Add Knowledge Fabric - Fast, multi-tenant hybrid RAG server on PostgreSQL

[sagarv48]

Summary

Add Knowledge Fabric to the community MCP servers and frameworks list.

Description

Knowledge Fabric is a vendor-neutral, multi-tenant hybrid RAG evidence retrieval engine built natively on PostgreSQL with pgvector HNSW indexes and full-text search (tsvector) fused via Reciprocal Rank Fusion (RRF, k=60). It provides sub-10ms retrieval latency, HIPAA/SOC 2 Row-Level Security (RLS) tenant isolation, and FastMCP server compliance.

Exposed Tools

• retrieve_evidence: Execute hybrid RRF retrieval across ingested documents and return structured evidence with citations.
• get_document: Fetch raw content and metadata for a specific document URI.
• explain_retrieval: Return detailed ranking diagnostics, lexical scores, and vector distances for a query.
• health_check: Verify database connectivity, embedding provider status, and row counts.
• list_sources: List distinct ingested source types and document counts.

Server Configuration Schema for Registry

{
  "$schema": "http://json-schema.org/draft-07/schema#",
  "name": "knowledge-fabric",
  "description": "High-performance hybrid RAG evidence retrieval on PostgreSQL (pgvector + BM25 RRF)",
  "repository": "https://github.com/sagarv48/knowledge-fabric",
  "license": "Apache-2.0",
  "packages": [
    {
      "registry": "pypi",
      "package": "knowledge-fabric"
    }
  ],
  "command": "uvx",
  "args": ["knowledge-fabric-mcp"],
  "env": {
    "DATABASE_URL": {
      "type": "string",
      "description": "PostgreSQL connection string (e.g. postgresql://user:pass@localhost:5432/knowledge_fabric)"
    },
    "EMBEDDING_PROVIDER": {
      "type": "string",
      "default": "ollama",
      "enum": ["ollama", "openai", "cohere", "mock"]
    },
    "RERANKER": {
      "type": "string",
      "default": "passthrough",
      "enum": ["passthrough", "cross_encoder", "cohere"]
    }
  },
  "tools": [
    {
      "name": "retrieve_evidence",
      "description": "Execute hybrid RRF retrieval across ingested documents and return structured evidence with citations."
    },
    {
      "name": "get_document",
      "description": "Fetch raw content and metadata for a specific document URI."
    },
    {
      "name": "explain_retrieval",
      "description": "Return detailed ranking diagnostics, lexical scores, and vector distances for a query."
    },
    {
      "name": "health_check",
      "description": "Verify database connectivity, embedding provider status, and row counts."
    },
    {
      "name": "list_sources",
      "description": "List distinct ingested source types and document counts."
    }
  ]
}

Checklist

• Alphabetically sorted under ### For servers in ADDITIONAL.md
• Open-source (Apache 2.0)
• FastMCP compatible and runnable via uvx / Docker