Why AI Agents Need Verifiable Evidence: Building an MCP-Native Retrieval Engine with PostgreSQL

Description and use

Knowledge Fabric delivers verifiable evidence for agentic workflows by combining BM25 and vector search in PostgreSQL, providing metadata and hashes for each retrieved passage.

Problem

Standard Retrieval-Augmented Generation (RAG) systems lack provenance, exact-term recall, and efficient context sizing, making them unsuitable for agentic workflows that require verifiable evidence and policy compliance.

Outcome

Agents can now retrieve context with exact-term recall, provenance, and efficient prompt sizing, reducing policy failures and improving execution reliability.

Workflow

• 1. Index documents into PostgreSQL with pgvector and tsvector for hybrid search.
• 2. Use Knowledge Fabric’s retrieve_evidence MCP tool to perform RRF‑based hybrid search and obtain cited evidence packages.
• 3. Agent calls get_document to fetch full content if needed.
• 4. Agent uses explain_retrieval to understand scoring breakdown.
• 5. Agent plans actions via Intent Fabric and executes writes through Enterprise Adapters, ensuring provenance and policy compliance.

Evidence

Knowledge Fabric is built directly on PostgreSQL + pgvector: Hybrid Search with Reciprocal Rank Fusion (RRF) fuses PostgreSQL BM25-style lexical search with semantic vector embeddings

Source

retrieve_evidence: Performs hybrid search and returns bounded, cited evidence packages

Source