Chat interfaces

Open WebUI, LibreChat and AnythingLLM give you a chat interface like ChatGPT that runs on your server. AnythingLLM can also answer from your own documents.

Local models

Ollama runs open language models on the server itself, so nothing is sent to an outside provider. Large models need a lot of memory and ideally a GPU. A regular VPS is enough for small models and for interfaces that call an outside model.

Workflows and agents

n8n, Flowise, Langflow and Dify build workflows with AI steps in a visual editor, for example a support bot or a weekly report.

API keys

Store keys as environment variables on the service in Dokploy, never in the code. Set a spending limit with the provider so a broken loop cannot get expensive.