RavenRustRAG¶

RavenRustRAG is a local-first RAG (Retrieval-Augmented Generation) engine written in Rust. It provides millisecond document indexing, microsecond query latency, and ships as a single static binary with zero runtime dependencies.

What is RAG?¶

Retrieval-Augmented Generation combines document retrieval with LLM prompting. Instead of feeding an entire corpus to a language model, RAG retrieves the most relevant chunks and provides them as context. This reduces hallucinations, keeps responses grounded in your data, and works within token limits.

Key Features¶

• Vector search with cosine similarity over local embeddings
• HNSW index for O(log n) approximate nearest neighbor search
• Hybrid search combining BM25 keyword matching with vector retrieval (RRF fusion)
• Knowledge graph for entity-relationship traversal and graph-vector fusion
• Semantic splitting via embedding cosine similarity between sentences
• Lock-free caching with DashMap for concurrent embedding cache access
• Memory-mapped SQLite with 256 MB mmap for zero-copy reads
• HTTP API (Axum) with Bearer auth, rate limiting, CORS, and OpenAPI schema
• MCP server for AI assistant integration (Claude, Copilot, Cursor)
• CLI with 20 commands for scripting and automation
• File watching with debounce for automatic re-indexing
• Incremental indexing via SHA-256 content fingerprinting
• Docker support with minimal scratch-based images (~15 MB)

Architecture¶

RavenRustRAG is a Cargo workspace with 10 crates:

Getting Started¶

See Installation to get the binary, then Quick Start to index your first documents.