search

yeagerai / genlayer-simulator

MIT License
16 stars 8 forks source link

SIM-GH Action to validate PR title and content #423

Closed cristiam86 closed 2 months ago

cristiam86 commented 2 months ago

Actions:

AgustinRamiroDiaz commented 2 months ago

@cristiam86 I think that checking that the PR body complies with the PR template is very hard, since the solutions that I'm thinking of would be very brittle (checking for certain words in the body)

I think that would be a feature we could implement with an intelligent contract to leverage LLMs. I'd leave that part for later

cristiam86 commented 2 months ago

@cristiam86 I think that checking that the PR body complies with the PR template is very hard, since the solutions that I'm thinking of would be very brittle (checking for certain words in the body)

I think that would be a feature we could implement with an intelligent contract to leverage LLMs. I'd leave that part for later

What about a regex? like this one: https://github.com/yeagerai/hackathon-2024-08-16/blob/main/DeepThoughtCoders/adr_validator.py#L68

AgustinRamiroDiaz commented 2 months ago

@cristiam86 I think that checking that the PR body complies with the PR template is very hard, since the solutions that I'm thinking of would be very brittle (checking for certain words in the body) I think that would be a feature we could implement with an intelligent contract to leverage LLMs. I'd leave that part for later

What about a regex? like this one: https://github.com/yeagerai/hackathon-2024-08-16/blob/main/DeepThoughtCoders/adr_validator.py#L68

Well we can do that to check the headers, but I'm not sure if that's valuable. I feel that we'll be leaving empty sections just to adhere with the regex, and that might not be valuable

cristiam86 commented 2 months ago

@cristiam86 I think that checking that the PR body complies with the PR template is very hard, since the solutions that I'm thinking of would be very brittle (checking for certain words in the body) I think that would be a feature we could implement with an intelligent contract to leverage LLMs. I'd leave that part for later

What about a regex? like this one: https://github.com/yeagerai/hackathon-2024-08-16/blob/main/DeepThoughtCoders/adr_validator.py#L68

Well we can do that to check the headers, but I'm not sure if that's valuable. I feel that we'll be leaving empty sections just to adhere with the regex, and that might not be valuable

ok, let's keep it as it is and look for an AI tool then :) something like https://github.com/Codium-ai/pr-agent

btw, having empty sections just with the titles will bring all the shame on you :D