skills.rest
ishandutta2007
ishandutta2007COMMUNITY
Explore all skills by this author →
Website𝕏@ishandutta2007GitHub

AgentDB Vector Search

Implements semantic vector search with AgentDB using HNSW indexing and quantization.

11|Updated Jan 20, 2026
Search Tags:#semantic search#embeddings#rag#agentdb#vector search#similarity search#hnsw
One-click install
npx skills add https://github.com/ishandutta2007/claude-agent-orchestration --skill agentdb-vector-search-ishandutta2007
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: AgentDB Vector Search
Source: https://github.com/ishandutta2007/claude-agent-orchestration/tree/main/.claude/skills/agentdb-vector-search
Command: npx skills add https://github.com/ishandutta2007/claude-agent-orchestration --skill agentdb-vector-search-ishandutta2007

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires agentdb, agentic-flow.

What problem does it solve? Building semantic search and retrieval-augmented generation (RAG) systems requires a fast vector database, but traditional solutions are slow at scale and consume excessive memory. This Skill provides CLI commands and TypeScript APIs to store embeddings, run similarity queries, and tune performance with AgentDB. ## Core Features & Use Cases - Vector Storage & Similarity Search: Store documents with embeddings and query by cosine, euclidean, or dot-product distance with sub-millisecond HNSW-indexed retrieval. - Hybrid & MMR Search: Combine vector similarity with metadata filters, and use Maximal Marginal Relevance for diverse, non-redundant result sets. - Memory-Efficient Quantization: Reduce memory 4-32x with binary, scalar, or product quantization for large vector collections. - Use Case: Build a RAG pipeline where user questions are embedded, matched against a knowledge base of 100K+ documents, and the top context is injected into an LLM prompt. ## Quick Start Initialize an AgentDB vector database and run a semantic similarity query against my document embeddings.

Frequently Asked Questions about AgentDB Vector Search

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I set up vector search with AgentDB?▼

Initialize a database with npx agentdb@latest init ./vectors.db, specifying the embedding dimension for your model. Then store documents with embeddings via the CLI or the createAgentDBAdapter TypeScript API and query with similarity thresholds.

How do I build a RAG pipeline with vector search?▼

Embed the user question, run a similarity search against stored document embeddings with a threshold around 0.7, and inject the top results as context into the LLM prompt. AgentDB's retrieveWithReasoning supports MMR for diverse context.

What embedding dimensions does AgentDB support?▼

AgentDB supports configurable dimensions set at initialization: 1536 for OpenAI ada-002, 768 for sentence-transformers, and 384 for all-MiniLM-L6-v2. Use the --dimension flag or a preset when running the init command.

Does AgentDB support metadata filtering with vector search?▼

Yes, hybrid search combines vector similarity with metadata filters such as category or date ranges. Pass a filters object alongside the query and limit to narrow results by structured fields.

Why is my vector search using too much memory?▼

High memory usage comes from storing full-precision embeddings. Enable binary quantization for 32x reduction, scalar for 4x, or product quantization for 8-16x reduction via the quantizationType adapter option.

Why is vector search returning irrelevant results?▼

Low relevance usually means the similarity threshold is too low or results are redundant. Raise the threshold toward 0.8, or enable MMR with useMMR: true to balance relevance and diversity in the result set.

Related Skills

Explore More Software Engineering Skills→
ACGSpgpACGSpgp

rag-implementation

Implement RAG systems with vector databases, embedding models, and retrieval strategies.

Official
Advanced
JhabbigJhabbig

rag-implementation

Build retrieval-augmented generation systems that ground LLM answers in external knowledge.

Community
Advanced
TheGraphicKingTheGraphicKing

kb-retriever

Retrieves top-K semantically relevant knowledge base chunks with citations and relevance scores.

Community
Intermediate
as4584as4584
2

hybrid-search-implementation

Combine vector and keyword search results using RRF, linear combination, or cross-encoder reranking.

Community
Advanced