AI stacks · Overview
Ready-made AI stack recipes for WhatsApp, RAG, agents and automation, all of them tested in production by the Rollin Host engineering team.
Each stack here is a combination of tools that works in production, not a blog tutorial. The articles are put together by our engineering team and updated whenever a new version breaks something.
Available stacks
n8n + EvolutionAPI + OpenAI
An AI agent answering on WhatsApp in 2h. The most requested combo: automation + WhatsApp + LLM.
RAG with Qdrant + LangChain
Index company documents and answer questions with context. For internal support and qualified customer service.
How to choose
Start with the problem, not the tool
The biggest trap in AI is building the stack before deciding what it will do. If you are still figuring out the use case, start with the n8n + EvolutionAPI + OpenAI combination: it covers 80% of scenarios (customer service, qualification, follow-up) and you can swap pieces later.
Minimum specifications
| Stack | RAM | CPU | Disk | Suggested plan |
|---|---|---|---|---|
| n8n + EvolutionAPI + OpenAI | 4 GB | 4 vCPU | 80 GB | VPS Pro |
| RAG (Qdrant + LangChain) | 8 GB | 4 vCPU | 160 GB | VPS Plus |
| Self-hosted LLM (7B) | 24 GB GPU | 8 vCPU | 200 GB | Servidor GPU Estação |
| Self-hosted LLM (70B) | 80 GB GPU | 16 vCPU | 500 GB | Servidor GPU Cluster |
Last updated:
