IoT Claw - AI Automation Platform

An AI-assisted no-code automation platform for IoT and robotics workflows.

This repository combines:

• A FastAPI backend for chat, workflow storage, and simulation state
• A React frontend for chat-driven automation, workflow editing, and dashboard monitoring
• A simulation engine that mimics sensors/devices so workflows can be tested without hardware

Project Idea

The core concept is:

Talk -> Workflow -> Execution

Users describe what they want in natural language. The AI assistant converts that into a structured workflow (trigger, conditions, actions). The backend stores and executes the workflow against simulated devices (and later real devices through adapters).

The product vision supports three experience levels:

• Consumer mode: plain-language automation chat
• Maker mode: visual workflow building/editing
• Power User mode: schema-focused JSON workflow output

How It Works

1. Chat to Workflow

• Frontend sends conversation context and active mode to POST /chat
• Backend uses Gemini with mode-specific system prompts
• If the AI reply includes a workflow JSON block, frontend can save it as a workflow

2. Workflow Management

• Workflows are stored in SQLite (backend/openclaw.db)
• CRUD endpoints allow create, update, toggle, list, and delete
• Each update increments the workflow version
• Audit records are stored for key workflow events

3. Simulation and State

• A background simulation loop updates virtual sensors every 2 seconds
• Simulated device and sensor states are exposed via /state
• Dashboard polls:
  • /state (live state)
  • /execlog (execution history)
  • /notifications (user-facing alerts)

4. Execution Engine (Current Snapshot)

• Execution engine module exists with support for:
  • mqtt_event / time triggers
  • time / numeric / state conditions
  • device_control / delay / notify / robot_move actions
  • error policy with retries
• In the current backend startup path, execution start is wired through a backward-compatible entry point.
• The architecture is ready for a fuller event-bus-driven runtime integration.

Current Features in This Repo

• Multi-mode chat UI (Consumer, Maker, Power User)
• AI-generated workflow detection and save flow
• Workflow list, edit, enable/disable, delete
• Real-time simulation dashboard with sensor/device states
• Backend workflow persistence and audit endpoint

Tech Stack

• Backend: FastAPI, Pydantic, SQLite
• AI: Google Generative AI (Gemini)
• Frontend: React + Vite + Tailwind CSS
• Simulation: Python background simulation services

Repository Layout

• backend/: FastAPI app, workflow/data logic, execution/simulation modules
• frontend/: React app (chat, workflow UI, dashboard)
• config/: shared settings module
• open_claw_ai_automation_platform_master_context.md: product architecture context
• open_claw_ai_automation_platform_v2.md: upgraded architecture and roadmap context

Local Setup

Prerequisites

• Python 3.10+
• Node.js 18+

1. Backend Setup

From the backend folder:

pip install -r requirements.txt

Create backend/.env with:

GEMINI_API_KEY=your_api_key_here
GEMINI_MODEL=gemini-1.5-flash

Run backend:

uvicorn main:app --reload --host 0.0.0.0 --port 8000

Backend URL:

• http://localhost:8000

2. Frontend Setup

From the frontend folder:

npm install
npm run dev

Frontend URL:

• http://localhost:5173

Main API Endpoints

• GET /: service status
• POST /chat: AI assistant response (+ optional workflow JSON)
• GET /workflows: list workflows
• POST /workflows: create workflow
• GET /workflows/{id}: get one workflow
• PUT /workflows/{id}: update workflow
• PATCH /workflows/{id}/toggle: enable/disable workflow
• DELETE /workflows/{id}: delete workflow
• GET /state: simulation state snapshot
• POST /state/{device_path}: manual device control in simulation
• GET /execlog: execution log
• GET /notifications: simulation notifications
• GET /audit: workflow audit records

Example Workflow Shape

{
  "name": "Night Lighting",
  "description": "Turn on bedroom light when motion is detected at night",
  "enabled": true,
  "trigger": {
    "type": "mqtt_event",
    "topic": "sensor/motion",
    "condition": "payload == 'detected'"
  },
  "conditions": [
    {
      "type": "time",
      "after": "22:00",
      "before": "06:00"
    }
  ],
  "actions": [
    { "type": "device_control", "device": "light", "command": "on" },
    { "type": "delay", "seconds": 300 },
    { "type": "device_control", "device": "light", "command": "off" }
  ],
  "error_policy": {
    "retry_count": 2,
    "on_failure": "notify_user"
  }
}

Roadmap Direction

Planned architecture in the context docs includes:

• API gateway + JWT auth
• MQTT-first event bus integration
• OpenClaw adapter for real skill execution
• Template library onboarding
• Stronger error handling and audit surfaces

License

See LICENSE for project licensing details.