Any plan to support to chatgpt?

[LaiZhou]

Any plan to support to chatgpt?

[csells]

ChatGPT is a chat app built around the LLM API provided by OpenAI. @domesticmouse and I are working with @davidmigloz on support for OpenAI LLMs right now (among others): #30

[bernaferrari]

I wonder how that is going to work without leaking your OpenAI keys in the app. On web there is NextJS which separates server and app, but I don't think anyone did that on Flutter yet...

[csells]

You can certainly build your own server-side proxy and plug that into the AI Toolkit.

[bernaferrari]

Maybe you could have that somewhere in AI Toolkit, because it feels like a very common request for people (without that, they can't use the AI Toolkit in production)

[csells]

You certainly can. Vertex has support for that.

[bernaferrari]

But Vertex doesn't support OpenAI

[davidmigloz]

You can build a simple proxy as @csells was mentioning, eg. using Firebase Cloud Functions.

[DominikStarke]

I've created a minimal open-webui api client as LlmProvider implementation.
I'm currently using it with ollama (local model), but since open-webui also supports chat-gpt it should be possible to use it as well.

LlmChatView(
  provider: OpenwebuiProvider(
    host: 'http://127.0.0.1:3000', /// Open-webui host
    model: 'llama3.1:latest', /// The chat-model you want to use, for example gpt-4o-mini, 
    apiKey: "YOUR_API_KEY", /// Open-webui api key, not the api key of external llm provider,
    history: [] /// Previous conversation
  ),
)

The fork can be found here

[csells]

hey, @DominikStarke. if you want to contribute your open-webui provider, you can do so into the
flutter_ai_community repo that @davidmigloz and I are pulling together. I created an issue on that repo.