b
Retrieval-Augmented Generation

Knowledge that
answers back.

Upload documents. Ask questions. Get answers with citations. b-rag transforms your files into an intelligent knowledge base with semantic search across PDFs, images, and audio.

Local-first
Multimodal
Source-cited
Supported Formats
PDF
Documents
PNG
Images
MP3
Audio
Natural language queries with semantic understanding

How it works

01

Ingest

Upload documents via drag-and-drop or URL. Files are chunked, embedded, and indexed for semantic retrieval.

02

Query

Ask questions in natural language. Vector similarity search finds the most relevant document chunks.

03

Retrieve

Get synthesized answers with source citations. Every response links back to original documents.

Architecture

Built for accuracy

b-rag uses dense vector embeddings to capture semantic meaning, not just keyword matching. Results are ranked by actual relevance to your query intent.

HNSW IndexCosine SimilarityChunk OverlapContext Window
Open Source

Self-hosted & private

Run entirely on your machine. Your documents never leave your infrastructure. No cloud dependencies, no API keys required.

Start building knowledge

Create a project, upload your documents, and start asking questions.