaling\nconst shards = {\n  'domain-a': await createAgentDBAdapter({ dbPath: '.agentdb\u002Fshard-a.db' }),\n  'domain-b': await createAgentDBAdapter({ dbPath: '.agentdb\u002Fshard-b.db' }),\n  'domain-c': await createAgentDBAdapter({ dbPath: '.agentdb\u002Fshard-c.db' }),\n};\n\n\u002F\u002F Route queries to appropriate shard\nfunction getDBForDomain(domain: string) {\n  const shardKey = domain.split('-')[0];  \u002F\u002F Extract shard key\n  return shards[shardKey] || shards['domain-a'];\n}\n\n\u002F\u002F Insert to correct shard\nconst db = getDBForDomain('domain-a-task');\nawait db.insertPattern({ \u002F* ... *\u002F });\n```\n\n---\n\n## MMR (Maximal Marginal Relevance)\n\nRetrieve diverse results to avoid redundancy:\n\n```typescript\n\u002F\u002F Without MMR: Similar results may be redundant\nconst standardResults = await adapter.retrieveWithReasoning(queryEmbedding, {\n  k: 10,\n  useMMR: false,\n});\n\n\u002F\u002F With MMR: Diverse, non-redundant results\nconst diverseResults = await adapter.retrieveWithReasoning(queryEmbedding, {\n  k: 10,\n  useMMR: true,\n  mmrLambda: 0.5,  \u002F\u002F Balance relevance (0) vs diversity (1)\n});\n```\n\n**MMR Parameters**:\n- `mmrLambda = 0`: Maximum relevance (may be redundant)\n- `mmrLambda = 0.5`: Balanced (default)\n- `mmrLambda = 1`: Maximum diversity (may be less relevant)\n\n**Use Cases**:\n- Search result diversification\n- Recommendation systems\n- Avoiding echo chambers\n- Exploratory search\n\n---\n\n## Context Synthesis\n\nGenerate rich context from multiple memories:\n\n```typescript\nconst result = await adapter.retrieveWithReasoning(queryEmbedding, {\n  domain: 'problem-solving',\n  k: 10,\n  synthesizeContext: true,  \u002F\u002F Enable context synthesis\n});\n\n\u002F\u002F ContextSynthesizer creates coherent narrative\nconsole.log('Synthesized Context:', result.context);\n\u002F\u002F \"Based on 10 similar problem-solving attempts, the most effective\n\u002F\u002F  approach involves: 1) analyzing root cause, 2) brainstorming solutions,\n\u002F\u002F  3) evaluating trade-offs, 4) implementing incrementally. Success rate: 85%\"\n\nconsole.log('Patterns:', result.patterns);\n\u002F\u002F Extracted common patterns across memories\n```\n\n---\n\n## Production Patterns\n\n### Connection Pooling\n\n```typescript\n\u002F\u002F Singleton pattern for shared adapter\nclass AgentDBPool {\n  private static instance: AgentDBAdapter;\n\n  static async getInstance() {\n    if (!this.instance) {\n      this.instance = await createAgentDBAdapter({\n        dbPath: '.agentdb\u002Fproduction.db',\n        quantizationType: 'scalar',\n        cacheSize: 2000,\n      });\n    }\n    return this.instance;\n  }\n}\n\n\u002F\u002F Use in application\nconst db = await AgentDBPool.getInstance();\nconst results = await db.retrieveWithReasoning(queryEmbedding, { k: 10 });\n```\n\n### Error Handling\n\n```typescript\nasync function safeRetrieve(queryEmbedding: number[], options: any) {\n  try {\n    const result = await adapter.retrieveWithReasoning(queryEmbedding, options);\n    return result;\n  } catch (error) {\n    if (error.code === 'DIMENSION_MISMATCH') {\n      console.error('Query embedding dimension mismatch');\n      \u002F\u002F Handle dimension error\n    } else if (error.code === 'DATABASE_LOCKED') {\n      \u002F\u002F Retry with exponential backoff\n      await new Promise(resolve => setTimeout(resolve, 100));\n      return safeRetrieve(queryEmbedding, options);\n    }\n    throw error;\n  }\n}\n```\n\n### Monitoring and Logging\n\n```typescript\n\u002F\u002F Performance monitoring\nconst startTime = Date.now();\nconst result = await adapter.retrieveWithReasoning(queryEmbedding, { k: 10 });\nconst latency = Date.now() - startTime;\n\nif (latency > 100) {\n  console.warn('Slow query detected:', latency, 'ms');\n}\n\n\u002F\u002F Log statistics\nconst stats = await adapter.getStats();\nconsole.log('Database Stats:', {\n  totalPatterns: stats.totalPatterns,\n  dbSize: stats.dbSize,\n  cacheHitRate: stats.cacheHitRate,\n  avgSearchLatency: stats.avgSearchLatency,\n});\n```\n\n---\n\n## CLI Advanced Operations\n\n### Database Import\u002FExport\n\n```bash\n# Export with compression\nnpx agentdb@latest export .\u002Fvectors.db .\u002Fbackup.json.gz --compress\n\n# Import from backup\nnpx agentdb@latest import .\u002Fbackup.json.gz --decompress\n\n# Merge databases\nnpx agentdb@latest merge .\u002Fdb1.sqlite .\u002Fdb2.sqlite .\u002Fmerged.sqlite\n```\n\n### Database Optimization\n\n```bash\n# Vacuum database (reclaim space)\nsqlite3 .agentdb\u002Fvectors.db \"VACUUM;\"\n\n# Analyze for query optimization\nsqlite3 .agentdb\u002Fvectors.db \"ANALYZE;\"\n\n# Rebuild indices\nnpx agentdb@latest reindex .\u002Fvectors.db\n```\n\n---\n\n## Environment Variables\n\n```bash\n# AgentDB configuration\nAGENTDB_PATH=.agentdb\u002Freasoningbank.db\nAGENTDB_ENABLED=true\n\n# Performance tuning\nAGENTDB_QUANTIZATION=binary     # binary|scalar|product|none\nAGENTDB_CACHE_SIZE=2000\nAGENTDB_HNSW_M=16\nAGENTDB_HNSW_EF=100\n\n# Learning plugins\nAGENTDB_LEARNING=true\n\n# Reasoning agents\nAGENTDB_REASONING=true\n\n# QUIC synchronization\nAGENTDB_QUIC_SYNC=true\nAGENTDB_QUIC_PORT=4433\nAGENTDB_QUIC_PEERS=host1:4433,host2:4433\n```\n\n---\n\n## Troubleshooting\n\n### Issue: QUIC sync not working\n\n```bash\n# Check firewall allows UDP port 4433\nsudo ufw allow 4433\u002Fudp\n\n# Verify peers are reachable\nping host1\n\n# Check QUIC logs\nDEBUG=agentdb:quic node server.js\n```\n\n### Issue: Hybrid search returns no results\n\n```typescript\n\u002F\u002F Relax filters\nconst result = await adapter.retrieveWithReasoning(queryEmbedding, {\n  k: 100,  \u002F\u002F Increase k\n  filters: {\n    \u002F\u002F Remove or relax filters\n  },\n});\n```\n\n### Issue: Memory consolidation too aggressive\n\n```typescript\n\u002F\u002F Disable automatic optimization\nconst result = await adapter.retrieveWithReasoning(queryEmbedding, {\n  optimizeMemory: false,  \u002F\u002F Disable auto-consolidation\n  k: 10,\n});\n```\n\n---\n\n## Learn More\n\n- **QUIC Protocol**: docs\u002Fquic-synchronization.pdf\n- **Hybrid Search**: docs\u002Fhybrid-search-guide.md\n- **GitHub**: https:\u002F\u002Fgithub.com\u002Fruvnet\u002Fagentic-flow\u002Ftree\u002Fmain\u002Fpackages\u002Fagentdb\n- **Website**: https:\u002F\u002Fagentdb.ruv.io\n\n---\n\n**Category**: Advanced \u002F Distributed Systems\n**Difficulty**: Advanced\n**Estimated Time**: 45-60 minutes\n","# AgentDB Advanced Features\n\n## What This Skill Does\n\nCovers advanced AgentDB capabilities for distributed systems, multi-database coordination, custom distance metrics, hybrid search (vector + metadata), QUIC synchronization, and production deployment patterns. Enables building sophisticated AI systems with sub-millisecond cross-node communication and advanced search capabilities.\n\n**Performance**: \u003C1ms QUIC sync, hybrid search with filters, custom distance metrics.\n\n## Prerequisites\n\n- Node.js 18+\n- AgentDB v1.0.7+ (via agentic-flow)\n- Understanding of distributed systems (for QUIC sync)\n- Vector search fundamentals\n\n---\n\n## QUIC Synchronization\n\n### What is QUIC Sync?\n\nQUIC (Quick UDP Internet Connections) enables sub-millisecond latency synchronization between AgentDB instances across network boundaries with automatic retry, multiplexing, and encryption.\n\n**Benefits**:\n- \u003C1ms latency between nodes\n- Multiplexed streams (multiple operations simultaneously)\n- Built-in encryption (TLS 1.3)\n- Automatic retry and recovery\n- Event-based broadcasting\n\n### Enable QUIC Sync\n\n```typescript\nimport { createAgentDBAdapter } from 'agentic-flow\u002Freasoningbank';\n\n\u002F\u002F Initialize with QUIC synchronization\nconst adapter = await createAgentDBAdapter({\n  dbPath: '.agentdb\u002Fdistributed.db',\n  enableQUICSync: true,\n  syncPort: 4433,\n  syncPeers: [\n    '192.168.1.10:4433',\n    '192.168.1.11:4433',\n    '192.168.1.12:4433',\n  ],\n});\n\n\u002F\u002F Patterns automatically sync across all peers\nawait adapter.insertPattern({\n  \u002F\u002F ... pattern data\n});\n\n\u002F\u002F Available on all peers within ~1ms\n```\n\n### QUIC Configuration\n\n```typescript\nconst adapter = await createAgentDBAdapter({\n  enableQUICSync: true,\n  syncPort: 4433,              \u002F\u002F QUIC server port\n  syncPeers: ['host1:4433'],   \u002F\u002F Peer addresses\n  syncInterval: 1000,          \u002F\u002F Sync interval (ms)\n  syncBatchSize: 100,          \u002F\u002F Patterns per batch\n  maxRetries: 3,               \u002F\u002F Retry failed syncs\n  compression: true,           \u002F\u002F Enable compression\n});\n```\n\n### Multi-Node Deployment\n\n```bash\n# Node 1 (192.168.1.10)\nAGENTDB_QUIC_SYNC=true \\\nAGENTDB_QUIC_PORT=4433 \\\nAGENTDB_QUIC_PEERS=192.168.1.11:4433,192.168.1.12:4433 \\\nnode server.js\n\n# Node 2 (192.168.1.11)\nAGENTDB_QUIC_SYNC=true \\\nAGENTDB_QUIC_PORT=4433 \\\nAGENTDB_QUIC_PEERS=192.168.1.10:4433,192.168.1.12:4433 \\\nnode server.js\n\n# Node 3 (192.168.1.12)\nAGENTDB_QUIC_SYNC=true \\\nAGENTDB_QUIC_PORT=4433 \\\nAGENTDB_QUIC_PEERS=192.168.1.10:4433,192.168.1.11:4433 \\\nnode server.js\n```\n\n---\n\n## Distance Metrics\n\n### Cosine Similarity (Default)\n\nBest for normalized vectors, semantic similarity:\n\n```bash\n# CLI\nnpx agentdb@latest query .\u002Fvectors.db \"[0.1,0.2,...]\" -m cosine\n\n# API\nconst result = await adapter.retrieveWithReasoning(queryEmbedding, {\n  metric: 'cosine',\n  k: 10,\n});\n```\n\n**Use Cases**:\n- Text embeddings (BERT, GPT, etc.)\n- Semantic search\n- Document similarity\n- Most general-purpose applications\n\n**Formula**: `cos(θ) = (A · B) \u002F (||A|| × ||B||)`\n**Range**: [-1, 1] (1 = identical, -1 = opposite)\n\n### Euclidean Distance (L2)\n\nBest for spatial data, geometric similarity:\n\n```bash\n# CLI\nnpx agentdb@latest query .\u002Fvectors.db \"[0.1,0.2,...]\" -m euclidean\n\n# API\nconst result = await adapter.retrieveWithReasoning(queryEmbedding, {\n  metric: 'euclidean',\n  k: 10,\n});\n```\n\n**Use Cases**:\n- Image embeddings\n- Spatial data\n- Computer vision\n- When vector magnitude matters\n\n**Formula**: `d = √(Σ(ai - bi)²)`\n**Range**: [0, ∞] (0 = identical, ∞ = very different)\n\n### Dot Product\n\nBest for pre-normalized vectors, fast computation:\n\n```bash\n# CLI\nnpx agentdb@latest query .\u002Fvectors.db \"[0.1,0.2,...]\" -m dot\n\n# API\nconst result = await adapter.retrieveWithReasoning(queryEmbedding, {\n  metric: 'dot',\n  k: 10,\n});\n```\n\n**Use Cases**:\n- Pre-normalized embeddings\n- Fast similarity computation\n- When vectors are already unit-length\n\n**Formula**: `dot = Σ(ai × bi)`\n**Range**: [-∞, ∞] (higher = more similar)\n\n### Custom Distance Metrics\n\n```typescript\n\u002F\u002F Implement custom distance function\nfunction customDistance(vec1: number[], vec2: number[]): number {\n  \u002F\u002F Weighted Euclidean distance\n  const weights = [1.0, 2.0, 1.5, ...];\n  let sum = 0;\n  for (let i = 0; i \u003C vec1.length; i++) {\n    sum += weights[i] * Math.pow(vec1[i] - vec2[i], 2);\n  }\n  return Math.sqrt(sum);\n}\n\n\u002F\u002F Use in search (requires custom implementation)\n```\n\n---\n\n## Hybrid Search (Vector + Metadata)\n\n### Basic Hybrid Search\n\nCombine vector similarity with metadata filtering:\n\n```typescript\n\u002F\u002F Store documents with metadata\nawait adapter.insertPattern({\n  id: '',\n  type: 'document',\n  domain: 'research-papers',\n  pattern_data: JSON.stringify({\n    embedding: documentEmbedding,\n    text: documentText,\n    metadata: {\n      author: 'Jane Smith',\n      year: 2025,\n      category: 'machine-learning',\n      citations: 150,\n    }\n  }),\n  confidence: 1.0,\n  usage_count: 0,\n  success_count: 0,\n  created_at: Date.now(),\n  last_used: Date.now(),\n});\n\n\u002F\u002F Hybrid search: vector similarity + metadata filters\nconst result = await adapter.retrieveWithReasoning(queryEmbedding, {\n  domain: 'research-papers',\n  k: 20,\n  filters: {\n    year: { $gte: 2023 },          \u002F\u002F Published 2023 or later\n    category: 'machine-learning',   \u002F\u002F ML papers only\n    citations: { $gte: 50 },       \u002F\u002F Highly cited\n  },\n});\n```\n\n### Advanced Filtering\n\n```typescript\n\u002F\u002F Complex metadata queries\nconst result = await adapter.retrieveWithReasoning(queryEmbedding, {\n  domain: 'products',\n  k: 50,\n  filters: {\n    price: { $gte: 10, $lte: 100 },      \u002F\u002F Price range\n    category: { $in: ['electronics', 'gadgets'] },  \u002F\u002F Multiple categories\n    rating: { $gte: 4.0 },                \u002F\u002F High rated\n    inStock: true,                        \u002F\u002F Available\n    tags: { $contains: 'wireless' },      \u002F\u002F Has tag\n  },\n});\n```\n\n### Weighted Hybrid Search\n\nCombine vector and metadata scores:\n\n```typescript\nconst result = await adapter.retrieveWithReasoning(queryEmbedding, {\n  domain: 'content',\n  k: 20,\n  hybridWeights: {\n    vectorSimilarity: 0.7,  \u002F\u002F 70% weight on semantic similarity\n    metadataScore: 0.3,     \u002F\u002F 30% weight on metadata match\n  },\n  filters: {\n    category: 'technology',\n    recency: { $gte: Date.now() - 30 * 24 * 3600000 },  \u002F\u002F Last 30 days\n  },\n});\n```\n\n---\n\n## Multi-Database Management\n\n### Multiple Databases\n\n```typescript\n\u002F\u002F Separate databases for different domains\nconst knowledgeDB = await createAgentDBAdapter({\n  dbPath: '.agentdb\u002Fknowledge.db',\n});\n\nconst conversationDB = await createAgentDBAdapter({\n  dbPath: '.agentdb\u002Fconversations.db',\n});\n\nconst codeDB = await createAgentDBAdapter({\n  dbPath: '.agentdb\u002Fcode.db',\n});\n\n\u002F\u002F Use appropriate database for each task\nawait knowledgeDB.insertPattern({ \u002F* knowledge *\u002F });\nawait conversationDB.insertPattern({ \u002F* conversation *\u002F });\nawait codeDB.insertPattern({ \u002F* code *\u002F });\n```\n\n### Database Sharding\n\n```typescript\n\u002F\u002F Shard by domain for horizontal scaling\nconst shards = {\n  'domain-a': await createAgentDBAdapter({ dbPath: '.agentdb\u002Fshard-a.db' }),\n  'domain-b': await createAgentDBAdapter({ dbPath: '.agentdb\u002Fshard-b.db' }),\n  'domain-c': await createAgentDBAdapter({ dbPath: '.agentdb\u002Fshard-c.db' }),\n};\n\n\u002F\u002F Route queries to appropriate shard\nfunction getDBForDomain(domain: string) {\n  const shardKey = domain.split('-')[0];  \u002F\u002F Extract shard key\n  return shards[shardKey] || shards['domain-a'];\n}\n\n\u002F\u002F Insert to correct shard\nconst db = getDBForDomain('domain-a-task');\nawait db.insertPattern({ \u002F* ... *\u002F });\n```\n\n---\n\n## MMR (Maximal Marginal Relevance)\n\nRetrieve diverse results to avoid redundancy:\n\n```typescript\n\u002F\u002F Without MMR: Similar results may be redundant\nconst standardResults = await adapter.retrieveWithReasoning(queryEmbedding, {\n  k: 10,\n  useMMR: false,\n});\n\n\u002F\u002F With MMR: Diverse, non-redundant results\nconst diverseResults = await adapter.retrieveWithReasoning(queryEmbedding, {\n  k: 10,\n  useMMR: true,\n  mmrLambda: 0.5,  \u002F\u002F Balance relevance (0) vs diversity (1)\n});\n```\n\n**MMR Parameters**:\n- `mmrLambda = 0`: Maximum relevance (may be redundant)\n- `mmrLambda = 0.5`: Balanced (default)\n- `mmrLambda = 1`: Maximum diversity (may be less relevant)\n\n**Use Cases**:\n- Search result diversification\n- Recommendation systems\n- Avoiding echo chambers\n- Exploratory search\n\n---\n\n## Context Synthesis\n\nGenerate rich context from multiple memories:\n\n```typescript\nconst result = await adapter.retrieveWithReasoning(queryEmbedding, {\n  domain: 'problem-solving',\n  k: 10,\n  synthesizeContext: true,  \u002F\u002F Enable context synthesis\n});\n\n\u002F\u002F ContextSynthesizer creates coherent narrative\nconsole.log('Synthesized Context:', result.context);\n\u002F\u002F \"Based on 10 similar problem-solving attempts, the most effective\n\u002F\u002F  approach involves: 1) analyzing root cause, 2) brainstorming solutions,\n\u002F\u002F  3) evaluating trade-offs, 4) implementing incrementally. Success rate: 85%\"\n\nconsole.log('Patterns:', result.patterns);\n\u002F\u002F Extracted common patterns across memories\n```\n\n---\n\n## Production Patterns\n\n### Connection Pooling\n\n```typescript\n\u002F\u002F Singleton pattern for shared adapter\nclass AgentDBPool {\n  private static instance: AgentDBAdapter;\n\n  static async getInstance() {\n    if (!this.instance) {\n      this.instance = await createAgentDBAdapter({\n        dbPath: '.agentdb\u002Fproduction.db',\n        quantizationType: 'scalar',\n        cacheSize: 2000,\n      });\n    }\n    return this.instance;\n  }\n}\n\n\u002F\u002F Use in application\nconst db = await AgentDBPool.getInstance();\nconst results = await db.retrieveWithReasoning(queryEmbedding, { k: 10 });\n```\n\n### Error Handling\n\n```typescript\nasync function safeRetrieve(queryEmbedding: number[], options: any) {\n  try {\n    const result = await adapter.retrieveWithReasoning(queryEmbedding, options);\n    return result;\n  } catch (error) {\n    if (error.code === 'DIMENSION_MISMATCH') {\n      console.error('Query embedding dimension mismatch');\n      \u002F\u002F Handle dimension error\n    } else if (error.code === 'DATABASE_LOCKED') {\n      \u002F\u002F Retry with exponential backoff\n      await new Promise(resolve => setTimeout(resolve, 100));\n      return safeRetrieve(queryEmbedding, options);\n    }\n    throw error;\n  }\n}\n```\n\n### Monitoring and Logging\n\n```typescript\n\u002F\u002F Performance monitoring\nconst startTime = Date.now();\nconst result = await adapter.retrieveWithReasoning(queryEmbedding, { k: 10 });\nconst latency = Date.now() - startTime;\n\nif (latency > 100) {\n  console.warn('Slow query detected:', latency, 'ms');\n}\n\n\u002F\u002F Log statistics\nconst stats = await adapter.getStats();\nconsole.log('Database Stats:', {\n  totalPatterns: stats.totalPatterns,\n  dbSize: stats.dbSize,\n  cacheHitRate: stats.cacheHitRate,\n  avgSearchLatency: stats.avgSearchLatency,\n});\n```\n\n---\n\n## CLI Advanced Operations\n\n### Database Import\u002FExport\n\n```bash\n# Export with compression\nnpx agentdb@latest export .\u002Fvectors.db .\u002Fbackup.json.gz --compress\n\n# Import from backup\nnpx agentdb@latest import .\u002Fbackup.json.gz --decompress\n\n# Merge databases\nnpx agentdb@latest merge .\u002Fdb1.sqlite .\u002Fdb2.sqlite .\u002Fmerged.sqlite\n```\n\n### Database Optimization\n\n```bash\n# Vacuum database (reclaim space)\nsqlite3 .agentdb\u002Fvectors.db \"VACUUM;\"\n\n# Analyze for query optimization\nsqlite3 .agentdb\u002Fvectors.db \"ANALYZE;\"\n\n# Rebuild indices\nnpx agentdb@latest reindex .\u002Fvectors.db\n```\n\n---\n\n## Environment Variables\n\n```bash\n# AgentDB configuration\nAGENTDB_PATH=.agentdb\u002Freasoningbank.db\nAGENTDB_ENABLED=true\n\n# Performance tuning\nAGENTDB_QUANTIZATION=binary     # binary|scalar|product|none\nAGENTDB_CACHE_SIZE=2000\nAGENTDB_HNSW_M=16\nAGENTDB_HNSW_EF=100\n\n# Learning plugins\nAGENTDB_LEARNING=true\n\n# Reasoning agents\nAGENTDB_REASONING=true\n\n# QUIC synchronization\nAGENTDB_QUIC_SYNC=true\nAGENTDB_QUIC_PORT=4433\nAGENTDB_QUIC_PEERS=host1:4433,host2:4433\n```\n\n---\n\n## Troubleshooting\n\n### Issue: QUIC sync not working\n\n```bash\n# Check firewall allows UDP port 4433\nsudo ufw allow 4433\u002Fudp\n\n# Verify peers are reachable\nping host1\n\n# Check QUIC logs\nDEBUG=agentdb:quic node server.js\n```\n\n### Issue: Hybrid search returns no results\n\n```typescript\n\u002F\u002F Relax filters\nconst result = await adapter.retrieveWithReasoning(queryEmbedding, {\n  k: 100,  \u002F\u002F Increase k\n  filters: {\n    \u002F\u002F Remove or relax filters\n  },\n});\n```\n\n### Issue: Memory consolidation too aggressive\n\n```typescript\n\u002F\u002F Disable automatic optimization\nconst result = await adapter.retrieveWithReasoning(queryEmbedding, {\n  optimizeMemory: false,  \u002F\u002F Disable auto-consolidation\n  k: 10,\n});\n```\n\n---\n\n## Learn More\n\n- **QUIC Protocol**: docs\u002Fquic-synchronization.pdf\n- **Hybrid Search**: docs\u002Fhybrid-search-guide.md\n- **GitHub**: https:\u002F\u002Fgithub.com\u002Fruvnet\u002Fagentic-flow\u002Ftree\u002Fmain\u002Fpackages\u002Fagentdb\n- **Website**: https:\u002F\u002Fagentdb.ruv.io\n\n---\n\n**Category**: Advanced \u002F Distributed Systems\n**Difficulty**: Advanced\n**Estimated Time**: 45-60 minutes\n",{"name":5,"description":7},"main","63caf799a0409377ce4fd434e24e9eac7c05c6fa",".claude\u002Fskills\u002Fagentdb-advanced",0,"https:\u002F\u002Fgithub.com\u002Fmanykarim\u002Frobotframework-agentguard","https:\u002F\u002Fgithub.com\u002Fmanykarim","2026-07-27 14:20:37",{"source":8,"skillName":5,"commands":56,"prompt":66,"zipAvailable":67},[57,60,63],{"tool":58,"command":59},"npx","npx skills add manykarim\u002Frobotframework-agentguard --skill \"AgentDB Advanced Features\"",{"tool":61,"command":62},"bunx","bunx skills add manykarim\u002Frobotframework-agentguard --skill \"AgentDB Advanced Features\"",{"tool":64,"command":65},"pnpm","pnpm dlx skills add manykarim\u002Frobotframework-agentguard --skill \"AgentDB Advanced Features\"","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fmanykarim\u002Frobotframework-agentguard\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fagentdb-advanced\u002FSKILL.md",true,[],[70,90,98,116,130,146],{"skillNo":71,"name":72,"nameZh":6,"description":73,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":74,"categories":75,"keywords":79,"publishedAt":43},"01M71S584X6G22VN1KZ42DZ33K","good-example","Demonstrate a minimal valid Agent Skill used by AgentGuard's own tests.","https:\u002F\u002Fgithub.com\u002Fmanykarim\u002Frobotframework-agentguard\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fskills\u002Fgood-skill\u002FSKILL.md",[76],{"categoryNo":77,"categoryName":78,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000023","Skill与智能体开发",[80,81,82,83,84,85,86,87,88,89],"good example","good","example","valid","agentguard","read","write","minimal","tests","agent skill",{"skillNo":91,"name":92,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":93,"categories":94,"keywords":95,"publishedAt":43},"01M71S55P2APRSAC0S9F5X556A","bad-example","https:\u002F\u002Fgithub.com\u002Fmanykarim\u002Frobotframework-agentguard\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fskills\u002Fbad-skill\u002FSKILL.md",[],[96,82,97,85],"bad","bad example",{"skillNo":99,"name":100,"nameZh":6,"description":101,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":102,"categories":103,"keywords":105,"publishedAt":43},"01M71S53FGWCB5XXX4QGW3RF4P","injection-skill","Demonstrates a prompt-injection payload embedded in the body.","https:\u002F\u002Fgithub.com\u002Fmanykarim\u002Frobotframework-agentguard\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fsecurity\u002Finjection-skill\u002FSKILL.md",[104],{"categoryNo":77,"categoryName":78,"taxonomyType":21,"taxonomyTypeDesc":22},[106,107,85,108,109,110,111,112,113,114,115],"injection skill","injection","payload","body","prompt","unknown","Bash","prompt-injection","author","embedded",{"skillNo":117,"name":118,"nameZh":6,"description":119,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":120,"categories":121,"keywords":123,"publishedAt":43},"01M71S518MA1JN7F1DZ5XVRH06","curl-pipe-skill","Skill whose installer script demonstrates the `curl ... | sh` antipattern.","https:\u002F\u002Fgithub.com\u002Fmanykarim\u002Frobotframework-agentguard\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fsecurity\u002Fcurl-pipe-skill\u002FSKILL.md",[122],{"categoryNo":77,"categoryName":78,"taxonomyType":21,"taxonomyTypeDesc":22},[124,125,126,127,128,111,112,114,129],"curl","pipe","curl pipe skill","installer","antipattern","script",{"skillNo":131,"name":132,"nameZh":6,"description":133,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":134,"categories":135,"keywords":137,"publishedAt":43},"01M71S4YYFE3HJKDEQR5Z944ZW","clean-skill","A minimal, well-formed skill used as the security-scanner happy path.","https:\u002F\u002Fgithub.com\u002Fmanykarim\u002Frobotframework-agentguard\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fsecurity\u002Fclean-skill\u002FSKILL.md",[136],{"categoryNo":77,"categoryName":78,"taxonomyType":21,"taxonomyTypeDesc":22},[138,139,84,140,87,141,142,143,144,85,145],"clean skill","clean","scanner","grep","true","as","security","path",{"skillNo":147,"name":148,"nameZh":6,"description":149,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":150,"categories":151,"keywords":152,"publishedAt":43},"01M71S4W60AZYGA57HJWYBQZMJ","rf-libdoc-search","Search Robot Framework library\u002Fresource\u002Fsuite documentation to find matching keywords for a use case. Use when asked to find keywords, search libdoc, match a use case to keywords, or scan multiple libraries\u002Fresources for relevant keywords.","https:\u002F\u002Fgithub.com\u002Fmanykarim\u002Frobotframework-agentguard\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fintegrations\u002Fskills\u002Frf-libdoc-search\u002FSKILL.md",[],[40,153,154,155,156,157,158,159,160,161,27,162,26,163,164,165,166,167],"rf libdoc search","case","for","resources","relevant","scan","libraries","documentation","robot","framework","match","library","suite","matching","find"]