Local capabilities with Ollama and ComfyUI
Connect a private local text or image service and verify its models before running a task.
Raydo Desktop connects to a running Ollama service for text tasks and a running ComfyUI service for image generation. Start the service and prepare a model or checkpoint in it before connecting from Raydo.
Use a local Ollama model
- Run Ollama and download a model suitable for this machine. Use the Ollama documentation if you need to install it.
- In Settings → Local → Text models, enter the service address. On the same machine, the usual address is
http://localhost:11434. Enter just the service address, without/apior/v1. - Select Test, check the returned model list, and Save. Open Models to pick a default, or select the local model in a new chat. Use a short sample prompt to check that it can reply.
Use a local ComfyUI checkpoint
- Run ComfyUI and make at least one compatible checkpoint available in its
models/checkpointsfolder. Check the address ComfyUI displays. - In Settings → Local → Image generation, enter the address, usually
http://127.0.0.1:8188on the same computer. Select Test, check the checkpoint list, then Save. - Open Studio, choose the ComfyUI source and a checkpoint when offered, and test with one small image. If a workflow calls an external API or uses an external model, review that service’s account and routing separately.
A connection test fails or returns an empty list
Use an address on your computer or a private network reachable from the computer running Raydo. For a service behind an authenticated reverse proxy, use the optional access-token field. An empty list means you should check that the service has a model or checkpoint loaded and test again. For timeouts and wrong service errors, check the host, port, network and the service’s own logs. See model and provider troubleshooting.