Privacy-First Document Intelligence
Grounded answers from your files, with citations and transparent execution.
OpenIntelligence implements a full Retrieval-Augmented Generation (RAG) pipeline on iOS. Documents are parsed and chunked, embeddings are generated locally, and retrieval blends semantic similarity and keyword relevance. Responses are presented with grounding and citations.
Key Capabilities
Local-first by default
Document ingestion, chunking, embeddings, and search execute on-device unless you explicitly enable a cloud pathway.
Multi-format ingestion
Supports common formats including PDF, text, Markdown, CSV, and more, with OCR fallback for scanned pages when needed.
Hybrid retrieval
Combines vector search and BM25 keyword scoring, then improves results via Reciprocal Rank Fusion (RRF) and MMR diversification.
Grounded answers and citations
Answers are generated from retrieved passages and include citations back to source content where available.
Power-user controls
Container isolation, retrieval depth controls, and execution indicators support repeatable, auditable workflows.
Observability
Telemetry views and diagnostics help you understand retrieval quality and performance characteristics.