Devin's use in improving Team Mirai policy
2025-05-22
Team Mirai Using Devin in Policy Improvement
I'll summarize it in bullet points as it might make an interesting case study for a Note article later.
There are 1400+ PRs for the policy
https://github.com/team-mirai/policy
Idombata policy making Effect of allowing modules to be proposed even by those who cannot use GitHub directly.
If we try to analyze all of these cases, we will hit the Rate Limit (5000 cases / hour) of the Github API.
If you ask Devin to do it, he tries his best to work with a web browser.
Wrote code to get hourly up-to-date data and convert to JSON using Github Actions.
https://github.com/team-mirai/random/tree/main/pr_analysis
Now you can get 1,000 PRs per hour.
Data is placed here.
https://github.com/team-mirai/policy-pr-data
By instructing Devin in Japanese, I was able to use this data to extract specific conditions from all PRs and create a CSV
Instruct from Slack and have the resulting CSV attached to Slack.
Results were obtained in two minutes.
This has created a system where non-engineers can direct "the type of analysis that requires some implementation" 24 hours a day and get results in a matter of minutes.
Team Mirai 2025-05
---
This page is auto-translated from /nishio/チームみらい政策改善でのDevinの活用 using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I'm very happy to spread my thought to non-Japanese readers.