Knowledge Fabric

Knowledge Fabric is an MCP-native evidence retrieval platform for open-source use. It ingests source material, builds lexical and vector indexes, performs hybrid retrieval, and returns structured evidence packages for AI assistants.

Project Overview

Knowledge Fabric provides vendor-neutral retrieval primitives that improve answer quality by grounding responses in retrieved evidence.

Architecture

User Query
    ↓
Knowledge Fabric
    ↓
Evidence Package
    ↓
AI Assistant

Core retrieval pipeline:

ingestion -> chunking -> embeddings -> lexical retrieval + vector retrieval -> hybrid fusion -> evidence packaging -> MCP retrieval tools

Features

• Document ingestion and parsing
• Chunking and metadata extraction
• Embedding provider abstraction
• Lexical retrieval
• Vector retrieval
• Hybrid retrieval with rank fusion
• Evidence package contract
• MCP-native retrieval endpoints
• Retrieval evaluation framework

Roadmap

1. Stabilize ingestion and chunking interfaces.
2. Expand retrieval evaluation benchmarks and datasets.
3. Add optional pluggable reranking components.
4. Improve deployment examples for local and containerized development.

Getting Started

git clone <repository-url>
cd knowledge-fabric

Read the project guides in docs/:

• docs/README.md
• docs/repo-boundaries.md
• docs/phase-1-knowledge-fabric/README.md

Contributing

See CONTRIBUTING.md for contribution workflow, coding expectations, and review process.

License

This project is licensed under the Apache License 2.0. See LICENSE.

Non-goals

• Runtime workflow execution
• Product-specific integrations
• Customer-specific adapters and deployment details
• Credential management for external enterprise systems