You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Post auto-patching, we want an agent to to run tests/validations optionally
The first simple version is to define it in the config and run them.
The second phase is, if that failed, to feed the errors back to an agent to fix the code.
The third phase could include detecting what to run alone based on the README (or code), here is a sample prompt that seems to work:
What commands can I use to run things in this project (multiple commands ok. both testing and normal runs. only commands that appear above. Ignore package installs.)
Please return a valid JSON format that includes each command, example format
{
"commands": [
"command 1",
"command 2",
]
}
Code only. Don't explain. No comments.
Post auto-patching, we want an agent to to run tests/validations optionally
The first simple version is to define it in the config and run them.
The second phase is, if that failed, to feed the errors back to an agent to fix the code.
The third phase could include detecting what to run alone based on the README (or code), here is a sample prompt that seems to work:
Returned:
The text was updated successfully, but these errors were encountered: