Meet Docura
Docura is a blazing-fast, privacy-first RAG framework built for internal tools, public chatbots, and everything in between. Engineered with real-time document awareness, smart chunking, and modular deployment in mind.
Why Docura?
- Built on Gemini 2.5 Flash for lightning-fast, real-time answers.
- Uses hybrid retrieval — combining semantic search with keyword matching and cross-encoder reranking.
- NLTK-powered chunking ensures smarter, context-aware results.
- Privacy-first with full local support — no external APIs required.
- Modular and production-ready with FastAPI + Docker for cloud and edge deployment.
Core Capabilities
- Ingest PDFs, DOCX, HTML, PPTX, and emails with support for
system_cachefolders. - Custom citation engine with contextual highlights and automatic metadata extraction.
- Powered by FAISS for fast local vector search and offline-ready embeddings (MiniLM or GGML).
- Multi-threaded pipelines optimized for speed, scalability, and control.
Documentation
Get full integration guides and feature docs covering setup, deployment, embedding strategies, vector stores, and advanced ingestion flows.
