Visual agent workflow platform for non-technical operators. Design AI-powered business processes through conversation, run and monitor them from any device.
| Doc | Description |
|---|---|
| 01 Vision | Product overview, goals, phase 1 scope |
| 02 Architecture | System components and data flow |
| 03 Core Concepts | Graphs, nodes, knowledge, runs, roles |
| 04 Knowledge System | MD files, versioning, health score, improvement loop |
| 05 Node Types | LLM Agent, Human Checkpoint, Router, Tool Executor |
| 06 Human-in-the-Loop | Escalation, notifications, rating |
| 07 Tool Registry | Tool categories, built-ins, scoping |
| 08 Data Models | Database schemas |
| 09 API Spec | REST endpoints, WebSocket events |
| 10 Frontend Spec | UI/UX flows, mobile design, progressive education |
| 11 Runtime Spec | LangGraph execution engine, folder-as-domain traversal |
| 12 MCP Spec | MCP server, Telegram/WhatsApp, Claude Desktop |
| 13 Use Cases | Actor overview and use case diagrams (Mermaid) |
| 14 Activity Diagrams | Key flow diagrams for all major scenarios (Mermaid) |
- Backend: Python, FastAPI, LangGraph
- Frontend: React, React Flow, Tailwind CSS
- Database: PostgreSQL
- Storage: Local FS (dev) → S3 (prod) via adapter
- Queue: Background worker for async runs