DocBento - Python AI Document Management

DocBento is a self-hosted document management application with OCR, document search and an AI assistant that provides page citations. It includes a Python API and background worker, a web client and a database deployment configuration. Source, Docker setup and installation documentation are included.

Features

• OCR with four engines
  • PaddleOCR PP-OCRv6 on ONNX Runtime: offline on CPU, about 50 languages, the default.
  • Tesseract 5 with English, Spanish, French and German included.
  • Docling for layout and tables (optional install).
  • Vision AI (Qwen or any vision model) for handwriting, stamps and photos.
  • Side-by-side view of each scan and its text.
• AI analysis (optional): title, summary, type, tags, language, dates, reference number, custom key fields per document type, and entities.
• AI assistant (optional)
  • An agent that searches, opens and reads documents before it answers.
  • Page citations for every statement.
  • 13 skills, including summarize, compare, extract data, obligations and deadlines, compliance, analyse numbers with charts, draft and translate.
  • Attachments, optional web search and export to Word.
• Hybrid search: full-text keywords plus semantic vectors across languages, with filters, snippets and page numbers.
• Documents: folders, multi-file upload (PDF, images, Office, CSV, HTML, e-mail), previews, versions, favorites, archive, trash, bulk actions, ZIP download and comments.
• Workflows: reviews and approvals with due dates, expiry reminders, and Word templates filled by hand or by AI.
• Administration and security
  • Four built-in roles, custom roles and groups.
  • Restricted folders that also apply to search and the AI.
  • Audit log with CSV export and a processing queue monitor.
  • Argon2id passwords, login lockout, API tokens and a REST API.
• Global: interface in English, Spanish, French and German.
• Ready to run
  • Docker Compose with PostgreSQL 17 + pgvector or MariaDB 11.8.
  • Secure setup script, sample data and a self-resetting demo mode.
  • HTTPS overlay, backup script and automated tests on both databases.

Requirements

• Docker Engine 24+ with Docker Compose v2
• 2 CPU cores and 4 GB RAM minimum (4 cores and 8 GB recommended), 10 GB free disk
• x86-64 or ARM64 Linux server, macOS or Windows with Docker
• Optional: an API key for an OpenAI-compatible AI provider
• A modern browser (Chrome, Edge, Firefox, Safari)

Instructions

1. Unzip the package.
2. Run sh scripts/setup.sh (add --db mariadb for MariaDB). It creates .env with unique secrets and prints the administrator e-mail and password.
3. Run docker compose up -d --build.
4. Open and sign in as the administrator.
5. Optional: to enable the AI features, open Settings, AI, and paste an OpenRouter key or point DocBento to your own OpenAI-compatible endpoint.
6. Optional: to load the sample company on a fresh install, add DEMO_PASSWORD to .env and run docker compose exec api python -m docbento.cli seed-demo.
7. For a public server with HTTPS, backups and upgrades, follow docs/DEPLOYMENT.md.

README.md and the docs/ folder cover installation, configuration, databases, OCR engines, AI features, security, permissions and the REST API.