https://www.anyscale.com/blog/building-an-llm-powered-github-bot-to-improve-your-pull-requests
Automatically get suggestions to improve the writing in your PRs from this GitHub app powered by Anyscale Endpoints.
Simply install the app on your project on GitHub: Docu-Mentor App
Then in any PR in your project, create a new comment that says:
@docu-mentor run
and I will start my analysis. I only look at what you changed in this PR. If you only want me to look at specific files or folders, you can specify them like this:
@docu-mentor run doc/ README.md
In this example, I'll have a look at all files contained in the "doc/" folder and the file "README.md".
Simply run
pytest . -s
to evaluate the doc sanitation bot against GPT-4.
The bot is built for deployment with Anyscale Services,
using the service.yaml
with the main.py
entrypoint provided in this repo.
You can also deploy it on Heroku, using the Procfile
and heroku.py
.
With Anyscale you can parallelize your bot using Ray, reducing the total
wallclock time, with Heroku that doesn't work.