How often do you find yourself writing a

How often do you find yourself writing a flow that is basically a very lightweight wrapper around a prompt/generate call? Where the

inputSchema

inputSchema

inputSchema

inputSchema are variables that affect the text of the prompt, and the

outputSchema

outputSchema

outputSchema

outputSchema is the output schema of the prompt (maybe with some post-processing)?

I'm wondering if it might make sense to have prompts be more explicitly executable:

const myPrompt = definePrompt({...});

await myPrompt({input: 'variables'});

// above is equivalent to...
const result = generate(await myPrompt({input: 'variables'});
return result.output();

const myPrompt = definePrompt({...});

await myPrompt({input: 'variables'});

// above is equivalent to...
const result = generate(await myPrompt({input: 'variables'});
return result.output();

const myPrompt = definePrompt({...});

await myPrompt({input: 'variables'});

// above is equivalent to...
const result = generate(await myPrompt({input: 'variables'});
return result.output();

const myPrompt = definePrompt({...});

await myPrompt({input: 'variables'});

// above is equivalent to...
const result = generate(await myPrompt({input: 'variables'});
return result.output();

a prompt would also have

myPrompt.render()

myPrompt.render()

myPrompt.render()

myPrompt.render() which would render into generate options like it does today, and

myPrompt.generate()

myPrompt.generate()

myPrompt.generate()

myPrompt.generate() which would not automatically return the output like it does today

How often do you find yourself writing a

Similar Threads

Similar Threads

Similar Threads