Build Agentic AI Apps with Postgres

Connect AI assistants directly to PostgreSQL

• Built-in PostgreSQL knowledge base

• Natural language SQL. Default read-only. Optional write mode

• Works with any PostgreSQL 14+ (RDS, Aurora, etc.)

• OpenAI, Voyage AI, or Ollama embeddings

Build AI apps with semantic search and retrieval

• Hybrid search — vector + BM25 + Reciprocal Rank Fusion

• Streaming responses via Server-Sent Events

• Multi-turn conversations with source attribution

• LLM providers: OpenAI, Anthropic, Ollama

Load docs into PostgreSQL for vectorization

• HTML, Markdown, reStructuredText, SGML

• Clone and load directly from Git repos

• Upsert mode for automated sync pipelines

• Ready for automatic embedding

PostgreSQL extensions for AI workloads

• pgVector — HNSW & IVFFlat similarity search

• VectorChord BM25 — Native keyword ranking

• pg_tokenizer — Text tokenization

Automatic embedding generation

• intelligently splits text into optimal-sized chunks

• handles embedding generation asynchronously using background workers without blocking.

• enables easy switching between OpenAI, Voyage AI, and Ollama.

• processes embeddings efficiently in batches for better API usage.

• automatically retries failed operations with exponential backoff

• provides built-in views for queue and worker monitoring.

• offers extensive GUC parameters for flexible configuration.