<2-3 sentences summarising what the project addresses and solves, the target audience, and the benefit.> GraphRAG-rs provides a high-performance, modular Rust implementation of Graph-based Retrieval Augmented Generation (GraphRAG). It enables building knowledge graphs from documents and querying them via natural language, featuring configurable entity extraction and local LLM integration. The project targets developers and researchers seeking efficient, client-side (WASM), or hybrid RAG solutions with GPU acceleration, prioritizing performance and advanced RAG techniques.

How It Works

GraphRAG-rs employs a Rust-native architecture integrating advanced RAG research, including LightRAG for token reduction and HippoRAG for efficient retrieval. It supports three deployment architectures: Server-Only, WASM-Only (100% client-side with WebGPU), and Hybrid. This design prioritizes high performance, privacy-focused client-side processing, and scalable server deployments, leveraging cutting-edge research for enhanced retrieval quality and cost efficiency.

Quick Start & Requirements

• Primary Install: Rust 1.70+ and Node.js 18+ (for WASM) required. Clone (git clone) and build (cargo build --release). CLI tools (graphrag-cli, simple_cli) simplify setup.
• Prerequisites: Platform build tools (Linux, macOS, Windows). Optional: Docker (Qdrant), Ollama (local LLM). GPU acceleration requires compatible hardware/drivers.
• Quick Start: Minimal Rust example: GraphRAG::quick_start(...). CLI: graphrag-cli setup or cargo run --bin simple_cli config.toml "question".
• Documentation: Key guides include HOW_IT_WORKS.md, PIPELINE_ARCHITECTURE.md, and graphrag-core/QUICKSTART.md.
• Resource Footprint: Optimized release binaries (e.g., 5.2MB server). WASM is 100% client-side. GPU acceleration supported.

Highlighted Details

• Performance: Achieves 25-40x embedding speedups (ONNX Runtime Web GPU), 6000x token reduction (LightRAG), and 27x PageRank retrieval performance boost (6x cost reduction).
• RAG Quality: Integrates 5+ research papers for +20% accuracy. Advanced features include Symbolic Anchoring, Dynamic Edge Weighting, Causal Reasoning, and Hierarchical Clustering.
• Deployment: Supports Server-Only, WASM-Only (client-side, privacy-first), and Hybrid architectures.
• LLM/Embeddings: Robust Ollama integration for local, GPU-accelerated LLMs/embeddings (streaming, caching, custom parameters). Supports multiple API embedding providers.
• Architecture: Built with 4 publishable crates (core, wasm, leptos, server) using feature flags. Utilizes a trait-based design with 15+ core abstractions.

Maintenance & Community

The project is actively developed by automataIA with a detailed roadmap outlining progress through planned phases for WASM/Web UI, advanced features, and enterprise capabilities. No specific community channels (like Discord/Slack) or external contributors/sponsorships are mentioned in the README.

Licensing & Compatibility

Licensed under the MIT License, generally permitting commercial use and integration into closed-source projects without significant restrictions.

Limitations & Caveats

Phase 2 (WASM & Web UI) is under active development (60% complete), with specific components like Burn + wgpu GPU acceleration at 70% completion. PDF document support is planned but not yet implemented. The persistent-storage and neural-embeddings features are mutually exclusive.