RustyRAG v0.2.0 — The lowest-latency open-source RAG app

Sub-200ms responses on localhost. Sub-600ms to a browser across continents. No GPU. 977 PDFs. 56,114 chunks in Milvus.

What's New

⚡ Cerebras & Groq as LLM providers — pick any model and go
⚡ Jina AI local embeddings — jina-embeddings-v5-text-nano-retrieval for the best speed-to-quality ratio
⚡ Contextual retrieval — LLM-generated context prefixes per chunk for better accuracy (opt-in)

Also

→ Rust + Actix-Web — full RAG pipeline in a single async binary
→ Milvus vector DB with Swagger UI
→ Docker Compose — pull, add .env, build, run
→ Supports PDFs and zipped PDF bundles
→ SSE streaming with source citations

See the README for full docs.