SelfInstruct

SelfInstruct is a Singleton task that generates user queries for a given AI application and context.

Inputs

• num_instructions (int): The number of user queries to generate.
• criteria_for_query_generation (str): The criteria for generating the queries.
• application_description (str): A description of the AI application.
• context (str): The context to which the queries should be applicable.

Outputs

• instructions (List[str]): The generated user queries.
• model (str): The model used for generation.

Example

Generate user queries for an AI application. This example uses the GEMMA2_9B_FP16 model.

import os
import asyncio
from dria.factory import SelfInstruct
from dria.client import Dria
from dria.models import Task, Model

dria = Dria(rpc_token=os.environ["DRIA_RPC_TOKEN"])

async def evaluate():
    self_instruct = SelfInstruct()
    res = await dria.execute(
        Task(
            workflow=self_instruct.workflow(
                num_instructions=5,
                criteria_for_query_generation="Diverse queries related to task management",
                application_description="A task management AI assistant",
                context="Professional work environment"
            ),
            models=[Model.GEMMA2_9B_FP16],
        ),
        timeout=45,
    )
    return self_instruct.parse_result(res)

def main():
    result = asyncio.run(evaluate())
    print(result)

if __name__ == "__main__":
    main()

Expected output

{
   "instructions":[
      "Prioritize my upcoming deadlines, considering project dependencies. ",
      "Can you schedule a meeting with the marketing team for next week to discuss the Q3 campaign?",
      "Generate a comprehensive list of actionable steps required for completing the client proposal.",
      "What tasks are currently assigned to me that are due within the next 7 days?",
      "Remind me to follow up with John about the budget approval at 2 PM tomorrow."
   ],
   "model":"gemma2:9b-instruct-fp16"
}