Designing prompts for complex task decomposition
To enable effective multi-step reasoning, prompts should guide the LLM to break down complex tasks into smaller, manageable steps. Here’s an example of a task decomposition prompt:
def task_decomposition_prompt(task, available_tools):
prompt = f"""Given the following complex task:
{task}
And the following available tools:
{' '.join(f'- {tool}' for tool in available_tools)}
Please break down the task into smaller, logical steps. For each step, indicate if a specific tool should be used. If no tool is needed, explain the reasoning required.
Your task decomposition:
Step 1:
Step 2:
Step 3:
...
Ensure that the steps are in a logical order and cover all aspects of the task.
"""
return prompt
# Example usage
task = "Analyze the sentiment of tweets about a new product launch and create a summary report with visualizations."...
