agentdb-semantic-vector-search
Semantische Vektorsuche mit AgentDB erstellen
Benutzer benötigen intelligente Dokumentensuche, die Bedeutung versteht, nicht nur Schlüsselwörter. Diese Skill bietet eine 5-Phasen-SOP zur Implementierung semantischer Vektorsuche mit AgentDB für RAG-Systeme, Wissensdatenbanken und kontextbewusste Abfragen.
스킬 ZIP 다운로드
Claude에서 업로드
설정 → 기능 → 스킬 → 스킬 업로드로 이동
토글을 켜고 사용 시작
테스트해 보기
"agentdb-semantic-vector-search" 사용 중입니다. Search for 'machine learning tutorials' with category filter
예상 결과:
- Document: Introduction to Neural Networks - relevance: 0.92
- Document: Python Machine Learning Guide - relevance: 0.89
- Document: Deep Learning Best Practices - relevance: 0.87
- Total results: 3 matching documents found in 45ms
"agentdb-semantic-vector-search" 사용 중입니다. Find documents similar to 'Python data analysis'
예상 결과:
- Document: Pandas Data Wrangling Guide - similarity: 0.94
- Document: NumPy for Beginners - similarity: 0.91
- Document: Statistical Analysis with Python - similarity: 0.88
- Documents retrieved and ranked by semantic similarity
"agentdb-semantic-vector-search" 사용 중입니다. Query knowledge base for 'troubleshooting database connections'
예상 결과:
- Result: Connection Pool Configuration - confidence: 0.89
- Result: MySQL Timeout Settings - confidence: 0.85
- Result: PostgreSQL SSL Certificates - confidence: 0.82
- Found 3 relevant articles from documentation knowledge base
보안 감사
안전Documentation-only skill containing markdown SOP files with TypeScript code examples. No executable code exists. All 31 static findings are false positives: (1) 'external_commands' patterns are TypeScript template literals and npm command examples, not Ruby backtick execution; (2) 'network' patterns are legitimate documentation URLs; (3) 'blocker' patterns (weak crypto, system reconnaissance) are technical terms in code comments (HNSW index parameters, SHA references). This skill provides guidance for implementing vector search and poses no security risk.
위험 요인
⚙️ 외부 명령어 (11)
🌐 네트워크 접근 (3)
📁 파일 시스템 액세스 (1)
품질 점수
만들 수 있는 것
RAG-Systeme erstellen
Retrieval-Augmented-Generation-Backends für LLMs mit semantischer Suchfunktion erstellen
Dokumentensuche-API
Semantische Such-Endpoints für Enterprise-Dokumenten-Retrieval-Systeme bereitstellen
Wissensdatenbank-Abfragen
Intelligente Wissensdatenbanken mit bedeutungsbasierter Dokumentensuche und Ähnlichkeitsabgleich erstellen
이 프롬프트를 사용해 보세요
Use AgentDB to set up a semantic vector search system. Configure it with 1536 dimensions for OpenAI ada-002 embeddings. Show how to initialize the database and embedding model.
Implement document embedding and storage. Process a corpus of documents, generate embeddings for each, and store them in AgentDB with metadata including title, content, and category.
Build a REST API endpoint for semantic search. Accept query text, generate embedding, search with topK parameter, apply metadata filters, and return ranked results.
Implement hybrid search combining vector similarity with keyword matching. Add re-ranking to improve relevance scores. Configure alpha parameter for vector and keyword balance.
모범 사례
- Verwenden Sie geeignete Embedding-Dimensionen für Ihr Modell (1536 für OpenAI ada-002, 768 für sentence-transformers)
- Wenden Sie Metadaten-Filterung an, um den Suchraum zu reduzieren und die Relevanz für bestimmte Dokumenttypen zu verbessern
- Überwachen Sie die Query-Latenz und passen Sie HNSW-Parameter (M, efConstruction) basierend auf Genauigkeitsanforderungen an
피하기
- Speichern von Dokumenten ohne Metadaten schränkt Filter- und Re-Ranking-Funktionen ein
- Verwendung nicht übereinstimmender Embedding-Dimensionen verursacht Indexierungs- und Suchfehler
- Ignorieren von Re-Ranking reduziert die Ergebnisqualität bei komplexen Abfragen