A web GUI for interacting with locally-run LLMs, (DeekSeek for example)
- This requires you have Ollama installed with a LLM model locally installed. Please do that first
- Clone this repository
- Navigate to Root directory
- Add .env with the entry
MODEL='<model name here>' - Run
npm install - Run
node app.js - On a web browser, navigate to
https://localhost:3000
- This project saves messages in SessionStorage. Closing the tab will delete your chats
- I don't intend to do much more with this, may prettify it more, but I just wanted to try this
- Text formatting is applied after the LLM is finished streaming, not immediately. This can be jarring for LLMs like DeepSeek where it likes to provide a gigantic thought process that is then removed. Code is also pulled out and enlarged for better readability, but jarring from the packed-in format it originally displays in
- The name for each response block does not change color based on color scheme. Workaround: used a color that looks good for both.