TrustRAG

Trustworthy Retrieval-Augmented Generation Knowledge Workbench

Verifiable Answers | Precise Citations | Document-Grounded Review

English | 简体中文 | 繁體中文 | 日本語

⬇️ Download · 📖 Documentation · 📋 Changelog

TrustRAG is a self-contained, multi-platform RAG knowledge workbench that runs entirely on your device. Upload documents, ask questions, and inspect the evidence behind each AI answer — with every citation traceable to its original source, page, and heading.

🎯 Core Features

• Multi-platform desktop — Windows (.exe installer + portable), macOS, Linux, Android, iOS, Web — all from a single Flutter codebase
• Self-contained desktop mode — Embedded SQLite + Rust backend bundled inside the app; no external database, no server setup
• RAG pipeline — Document-grounded retrieval-augmented generation with configurable LLM and embedding providers
• Citation tracking — Every AI response includes traceable source citations linked to document, chunk, page, and heading
• Citation review — Approve, reject, or flag citations for accuracy with full review history
• Full-text search — FTS5-powered search across all uploaded documents
• Knowledge graph — Entity and relation extraction for cross-document exploration
• Workspace collaboration — Multi-user workspaces with role-based member management
• Server mode — Full-stack deployment with PostgreSQL + pgvector, Redis, MinIO, and Docker Compose

📥 Download

Platform	Files
Windows	.exe installer, portable .zip
macOS	.tar.gz (universal)
Linux	.tar.gz (x64)
Android	.apk
iOS	.tar.gz (unsigned)
Web	.tar.gz (static files)

🏗 Architecture

TrustRAG operates in two modes:

Desktop Mode (Self-Contained)

Flutter Client (Windows / macOS / Linux / Android / iOS)
  └── HTTP/SSE
      └── Embedded Rust Backend (Axum)
          ├── Auth, workspace, document, search, chat, citation, review APIs
          ├── RAG orchestration and provider registry
          └── SQLite + FTS5 (embedded, zero config)

Note: Desktop mode currently supports TXT, Markdown, and HTML documents. For PDF/DOCX parsing, please use Server Mode which includes the Python Document Processor.

Server Mode (Full Stack)

Flutter Client (Web / Desktop)
  └── HTTP/SSE
      └── Rust Backend (Axum)
          ├── Auth, workspace, document, search, chat, citation, review APIs
          ├── RAG orchestration and provider registry
          ├── PostgreSQL + pgvector / pg_trgm
          ├── Redis
          ├── MinIO / S3-compatible object storage
          └── Python Document Processor (FastAPI)
              ├── PDF parsing (PyMuPDF)
              ├── DOCX parsing (python-docx)
              ├── OCR (Tesseract)
              └── Format conversion (Pandoc)

📁 Project Structure

TrustRAG/
├── apps/client/          # Flutter multi-platform client
├── backend/              # Rust Axum backend (dual-mode: postgres/desktop)
├── doc-processor/        # Python FastAPI document processing service
├── infra/                # Docker Compose, Caddy, PostgreSQL init scripts
├── scripts/              # Release notes generator
├── docs/                 # Product docs, API docs, roadmap
├── CHANGELOG.md          # Release changelog
└── .github/workflows/    # CI/CD: ci.yml, test-build.yml, release.yml

🔧 Tech Stack

Layer	Technology
Client	Flutter, Dart, Riverpod, go_router, dio
Backend	Rust, Axum, tokio, SQLx, JWT, argon2, tracing
Document processor	Python, FastAPI, PyMuPDF, python-docx, pytesseract
Desktop data	SQLite + FTS5
Server data	PostgreSQL 16, pgvector, pg_trgm, Redis 7, MinIO
CI/CD	GitHub Actions (ci + test-build + release)

🚀 Quick Start

Desktop App (Recommended)

1. Download the installer for your platform from Releases
2. Install and launch TrustRAG
3. Create a workspace, upload documents, start asking questions

Server Mode (Docker Compose)

cp .env.example .env
cd apps/client && flutter pub get && flutter build web --dart-define=API_BASE_URL=/api && cd ../..
cd infra && docker compose up --build

Open http://localhost to access the web app.

Development

# Backend
cd backend && cargo run

# Flutter client
cd apps/client && flutter run -d chrome --dart-define=API_BASE_URL=http://localhost:8080

# Document processor
cd doc-processor && pip install -r requirements.txt && uvicorn app.main:app --port 8081 --reload

⚠️ Notes

• Windows: May show SmartScreen warning on first launch — click More info → Run anyway
• macOS: App is unsigned — right-click and select "Open", or allow in System Settings > Privacy & Security
• Desktop mode: Uses SQLite with FTS5 for search; no external database needed
• Server mode: Requires Docker, PostgreSQL, Redis, and MinIO

📄 License

Apache License 2.0 — see LICENSE.

🙏 Acknowledgments

• LINUX.DO community

---

Made by XimilalaXiang