[External] Running WEBCON AI Proxy with a fully local LLM

External content by Sébastien Anselment
The original post has appeared on my blog.

Since WEBCON 2026 R1, the WEBCON AI Proxy can be operated as a self-hosted service. Combined with an OpenAI-compatible API such as LiteLLM, this also makes it possible to integrate locally hosted language models into WEBCON.

I have documented a fully local setup using:

WEBCON > Self-hosted AI Proxy > LiteLLM > Ollama > Qwen3:8b

All AI components run as Docker containers on a dedicated Ubuntu VM. Only the HTTPS endpoint of the WEBCON AI Proxy is exposed, while LiteLLM, Ollama and the LLM remain within the internal Docker network.

This approach can be particularly interesting when digital sovereignty, data protection and control over the AI infrastructure are important requirements.

The complete step-by-step installation guide, including VM sizing, Docker, networking, TLS certificates, configuration, testing and troubleshooting, is available here:

⁠WEBCON Self-hosted AI Proxy installation guide

For additional guidance on selecting the appropriate LLM for different WEBCON AI features and use cases, see the WEBCON Knowledge Base article from Lily Adamowicz:

Selecting an LLM for AI features in WEBCON