Design a minimal Python agent loop on paper

Single-goal agent loop structure

You’re designing a tiny “agent” that chugs along toward one clear goal, step by step, in Python.
On paper, that means: list out the steps in words, then turn each step into a Python operation.

A simple mental model:

Start with a goal and an empty history.
Look at the goal and the history.
Decide what to do next.
Call a tool (a helper function) to do that thing.
Record what happened in the history.
Check if you’re done; if not, go back to step 2.

Let’s write this as a small, runnable-style skeleton (no real APIs):

python

def agent_loop(goal):
    # 1. Initialize state
    history = []  # will store steps like dicts or strings
    step = 0
    max_steps = 5

    # 2. Loop until we succeed or hit limits
    while step < max_steps:
        step += 1

        # 3. Decide what to do next based on goal + history
        # For now, we hardcode a simple “plan”
        if step == 1:
            action = "search"
        else:
            action = "summarize"

        # 4. Call a tool based on the action
        if action == "search":
            result = tool_search(goal)
        elif action == "summarize":
            result = tool_summarize(history)
        else:
            result = "unknown action"

        # 5. Record the step into history
        history.append({
            "step": step,
            "action": action,
            "result": result,
        })

        # 6. Check for stopping condition
        if goal_satisfied(goal, history):
            break

    # Return whatever final answer we think we have
    final_answer = extract_answer(history)
    return final_answer

On paper, your job is to sketch something like this:

A single function agent_loop(goal)
A loop (while or for) that:
- Chooses an action
- Calls a tool
- Logs result into history
- Checks if the goal is satisfied or if you hit a limit

You don’t need real logic for planning yet; just fake it with if step == 1: ....

1 / 4