01 · INGEST
Upload documents
PDF, DOCX, TXT, MD, HTML, CSV, JSON — 50+ file types, processed in the background with progress tracking.
Retrieval-augmented generation
Ask your documents. Keep the answers grounded.
An enterprise-grade RAG platform with multi-provider LLM support, document intelligence and real-time streaming. Your documents stay on your infrastructure.
MITChromaDBSSE streamingGDPR-minded
$ git clone https://github.com/iwebbo/RAG.io
$ cd RAG.io && docker compose up -d
# upload documents, then ask
$ _
How it works
From files to cited answers.
02 · CHUNK
Smart chunking
Adaptive chunk size from 100 to 2000 tokens with configurable overlap, plus filename, page and type metadata.
03 · RETRIEVE
Semantic search
ChromaDB-powered retrieval with adjustable parameters, isolated per project.
04 · ANSWER
Any LLM provider
Pick the model per conversation. Answers stream over Server-Sent Events.
Token counting runs in real time so you can estimate cost before you commit to a model.
Features
Everything between the PDF and the answer.
SEARCH
Intelligent document search
ChromaDB-powered semantic search with adjustable retrieval parameters.
PROVIDERS
Multi-provider LLM support
OpenAI, Claude, Gemini, Ollama and 10+ providers behind one interface.
PROJECTS
Project-based organization
Isolate document collections and conversations by project.
STREAMING
Real-time responses
Server-Sent Events deliver the answer progressively.
CONTROL
Per-conversation tuning
Temperature from 0.0 to 2.0, top-k and context window management.
SECURITY
Enterprise security
JWT authentication, AES-256 encryption and GDPR compliance.
Every capabilityexpand
Ingestion
- Supported formats — PDF, DOCX, TXT, MD, HTML, CSV, JSON (50+ file types)
- Smart chunking — 100–2000 tokens with configurable overlap
- Metadata extraction — filename, page number and document type
- Batch processing — background async with progress tracking
Chat
- Token tracking — real-time counting for cost estimation
- Temperature control — per-conversation, 0.0–2.0
- Top-k and context — per-conversation limits
- Streaming — Server-Sent Events
Providers
- OpenAI
- Anthropic Claude
- Google Gemini
- OpenRouter
- xAI Grok
- Groq
- Hugging Face
- Ollama · LM Studio
- vLLM · LMDeploy
- Oobabooga
Built withPythonReactChromaDBDockerLinux
Star or fork on GitHub ↗