Coder Agent Improvements

[0Xuser100]

Coder Agent Improvements

This PR introduces several key improvements to the coder_agent.py script to make it a functional and robust CLI tool.

1. Fix: Infinite Loop Termination

Problem: The agent would enter an infinite loop because is_goal_achieved logic failed when tool_calls was present but empty/None (a behavior common in newer OpenAI SDK versions).
Fix: Updated the termination condition to robustly check for falsy tool_calls:

and not last_message.get("tool_calls")

2. Enhancement: Visible Assistant Responses

Problem: The agent's text responses were being generated but not printed to the console, leaving the user to see only "Thinking..." steps.
Fix: Added a print statement to output completion.choices[0].message.content when present.

3. Enhancement: Continuous Chat Loop (REPL)

Problem: The script would exit after a single turn, forcing the user to restart the process for every follow-up question.
Fix: Wrapped the interaction logic in a while True loop to allow for continuous conversation. Added an exit condition (exit/quit) and KeyboardInterrupt handling.

Verified By

• Verified the agent correctly terminates "thinking" loops when no tools are called.
• Verified assistant text replies are visible in the console.
• Verified the agent allows multiple conversational turns and exits gracefully on "quit".