NEW — Self-Hosted AI Infrastructure

Self-Hosted AI Infrastructure — Deploy in Minutes

Production-ready AI stacks that deploy automatically to your server. n8n workflows, AI agents, vector search, observability. Zero manual config. Paddle-compliant software products.

View Products View on GitHub

✓ 100% Paddle-compliant — Software products: infrastructure templates, deploy scripts, configs. No human services, no prohibited categories.

158★GitHub stars (hwdsl2/self-hosted-ai-stack)

180+Hermes skills available

<35 minMax deploy time (full fleet)

ZeroData leaves your server

3 VPSProviders supported (Hetzner, Contabo, DO)

€6/moStarting VPS cost (Hetzner CX23)

From Purchase to Production in 4 Steps

Zero manual config. Scripted deploys. You own everything.

Pick Product

Browse products, compare tiers. Each shows exact deliverables, requirements, deploy time, VPS cost estimates.

Checkout (Paddle)

Instant access to private repo + deploy scripts + docs. License key emailed.

Provision VPS

Create VPS on Hetzner/Contabo/DigitalOcean (5 min) or use existing. Run provisioning script.

Deploy & Run

Run deploy script (5-35 min). Services start on boot. Access Web UIs via HTTPS. Sérénité = we monitor for you.

Our Products

Production-ready AI stacks that deploy automatically to your server

View All Products →

⚙️

automation🎯 Non-Dev

n8n AI Automation Stack

400+ integrations + local LLMs + vector DB — one deploy

Complete n8n environment with Ollama (local LLMs), Qdrant (vector search), PostgreSQL. Pre-loaded with 15+ AI workflow templates: document Q&A, Slack bots, appointment scheduling, email processing. Deploys via Docker Compose in <5 minutes. Based on the 158-star hwdsl2/self-hosted-ai-stack pattern.

n8n (latest) with 400+ nodes
Ollama for local LLMs (Llama 3.1, Mistral, CodeLlama, Qwen 2.5)
Qdrant vector DB for RAG
PostgreSQL for n8n data
Traefik reverse proxy + auto-SSL
+5 more

99€one-timeDeploy: 3-5 minutes (scripted)

View Details Compare Tiers

✓ Paddle-compliant: Software product: infrastructure template + workflow JSON. Customer pays for deployable stack, not AI output.

🤖

agents⚡ Developer

AI Agent Runtime (Hermes + OpenCode)

Autonomous agents that run on YOUR server

Deploy Hermes Agent (autonomous ops, memory, skills, cron, Telegram gateway) + OpenCode (headless coding agent) as systemd services. Pre-configured with NVIDIA NIM models, 180+ skills library, project templates. Agents start on boot, persist across reboots. Based on Fractera's Hermes orchestrator pattern.

Hermes Agent v0.21+ (systemd service, multi-profile)
OpenCode v1.18+ (headless coding agent, 7 built-in agents)
NVIDIA NIM model routing (free tier) + Ollama local fallback
7 built-in coding agents (reviewer, debugger, refactorer, etc.)
Telegram/Discord/Slack gateway
+5 more

149€one-timeDeploy: 10-15 minutes (scripted)

View Details Compare Tiers

✓ Paddle-compliant: Software product: agent runtime + config templates. Customer pays for infrastructure deployment, not agent output.

📚

knowledge🎯 Non-Dev

RAG Knowledge Base Stack

Private ChatGPT for your docs — deploys in minutes

Complete RAG pipeline: document ingestion (PDF, MD, HTML, Notion, Confluence) → embedding (local or API) → vector search (Qdrant) → chat UI (Open WebUI) → API. Private, no data leaves your server. Pre-configured for internal knowledge, support docs, onboarding. Based on AY Automate's self-hosted AI stack guide.

Open WebUI (chat interface, multi-user, password-protected)
Qdrant vector DB (persistent, filtered search)
Ollama (local embeddings: nomic-embed-text) or OpenAI/Cohere API
Document ingestion pipeline (watch folder + API + Docling)
Multi-tenant collections (per team/project)
+5 more

79€one-timeDeploy: 5-8 minutes (scripted)

View Details Compare Tiers

✓ Paddle-compliant: Software product: RAG pipeline + chat UI. Customer pays for search infrastructure, not generated answers.

By Category

Automation

1 product

→

Agents

2 products

→

Knowledge

1 product

→

Observability

1 product

→

Devtools

1 product

→

Templates

1 product

→

Ready to Deploy Your AI Infrastructure?

Pick a product, checkout via Paddle, run the deploy script. Production in minutes.

Browse Products Custom Deployment

Frequently Asked Questions

Do these run on my server or yours?

100% on YOUR server (VPS, bare metal, cloud VM). You get root access, the deploy scripts, and full control. We never see your data, API keys, or logs.

What if I don't have a server?

We recommend: Hetzner CX43 (8 vCPU, 16GB RAM, 160GB NVMe, ~€16.50/mo) for EU/GDPR; Contabo Cloud VPS L (6 vCPU, 16GB RAM, 400GB NVMe, ~$25/mo) for max RAM/storage per dollar; DigitalOcean Basic (4 vCPU, 8GB RAM, 160GB SSD, ~$48/mo) for global low-latency. You provision it, give us root (for managed tiers) or run the script yourself (starter). You pay the provider directly.

Are there recurring fees to you?

No. One-time payment for the product. Optional: Sérénité maintenance at €149/mo (updates, health checks, backup audits, 1h support/mo). Cancel anytime.

How does deployment work?

Starter: you run `bash deploy.sh` on your server (5-35 min depending on product). Pro: we hop on a call, you share screen/SSH, we run it together. Enterprise: you give us temporary SSH, we deploy, harden, test, hand over with runbook.

What models do the agents use?

Default: NVIDIA NIM (free tier, 40 req/min) + Ollama (local, free: Llama 3.1, Mistral, CodeLlama, Qwen 2.5). Optional: OpenAI/Anthropic API (your keys, your bill), or any OpenAI-compatible endpoint (LiteLLM gateway included). You choose at deploy time.

Is this Paddle-compliant?

Yes. Every product is a SOFTWARE PRODUCT (deployable stack, templates, scripts, config). You pay for infrastructure automation, not AI-generated output. No human services, no outbound marketing, no prohibited categories. VPS costs paid directly to provider (Hetzner/Contabo/DigitalOcean).

Can I customize after purchase?

Absolutely. You get the source: Docker Compose files, scripts, configs, workflow JSON, agent skills, systemd units. Modify anything. Starter/Pro include docs for self-modification. Enterprise includes custom dev hours.

What happens after support period ends?

You keep everything forever. No license expiration. Renew Sérénité (€149/mo) for continued updates/health checks, or manage yourself. We provide upgrade guides for major versions.

Which VPS provider should I choose?

Hetzner: Best price/performance in EU, GDPR-native, great for EU customers. Contabo: Best RAM/storage per dollar, good for storage-heavy workloads, global locations. DigitalOcean: Best global latency, easiest API, per-second billing, good for distributed teams. We provide scripts for all three.