Jarvis Production System

Production-ready modular assistant backend with always-on event processing, secure tool execution, speech I/O and proactive notifications.

Dockerfiles

• services/core-api/Dockerfile
• services/event-engine/Dockerfile
• services/speech-service/Dockerfile

Python Voice Scripts

• services/speech-service/app/main.py (FastAPI: /stt, /tts, /health)
• services/speech-service/requirements.txt
• services/speech-service/Dockerfile

The Python voice service is started by Docker Compose as service speech-service.

Simple Install + Setup (from Git)

1. Install Docker Desktop (Windows/macOS) or Docker Engine + Compose plugin (Linux).
2. Clone project:

   git clone https://git.radionet.app/UkY/jarvis.git
   cd jarvis
   

3. Copy environment:

   cp .env.example .env
   

4. Start complete stack (auto-installs and initializes DB/Redis/Ollama model):

   docker compose up --build
   

5. Open webapp:
   • http://localhost:8080/jarvis/
   • unlock with WEBAPP_ACCESS_PASSWORD
6. Verify voice container is running:

   docker compose ps speech-service
   curl http://localhost:7000/health
   

Docker Setup (Local Ollama)

1. Copy env:

   cp .env.example .env
   

2. Start full stack (includes automatic Ollama model pull via ollama-init):

   docker compose up --build
   

3. Optional health checks:
   • GET http://localhost:8080/api/health
   • GET http://localhost:7000/health
   • GET http://localhost:11434/api/tags

Quick Start

1. Login:
   • POST http://localhost:8080/api/auth/login
   • default users: admin, operator, viewer
   • default password for all seeded users: password
2. Trigger AI tool decision:
   • POST http://localhost:8080/api/ai/tool-call
   • body example: { "prompt": "restart core api please" }
3. Jarvis voice webapp:
   • say: Jarvis wie viel grad ist es gerade in Saarbruecken?
   • mic listens only to commands starting with Jarvis

Admin Mode + Server Control

1. Activate admin mode (role admin token required):
   • POST /api/admin/activate with { "password": "ADMIN_MODE_PASSWORD" }
2. Test SSH connection:
   • POST /api/admin/connect with { "targetId": "server-main" }
3. Execute allowlisted action:
   • POST /api/admin/execute
   • example body: { "targetId": "server-main", "actionId": "service.status", "params": { "service": "nginx" } }
4. Approve risky action:
   • POST /api/admin/approve with { "requestId": 123 }

Server targets are defined in infra/admin/servers.json. Allowed remote actions are defined in infra/admin/actions.json.

API Surface (Core)

• POST /api/auth/login
• POST /api/events
• POST /api/health/wearable
• GET /api/tools
• POST /api/tools/execute
• POST /api/ai/tool-call
• POST /api/assistant/chat
• POST /api/webapp/access
• POST /api/admin/activate
• POST /api/admin/connect
• POST /api/admin/execute
• POST /api/admin/approve
• GET /api/health
• WS /ws/events

Speech Service

• POST http://localhost:7000/stt
• POST http://localhost:7000/tts
• GET http://localhost:7000/health

Docs

• docs/architecture.md
• docs/extension-guide.md
• docs/docker-setup.md