An enterprise-grade, self-hosted AI assistant that turns your organizational knowledge into instant answers.
OPAA transforms scattered knowledge — stored in wikis, emails, documents, and files — into a unified intelligence layer. Ask questions in natural language and get sourced answers from your entire knowledge base, no matter where it's stored.
OPAA is an open-source RAG (Retrieval-Augmented Generation) system for organizations that need:
- 🔍 Unified search across Confluence, email, file systems, and custom sources
- 🧠 Intelligent Q&A using configurable LLM providers (OpenAI, Anthropic, local models)
- 🏢 On-premises deployment with full data sovereignty
- 🔐 Multi-team support with workspace isolation and fine-grained permissions
- ⚙️ Flexible architecture — swap databases, LLMs, and data sources without code changes
- Multiple User Interfaces: Web chat, chat bot integrations (Mattermost, RocketChat, Slack, Telegram, Signal, WhatsApp), REST API
- Flexible Data Sources: Confluence, Jira, email archives, file systems, cloud storage, issue trackers, custom APIs
- Configurable LLM Providers: OpenAI, Anthropic, open-source models, or local deployments
- Multiple Vector Databases: Elasticsearch, PostgreSQL + pgvector, Milvus, or cloud options
- Workspace Isolation: Multi-team support with role-based access control
- Audit & Compliance: Full audit logging, permission enforcement, GDPR/HIPAA support
- Enterprise Deployment: Kubernetes, Docker Compose, AWS, Azure, GCP, or air-gapped environments
Read the documentation:
- New to OPAA? Start here: GETTING-STARTED.md (5 min)
- Learn key concepts: CONCEPTS.md (10 min)
- See the full vision: VISION.md (15 min)
- Deep dive into features: See INDEX.md for role-based reading paths
Complete documentation in docs/:
- VISION.md — Complete product vision, use cases, architecture, principles
- CONCEPTS.md — Glossary and explanation of key concepts
- GETTING-STARTED.md — Guide to finding the right documentation
- INDEX.md — Complete documentation index with reading paths by role
Detailed specifications for each major feature:
- User Frontends — Web UI, chat integrations, REST API
- Data Indexing & RAG — Document indexing, semantic search, retrieval
- LLM Integration — Model configuration, providers, cost optimization
- Deployment & Infrastructure — On-premises, cloud, operations, scaling
- Access Control & Workspaces — Permissions, multi-tenancy, audit logging
- Architecture Decisions — Design rationale and technical decisions
Fortune 500 company with 5,000+ employees uses OPAA to make internal wiki, documentation, and archived emails searchable. Employees ask "What's our approval process for international hiring?" and get instant, sourced answers with data governance compliance.
50-person SaaS company deploys OPAA with Mattermost integration. Team members ask "@opaa-bot" questions. The system searches wikis, project documentation, and decision records. Weekly reports run automated queries.
Support team uses OPAA to provide better customer answers. Instead of searching multiple systems, they ask for product information and share sourced answers with customers.
Healthcare organization uses OPAA to index compliance policies and audit documents. When questioned, the system provides exact source references for audit trails.
- 🔧 Configurability First — Every component is swappable (LLM, vector DB, data sources)
- 🏢 On-Premises by Default — Data stays in your infrastructure, not external services
- 🔌 Extensible Architecture — Plugin system for data sources, LLM adapters, custom frontends
- 🔐 Security & Privacy Built In — Workspace isolation, permissions, audit trails, no data logging
- 📖 Source Attribution Always — Every answer includes source documents and confidence scores
OPAA is in early product definition phase. The documentation defines the complete vision and feature set. Implementation roadmap coming soon.
See CONTRIBUTING.md for guidelines on how to contribute.
For AI agents: Read AGENTS.md for project conventions and collaboration guidelines.
Technology choices will be made during implementation. OPAA is intentionally technology-agnostic:
- LLM Provider: Any OpenAI-compatible API (OpenAI, Anthropic Claude, Ollama, vLLM, etc.)
- Vector Database: Elasticsearch, PostgreSQL + pgvector, Milvus, cloud alternatives
- Deployment: Kubernetes, Docker Compose, AWS, Azure, GCP, or on-premises
- Data Sources: Confluence, Jira, Gmail, S3, SharePoint, Google Drive, Dropbox, issue trackers, and more
GNU Affero General Public License v3.0 (AGPL-3.0) — Free and open source. Commercial licenses available for organizations that cannot comply with AGPL terms. See CLA.md for contributor requirements.
- Want to learn more? Start with CONCEPTS.md
- Ready to contribute? See CONTRIBUTING.md
- Have feedback on the vision? Open an issue or discussion in GitHub