Next to do:
- Try processing several questions together to save input tokens (i have always attached the doc for EVERY question)
- Try conversation-style generation. Re-feed the error message to the next try (actually these 2 options are kind of mutually exclusive...)
Next to do:
- Log console output to
output
folder