It's interesting that the screenshots advertising the bot make it seem totally useless. In the first example the bot comments "You should add tests for this function". The author asks "do I really need to?" and the bot replies "No actually you don't, just make sure it works". In the second example the author asks the bot for a review, and it just copies and pastes the code change and says basically "looks good to me".
I suppose the productivity improvement comes from not having to wait for those same lines from a colleague?
I agree with you, I think the examples need to show the full potential of the bot. This being based on ChatGPT, the feedback can really be astonishing if the use case makes sense.
While I don't necessarily agree with the somewhat-uncharitable root level comments currently around here, I think it's probably worth replacing those screenshots with more complex examples. If one advertises the bot as being able to "provide insightful responses to comments", one should probably demonstrate that. Like, these relatively-small, "human readable" commits aren't really where something like ChatGPT would shine, right?
ChatCody is a new Github bot that utilizes the latest GPT model to streamline the code review process and boost team productivity. The bot can perform a variety of tasks to help software development teams save time with comments on both pull requests and issues.
Seems way too wordy. I hate automatically generated code review feedback in PR comments.
If the feedback is valuable, I should be able to identify and fix it before anyone else needs to waste their time reviewing it.
I think the bot should avoid unnecessary comments on every change.