v0.1.0 - Initial Release
OpenDocuments v0.1.0 - Initial Release
Self-hosted open-source RAG platform that unifies organizational documents and answers natural language queries with AI.
Highlights
Core Platform
- Full RAG pipeline: Intent classification → Query decomposition → Cross-lingual search (Korean/English) → Hybrid search (dense + FTS5 sparse) → Reranking → Hallucination guard → Confidence scoring
- 3 interfaces: Web UI (React), CLI (17 commands), MCP Server (19 tools)
- Plugin architecture: 21 plugins with typed interfaces and lifecycle management
File Formats (22+)
Markdown, PDF, DOCX, XLSX/CSV, HTML, Jupyter Notebook, Email (.eml), Source Code (.js/.ts/.py/.java/.go/.rs and more), PowerPoint, JSON/YAML/TOML
Data Sources (9)
Local files, File upload, GitHub, Notion, Google Drive, S3/GCS, Confluence, Swagger/OpenAPI, Web Crawler + real-time Web Search (Tavily)
Model Providers (5)
- Local: Ollama (Qwen 3.5, Llama 4, DeepSeek V3.2, Gemma 3, K-EXAONE, Phi-4)
- Cloud: OpenAI (GPT-5.4), Anthropic (Claude Opus/Sonnet 4.6), Google (Gemini 3.1), xAI (Grok 4)
Security
- API key auth with RBAC (admin/member/viewer)
- PII auto-redaction (email, phone, credit card, IP)
- Audit logging, security alerts, rate limiting
- OAuth SSO (Google, GitHub)
- Workspace isolation
Developer Experience
opendocuments init-- interactive setup with hardware detection and model recommendationopendocuments plugin create-- scaffold custom plugins- TypeScript SDK (
@opendocuments/client) - Embeddable chat widget
- Docker Compose with optional Ollama
- Shell completions (zsh/bash)
Quick Start
npm install -g @opendocuments/cli
opendocuments init
opendocuments startStats
- 26 packages (5 core + 21 plugins)
- ~300 tests across 51 test suites
- 110+ commits
- MIT Licensed