I want to stop comments under certain conditions.

[yamamoto-ehhMH]

Please review for the following issues:
Bugs
N+1 queries
Possibility of refactoring
Only comment if there is a clear issue. Otherwise, just respond with 'no comment.'

I'm writing a prompt like this, but I want to prevent "no comment" from being included as a comment.

Is it possible to set it so that if the phrase "no comment" is included, no comment is made at all?

If there are other ways to stop specific comments, please let me know.

[anc95]

@yamamoto-ehhMH it makes sense, I can do this by the end of this week. stay tuned

[yamamoto-ehhMH]

@anc95

Thank you for your prompt response.
I’m very excited to hear that the implementation will be ready by this weekend. Looking forward to it!

[anc95]

@yamamoto-ehhMH hello, just supported this in the latest version
like the comment, if all reviews are good, will response with a LTGM 👍 message

[yamamoto-ehhMH]

@anc95

Thank you for your prompt response! I'm very happy.

However, I may not have explained myself clearly.
Would it be possible to disable bot comments altogether?

Below, I’m sharing my current configuration file.
The issue I'm facing is that ChatGPT responds even when there are no issues in the code.

I was considering the following approach as a possible solution:

Instructing ChatGPT to return "no comment" when there are no issues
Configuring the bot to check if the response contains "no comment" and, if so, prevent it from being posted as a GitHub comment

You are already doing an amazing job, and I feel sorry for making additional requests...

name: Code Review

permissions:
  contents: read
  pull-requests: write

on:
  pull_request:
    types: [opened, reopened, synchronize]

jobs:
  test:
    if: ${{ (github.event.pull_request.base.ref == 'main' && github.event.pull_request.head.ref != 'develop') || (github.event.pull_request.base.ref == 'develop' && github.event.pull_request.head.ref != 'main') }}
    runs-on: ubuntu-latest
    steps:
      - uses: anc95/ChatGPT-CodeReview@main
        env:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
          # Optional
          LANGUAGE: Japanese
          OPENAI_API_ENDPOINT: https://api.openai.com/v1
          MODEL: gpt-4-turbo-2024-04-09 # https://platform.openai.com/docs/models
          PROMPT: |
            以下の点で問題があるかどうかレビューしてください。
            - バグ
            - N+1
            - リファクタできるか

            明確な問題がある時だけコメントしてください。
            それ以外は絶対に何もコメントしないでください。

          top_p: 1 # https://platform.openai.com/docs/api-reference/chat/create#chat/create-top_p
          temperature: 1 # https://platform.openai.com/docs/api-reference/chat/create#chat/create-temperature
          max_tokens: 4096
          MAX_PATCH_LENGTH: 10000 # if the patch/diff length is large than MAX_PATCH_LENGTH, will be ignored and won't review. By default, with no MAX_PATCH_LENGTH set, there is also no limit for the patch/diff length.
          # IGNORE_PATTERNS: /node_modules/**/*,*.md # glob pattern or regex pattern to ignore files, separated by comma
          INCLUDE_PATTERNS: "*.html,*.js,*.css,*.scss,*.sass,*.rb,*.erb,*.rake" # glob pattern or regex pattern to include files, separated by comma

[anc95]

hi @yamamoto-ehhMH I think I understand your meaning. and my changes already met our requirements.
let's say there are 2 files commited and pushed in this pr

• if GPT thinks both of them are good. would only send a LTGM 👍 comment (meaning looking good to me, you don't need to resolve it), it's also a way to differ LGTM and the action not working, so will at least send one comment even there is no comment as you said in your case.
• if GPT thinks one of them is good, another is not good (with review comment), then will only do the code review to one file (in previous version. it will send the review comment for both files.)

does it make sense? looking forward to your more feedback

[yamamoto-ehhMH]

@anc95

Thank you for your polite explanation!

I understand that this bot takes a prompt written in a YML file, inputs it into ChatGPT, and inserts the response as a comment. In this case, what would be the ideal way to write the prompt?

Also, how can I determine whether the GPT has judged the response to be appropriate?

I apologize for my lack of understanding 🙏

[anc95]

@yamamoto-ehhMH

your prompt in yaml file should focus in code review work, I will append a prompt to ensure GPT responds a json format result to indicate whether 'lgtm'. you can refer to this code for details

[yamamoto-ehhMH]

@anc95

I see!
The code you shared helped me understand how it works.

I'll try it out right away and report back to you later.

I sincerely appreciate your prompt response once again.

[yamamoto-ehhMH]

@anc95

Sorry for the late reply
It looks great! Thank you so much!