Self-Hosted Model with ollama

How to configure self-hosted models in a Curiosity Workspace

This is a preview feature. Contact us if you experience any issues!

This guide will walk you through setting up ollama and connecting it to Curiosity so you can use models seamlessly.

The Curiosity Workspace needs to be able to access ollama via the network (Running in WSL for Windows is not currently supported)

Install & Start Ollama

• Download and install Ollama from the official site.

• Start Ollama on your system.

• Open a new chat and select a model in ollama. Then start a chat, this will trigger the download automatically.

Configure Ollama Settings

• Open Ollama’s settings.

• Enable: Expose Ollama to the network. This makes Ollama accessible from Docker containers and other applications.

Add an OpenAI Provider

• In Curiosity running inside Docker, add a new OpenAI-compatible AI provider with these settings:

Setting	Value
Model	The model you selected in ollama (e.g. deepseek-r1:8b)
API Key	ollama
Custom OpenAI Host	http://host.docker.internal:11434/{0}/{1

💡 Note: Inside a Docker container, host.docker.internal points to the host machine’s localhost.

Set Provider Priority

• Either remove other AI providers

• Or set Ollama as the default provider

• Alternatively, you can choose Ollama per chat session.

✅ That’s it! You should now have Ollama running inside Docker with your chosen model ready to use.