Ollama + Open-WebUI integration

[bryzgaloff]

Hey there! Thanks a bunch for sharing this awesome project 👏

Is it possible to replace Fireworks AI with the local Ollama model and the agent UI with the Open-WebUI (or any other existing UI for LLMs)?

If that’s not an option right away, could you please guide me through the process of implementing it? Which components require replacing to migrate the implementation to existing solutions while keeping the core browser navigation logic of your project.

[shubcodes]

yes this can be done for the local ollama piece you will need to update config/config.toml with the correct model from ollama and also point it towards the correct endpoint by updating the base_url with the local ollama endpoint. You must remember that you'll a vision model. For open-webui this is a little more difficult because the current frontend is set up to receive updates via websockets and also render screenshots in realtime which might be hard to implement using open-webui. Feel free to submit a PR for the implementation

[bryzgaloff]

Thank you for your reply. So, you mean the connection between the current UI and the internal logic is quick and tight, so it is preferable to stick with the current UI since it is not easy to decouple?

However, instead of Fireworks AI, any vision AI model can be used, but only requires some tweaks to run with Ollama?