Skip to main content
This setup keeps text-generation requests on your machine. Image or search requests remain external unless you also choose local providers or disable those features.

​
1. Install and start Ollama

Install Ollama for your operating system, then pull a model:
Confirm the service responds:

​
2. Start Presenton with the Ollama settings

On Docker Desktop, host.docker.internal normally works without the extra host mapping. Keeping it in the command is useful for compatible Linux Docker setups.

​
3. Verify the connection

Open http://localhost:5001, go to text-provider settings, choose Ollama, and use the model check. Confirm llama3.2:3b appears as available.

​
4. Generate a small test deck

Create a three- or four-slide presentation with a straightforward prompt. Smaller local models may need simpler instructions and may be less reliable with complex structured layouts.

​
5. Add local capabilities gradually

  • Enable Mem0 after configuring a compatible local memory model.
  • Use ComfyUI for local image generation.
  • Use SearXNG for self-hosted web search.
  • Give Docker GPU access when the model runtime can use it.
localhost inside the Presenton container points to the container itself. Use a host gateway or container-network hostname that Presenton can reach.