The Automated agent_loop() Controller

~20 min read

Replace the human controller with a Python function that parses the agent's Thought/Action/Answer output, runs the right tool via regex extraction, and keeps looping until a final Answer appears.

The Automated agent_loop() Controller is a Pro topic

Sign in, then upgrade to Pro or Power to unlock this topic and the full AI Engineering curriculum.

Key points

  • agent_loop() automates exactly what the human did in the manual trace: read output, act, feed back an Observation, repeat
  • current_prompt and previous_step are the loop's state — tracking what to send next and what stage was last seen
  • A regex extracts the tool name and argument from lines like 'Action: lookup_population: India'
  • The tools dictionary's keys must exactly match what the model writes in its Action lines, or the lookup fails
  • An 'Answer:' line in the response is the signal to stop looping and return the final result