sturdy-dev / codereview.gpt Goto Github PK

How to use this for hosted GitLab PR?

It could be interesting to do same for pipeline in devops (Azure)

I think it would be good if we parse the fetched diff using https://github.com/sergeyt/parse-diff so that each hunk can be send individually, allowing us to send more context to GPT.

It was mentioned before, but maybe there are other prompt engineering techniques we can leverage from https://thakkarparth007.github.io/copilot-explorer/posts/copilot-internals.html.

Suggesting bitbucket integration.

Is there somewhere I can get more info on why it's failing?

Where do I enter the OpenAI API key?
You simply say to enter it in the extension options, but I can't find this option menu anywhere.

Hi,

is there a possibility to integrate this easily also into Gitlab? Thank you.

Some PRs are larger than the limit of tokens. Enable directing the review to only the files the user is concerned about.

We should improve error handling as the extension does not work if the Open AI key was not entered

I would actually really love to be able to run this as a script on my machine before submitting a PR as a gut check

I tried the extension on Gitlab's MR and I got the Error from OpenAI:

Status: 413
{"detail":{"message":"The message you submitted was too long, please reload the conversation and submit something shorter.","code":"message_length_exceeds_limit"}}

From looking at the payload of the request and also at the code itself, it seems that the tool does not capture the code well, it sends too many lines (not only code) in one single request

Wanted to try this extension but I get an unexplained error when running this on my PR (this is in a private Github Repo). I'm already logged into to OpenAI and GitHub. Any idea what the issue might be?

Hi!
I'm worried about proprietary of code.

How does it do stuff at the backend level?
Does it try to git pull, and if so, how will that work for proxy-protected (or private organizations)?
Does it use ChatGPT API ?
If so, doesn't ChatGPT has access to the code as-well?

Cool tool though!

Navigated to pull request and clicked on extension on chrome:

This is a development tool and should be written in a way that exposes all options to the user, rather than simplifying those options.

Allowing customization of the prompt and any logic associated with the prompt for each run would enable a greater depth of review:

• Allows for additional contextual information to be provided on the codebase: language version, standards, etc
• Allows for a faster feedback process with prompts, furthering the speed of the development of this tool
• Allows requesting additional information with the review

If there are many files na changes in the PR (for me it fails with +256 −41 PR), GPT fails with the message:

{"message":"The message you submitted was too long, please reload the conversation and submit something shorter.","code":"message_length_exceeds_limit"}

Maybe it would be useful to split the patch changes and send separate requests for each change?

Hi there!
Just found your extension! Looks incredible !
was wandering if you could share how you managed to pull off gathering data from the pull request and feeding it into chatgpt

I wish to do similar thing but for an entire repo, or just any webpage
I assume that somewhere you have a pre-prompt which says something along the lines of.. "The following are the changes from a pull request, please review it: "

So for general sites you would have some like "The following is the contents of a webpage, get all the information you can from it as i will be asking you questions about it shortly"

For a repo i would like "Based on the Readme of this repo, write a full project that answers the authors wishes. Pick the best technology for the use case. Give me all the files I will need (and their contents) for to fill this repo. The repo should be easily deployable. Generate all the necessary test. etc

Would love to know what you think
Thanks a lot and have an amazing day!

Whatever I try I'll be getting: Please login at chat.openai.com first.

It seems like it's starting the review, but after 2/3 sentences it continue with loading and stops generating text

As an accomplished open-source project developer, I find it my obligation to address a pressing concern pertaining to the User Experience (UX) encountered within the environment. The essence of this issue is intrinsically tied to a fragile User Interface (UI), manifesting in the form of an operational window. A poignant drawback emerges as this window, devoid of forewarning, abruptly terminates upon the user's interaction with any region lying external to its confines. This includes seemingly innocuous actions such as clicking anywhere outside the window, even within the browser bookmarks. This lamentable occurrence engenders the untimely forfeiture of valuable data, rendering the user's efforts and inputs naught.

Steps to Reproduce:

1. Launch the application or designated interface.
2. Initiate a sequence of prompts or actions within the UI.
3. Progress to a point where user inputs or data are solicited.
4. Attempt to interact with any region outside the operational window.
5. Observe the window's unexpected and unannounced closure.
6. Realize that the data, painstakingly inputted thus far, has been irrevocably lost.

Expected Behavior:

The UI should ideally incorporate a more robust design, wherein the operational window remains impervious to inadvertent closure triggered by interactions external to its boundaries. This is essential to preserve user data and ensure a seamless and uninterrupted workflow.

Proposed Solution:

To address this concerning deficiency in UX, a couple of potential solutions are worth considering:

1. Enhanced Window Behavior: Modify the UI to ensure that the operational window remains open despite interactions outside its confines. This will prevent inadvertent closure and data loss.

2. Background Operation Persistence: Implement a mechanism whereby the extension continues its operation in the background, allowing users to interact with other elements without risk of losing data. Upon returning to the operational window, users can seamlessly resume their tasks.

Additional Notes:

While it is noteworthy that the issue has been observed on a private, corporate, self-hosted GitLab instance, the focus here is on the broader concept of User Experience design and its implications. The mentioned GitLab instance serves merely as a reference for context.

I look forward to collaborative efforts to resolve this UX predicament, thus contributing to the enhancement of our product's overall usability and appeal. Your consideration and attention to this matter are greatly appreciated.

I wasn't able to get the extension working on a private instance of GHE.

Would be nice to tell GPT to respond in a given language.