Changelog
v0.1.0 (April 2026) - Initial Open Source Release
Added
- Dual Memory Architecture: Semantic memory (Weaviate) + knowledge graph (Neo4j)
- 6-Stage Ingestion Pipeline: Sync → Extract → Validate → Store → Cluster → Wiki
- Multi-Platform Support: Slack, Discord, Microsoft Teams adapters
- MCP Server: Model Context Protocol for LLM tool integration
- Wiki Generation: Auto-generated hierarchical wiki pages from conversations
- Smart Query Router: LLM-powered routing between semantic, graph, and web search
- Knowledge Graph: Entity extraction and relationship mapping with Neo4j
- Web Dashboard: React-based UI with graph visualization (cytoscape.js)
- Google ADK Agents: Orchestrated extraction, validation, and generation agents
- Mock Mode: Zero-API-key development with fixture data
- Docker Compose: Full stack deployment with one command
- Multilingual Support: Language detection and cross-language memory
- Observability: OpenTelemetry tracing, health checks, structured logging
Technical Details
- Backend: Python/FastAPI with 16 API route groups
- Bot: TypeScript with Vercel Chat SDK
- Frontend: React 19 + TypeScript + Vite + shadcn/ui
- Embeddings: Jina v4 (2048-dim, multimodal)
- LLM: Gemini 2.0 Flash (primary), Claude via LiteLLM (fallback)
Documentation
- Comprehensive getting started guides
- Interactive tutorials for common workflows
- In-depth concept documentation
- API reference with examples
- Integration guides for all platforms
- Contributing guidelines and development setup