zeke / yolox

🤞 Use AI language models to write one-liner shell commands, then execute them.
14 stars 3 forks source link

Keep a papertrail of results #10

Open zeke opened 2 weeks ago

zeke commented 2 weeks ago

Sometimes the commands work, sometimes they don't.

It could be interesting to stuff all the inputs, outputs, and exit codes of yolox invocations in a directory like ~/.yolox/{uuid}.json or similar, then later use that to help diagnose and improve the system prompt...