Adapting the App to work with Multiple LLMs

[TSSFL]

How could one adapt the app to work with multiple LLMs, especially, Llama3 and Claude? Would be great to generalize the app's functionality and identify any necessary modifications.

Any guidance, suggestions, or technical assistance from the community would be greatly appreciated.

TSSFL

[fjosue4]

Hey @TSSFL in order to call different API services you could add a Selector with the available options on the Setup.tsx

For each option on the select input, you will need to associate it with a different endpoint, so you can create a state on userSlice.tsx named selectedApi which will store the string you need to call the API

Using setUser as is now, you can store the selectedApi based on the user Selection.

    setUser: (state, action) => {
      state.name = action.payload.name
      state.API_KEY = action.payload.API_KEY
      state.proxy = action.payload.proxy
      state.selectedAPI = action.payload.selectedApi
    }

Then use that selectedApi state on the dispatcher

${proxy ? proxy : ''}https://generativelanguage.googleapis.com/v1beta/models/gemini-pro:generateContent?key=${apiKey}

Then replace the Google API with the restructured API endpoint

const { API_KEY: apiKey, proxy, conversation, selectedApi } = currentState.user

${proxy ? proxy : ''}${selectedApi}${apiKey}

Let me know if this helps you to have one path to complete your objective.

[TSSFL]

Thank you @fjosue4 , will work on this and come up for any help/clarification.

[fjosue4]

@TSSFL with the last update I sent there's a selector for the model, you could use the same but instead of a model for Gemini, you could ask for the AI endpoint you want to point in the setup page, so you provide the API key according yo the AI provider.

[TSSFL]

Thank you @fjosue4 for the updates. This is a great way to provide the API key based on the AI provider. I’ll go ahead and test the LLM web app.