|
|
|
|
|
by joken0x
1575 days ago
|
|
Maybe giving more examples can give a better result, for the case of GPT-3 I kept it to a few examples because of the cost of the amount of tokens. Another option is to improve formatResults to fix more cases of invalid JSON responses. If you get any improvement a pull request is welcome. |
|
I'll come up with a pull request.