Improve code generation & Fine tuning the model

[zernonia]

I realised that the current multipass approach from https://openv0.com/ might not be the best.

Problem:

1. After generating the code, we didnt let ChatGPT know if the generated code is /
2. The result for follow up iteration might be very different from previous version.
3. The generation is taking quite (as compare to v0)

More reading and research need to be done!! Would love to hear your thoughts! 😁