Bring Embeddings
to LlamaIndex

Connect your PostgreSQL + pgvector database to any AI agent and manage vector embeddings, similarity searches, and index optimizations through natural conversation.

What you can do

• Vector Similarity Search — Run nearest-neighbor queries using cosine, L2, or inner product distance metrics across millions of embeddings with a single prompt.
• Table Management — Discover which tables contain vector columns, create new embedding tables with custom dimensions, and inspect your schema.
• Embedding CRUD — Insert, update, and delete individual vector entries with metadata, keeping your knowledge base fresh and accurate.
• Index Optimization — Create HNSW or IVFFlat indexes on vector columns to accelerate approximate nearest-neighbor (ANN) queries by orders of magnitude.

How it works

1. Subscribe to the pgvector integration on the marketplace.
2. Paste your PostgreSQL connection string (e.g., postgresql://user:pass@host:5432/db).
3. Ask your AI agent to search vectors, create tables, or optimize indexes.

Who is this for?

• AI Engineers — Build RAG (Retrieval-Augmented Generation) pipelines that query production embeddings without writing custom API endpoints.
• Data Architects — Manage vector schemas, monitor index performance, and optimize query latency from a single conversational interface.
• Fullstack Developers — Add semantic search to existing apps by querying pgvector directly through your AI agent, no new microservices needed.