Not getting correct outputs

[KananVyas]

Hi,
Great work on mimicking o1 CoT type of output. I have used your code and generated my own Gemini Flash API key, and using your code as it is, I'm not getting similar results to what you've shown in the videos, for the query such as "finding the cities containing 'a' as third spot?"

I also found there's a code difference between your video and the github code. I see in your source code, the prompt is quite huge" where in agent.py it is like this:

        if previous_solution:
            messages.append({"role": "user", "content": f"Problem:\n{problem}\n\nPrevious Solution:\n{previous_solution}\n\nPlease proceed with your analysis."})
        else:
            messages.append({"role": "user", "content": f"Problem:\n{problem}\n\nPlease provide your solution."})

Can you explain why there's a different?