Sovereign Stack Starter Kit & Multi-LLM Router
Description
Docker Compose + n8n flows + routing prompts to set up your own multi-AI command center. Stop depending on a single API and distribute the work across 11 models.
This product does not yet have a purchase link configured.
Customer Portal (Odoo)
Payment processing, invoicing, and download management are handled in our secure Odoo space.
🔑 Already purchased? Log in here to download it →This kit solves a concrete problem: dependence on a single AI provider. When your workflow depends on an API and that API changes prices, limits tokens, or goes down, your operation stops.
The Starter Kit includes the complete configuration for:
- Multi-LLM Router: Routing logic by task type (code → DeepSeek, writing → GPT-4o-mini, analysis → Qwen, vision → Kimi). Based on the failover matrix documented in the digital garden.
- Sovereign Stack in Docker: n8n + Redis + PostgreSQL + Astro in a single
docker-compose.ymltested in production (4GB RAM VPS). - Routing Prompts: System prompt templates optimized for each model, with format instructions and specific constraints.
- Failover Matrix: Automatic backup configuration — if the primary provider fails, the router scales to the secondary without manual intervention.
The kit is designed for someone who already uses AI in their workflow and wants to stop depending on a single API. You don't need to know how to program, but you should feel comfortable with Docker and editing YAML/JSON files.
- Complete
docker-compose.yml: n8n + Redis + PostgreSQL + documented environment variables - 8 n8n flows in JSON (direct import): task router, automatic translation, content generation, data analysis, scraping + summary, email drafting, code review, research pipeline
- Routing matrix in YAML: 11 models × 8 task types with real performance scores
- 11 optimized system prompts (one per model) with format, tone, and constraint instructions
- Step-by-step deployment guide: from empty VPS to operational router in 45 minutes
- Monitoring template: n8n dashboard to see consumption per model, latency, and error rate
- README with troubleshooting: the 12 most common errors and how to resolve them
- Prerequisites: A VPS with at least 4GB RAM (Ubuntu 22.04+), Docker and Docker Compose installed, and at least 2 API keys for AI models (OpenAI + OpenRouter is the fastest combination to start).
- Clone and configure: Copy the
docker-compose.ymlto your server, edit the.envfile with your API keys, and adjust ports if necessary. - Bring up the stack: Run
docker compose up -d. Wait 2 minutes for n8n to initialize. - Import the flows: Open n8n at
http://your-server:5678, go to Workflows → Import from File, and import the 8 included JSONs. - Configure the router: Edit the routing matrix (
routing-matrix.yaml) with your available models. Mark the ones you don't have asdisabled. - Test the pipeline: Run the "router-test" flow with an example task. Verify that the response comes from the correct model.
- Iterate: Adjust the system prompts according to your tone and needs. The guide includes a chapter on prompt fine-tuning per model.