b9547c2e6b
This PR switches up how PR labels are validated to allow for more freeform label tweaking in the future. Basically **huppy will now only check that your PR is labelled, it doesn't care how it's labelled**. I also updated the PR template with a new labelling scheme that we can tweak over time. So before Huppy bot had to know about the specific set of allowed labels, and now as long as the label exists you're allowed to add it. So to add a new label to the PR template, just create the label and then add an option for it in the .md file. ### Change Type - [ ] `patch` — Bug fix - [ ] `minor` — New feature - [ ] `major` — Breaking change - [ ] `dependencies` — Changes to package dependencies[^1] - [ ] `documentation` — Changes to the documentation only[^2] - [ ] `tests` — Changes to any test code only[^2] - [x] `internal` — Any other changes that don't affect the published package[^2] - [ ] I don't know
2.2 KiB
Repo-tools
Repo-tools is responsible for the huppy-bot app.
Development
To develop huppy-bot, you'll need to create a .env file that looks like this:
REPO_SYNC_PRIVATE_KEY_B64=<base64 encoded private key here>
REPO_SYNC_HOOK_SECRET=<hook secret here>
DM alex to get hold of these credentials.
To start the server, run yarn dev-huppy. Once running, you can go to
http://localhost:3000/deliveries to get to a list of github webhook event
deliveries. To test your code, pick an event that does roughly what you want and
hit 'simulate'. You can also ask GitHub to re-deliver events to the production
version of repo-sync through this UI.
Huppy-bot isn't currently deployed automatically. To deploy, use:
fly deploy --config apps/huppy/fly.toml --dockerfile apps/huppy/Dockerfile
from the repo root.
How it works
Huppy runs on a server with persistent disk storage attached. It maintains local mirrors of both our github repos on that disk. When events come in that mean we need to do some work in a repo, it updates the local mirror, then clones them to a temporary directory for work. This sort of pull + local clone is much faster (~1s) than normal from-scratch clones (~1m).
Huppy's reponsibilities are organized into "flows". These are defined in
src/flows. A flow is an object with webhook handlers that implement some
complete set of functionality. Right now there aren't many, but we could add more!
There's an alternative universe where huppy would exist as a set of github actions instead. We didn't pursue this route for three reasons:
- Huppy needs to operate over multiple github repos at once, which isn't well supported by actions.
- Giving actions in our public repo access to our private repo could be a security risk. We'd have to grant permission to OSS contributors to run certain actions, which could mean accidentally giving them access to more than we intend.
- Having access to the full range of webhook & API options provided by GitHub means we can create a better DX than would be possible with plain actions (e.g. the "Fix" button when huppy detects that bublic is out of date).
It also lets us make use of that local-clone trick, which means huppy responds to requests in seconds rather than minutes.