Does the js sdk expose a way to use the default AI provider set in the settings?

[armanckeser]

One of the first things I assumed for the pipe sdk would be to expose a way to send a message to the default ai model set in the settings (e.g. pipe.generate_with_default_model, or pipe.get_model, etc.). Is this not the case, is it planned, or did I just miss it somewhere? Maybe @louis030195? Thanks!

[louis030195]

@armanckeser hey

can you try this

screenpipe/screenpipe-js/main.ts

Line 722 in 7b9ae36

/**

i just implemented it

[louis030195]

eg @armanckeser

import { generateText } from 'ai';
import { openai } from '@ai-sdk/openai';
import { pipe } from '@screenpipe/sdk';

async function generateWithScreenpipeSettings() {
  // get settings from screenpipe
  const settings = await pipe.settings.getAll();
  
  // use openai key and model from screenpipe settings
  const { text } = await generateText({
    model: openai(settings.aiModel),
    apiKey: settings.openaiApiKey,
    system: settings.customPrompt,
    prompt: "what's the weather like?",
  });

  console.log('ai response:', text);
}

// usage
generateWithScreenpipeSettings().catch(console.error);

[louis030195]

import { generateText } from 'ai';
import { ollama } from 'ollama-ai-provider';
import { pipe } from '@screenpipe/sdk';

async function generateWithScreenpipeSettings() {
  const settings = await pipe.settings.getAll();
  
  // use ollama if the provider type is native-ollama
  if (settings.aiProviderType === 'native-ollama') {
    const { text } = await generateText({
      model: ollama(settings.aiModel),
      system: settings.customPrompt,
      prompt: "what's the weather like?",
    });

    console.log('ai response:', text);
  }
}

// usage
generateWithScreenpipeSettings().catch(console.error);

import { streamText } from 'ai';
import { ollama } from 'ollama-ai-provider';
import { pipe } from '@screenpipe/sdk';

async function streamWithScreenpipeSettings() {
  const settings = await pipe.settings.getAll();
  
  if (settings.aiProviderType === 'native-ollama') {
    const stream = await streamText({
      model: ollama(settings.aiModel),
      system: settings.customPrompt,
      prompt: "what's the weather like?",
    });

    for await (const chunk of stream) {
      process.stdout.write(chunk);
    }
  }
}

[armanckeser]

Thanks! I will say though, I feel like it might make more sense to expose the already created model from the settings. Just so there is a central place of creation for the Ai model and pipes don't need to write conditionals around the provider