Architecture Overview
Gnovi is a knowledge-graph-native intelligence platform with a seven-layer architecture.
System Layers
- Data Ingestion — MCP connectors pull data from workplace tools
- Knowledge Graph Engine — FalkorDB stores entities and relationships with self-discovering schema
- Memory Architecture — Graphiti temporal memory with consolidation and forgetting
- Retrieval Engine — Hybrid search (vector + graph + BM25) with adaptive routing
- Agent Orchestration — LangGraph agents with graduated autonomy
- API Layer — FastAPI with authentication and rate limiting
- Web UI — Next.js chat interface, graph explorer, and settings
Technology Stack
| Component | Technology |
|---|---|
| Backend | Python 3.13+, FastAPI |
| Frontend | Next.js 16, React 19, TypeScript |
| Knowledge Graph | FalkorDB |
| Relational + Vector DB | PostgreSQL 18 + pgvector |
| Cache / Queue | Redis |
| Task Queue | Celery |
| LLM Gateway | LiteLLM (BYOM) |