タスク分割

このページでは、あるタスクを小さなタスクに分割するプロンプトを作成する方法を説明します。

ソースコード

task-splitter.py (GitHub)

準備

openai_util.OpenAITextLLM はOpenAITextLLMページ で定義した TextLLM です（参考: TextLLM）。 同じディレクトリの openai_util.py にそれらを定義しておくと、import できます。

import promptogen as pg
from openai_util import OpenAITextLLM
from promptogen.prompt_collection import PromptCreatorPrompt

llm = OpenAITextLLM(model="gpt-3.5-turbo")

formatter = pg.KeyValuePromptFormatter()
prompt_runner = pg.TextLLMPromptRunner(llm=llm, formatter=formatter)

prompt_creator_prompt = PromptCreatorPrompt()

プロンプトの作成

def setup_task_splitter_prompt() -> pg.Prompt:
    input_value = {
        "description": "Split a task into subtasks.",
        "background": "(task: str) -> (strategy: str, subtasks: List[str])",
    }
    resp = prompt_runner.run_prompt(prompt_creator_prompt, input_value=input_value)
    return pg.Prompt.from_dict(resp["prompt"])


task_splitter_prompt = setup_task_splitter_prompt()
print(task_splitter_prompt)
# TaskSplitter: (task: str) -> (strategy: str, subtasks: list)

# Description
# -----------
# Split a task into subtasks.

# Input Parameters
# ----------------
# - task (str): The task to be split into subtasks

# Output Parameters
# -----------------
# - strategy (str): The strategy used to split the task
# - subtasks (list): The list of subtasks generated from the task

# Examples Count
# --------------
# 1

プロンプトの実行

input_value = {
    "task": "Make a system that can remind me to go running every day.",
}

output_value = prompt_runner.run_prompt(task_splitter_prompt, input_value=input_value)

print(output_value["strategy"])
# "Break down the task into smaller steps related to developing the reminder system"

print(output_value["subtasks"])
# ['Design a user interface for the reminder system', "Create a database to store the user's running schedule", 'Develop a notification system to send reminders to the user', "Implement a feature to track and record the user's running progress", 'Test and debug the reminder system for any issues', 'Deploy the reminder system on a suitable platform']