Spike implementation of dwylbot

[SimonLab]

• How to capture events from Github: new PR, new issues, changes label, change assignees...?
• How to listen to these event with Elixir/Phoenix?
• Do we need an authentication process to capture Githbub events?

Implement a prototype of the dwylbot to get a better idea of the technical challenges
Create implementation doc for creating the first dwylbot version

[SimonLab]

some links to read:

• Github webhook API doc
• Github webhook Guide
• Elixir wrapper for Github API, Tentacat
• Elixir package to manage webhooks, Grapple
• [Elm process-bot] (https:/process-bot/contribution-checklist)
• PubSubHubbub
• Phoenix & webhooks discussion
• Simpler alternative to phoenix for web-hooks endpoint?

[SimonLab]

So:

• 1. Create manually the webhooks (go to settings then webhooks) on the repository (or organisation). We can later on do this step automatically with the Github API but for a first version the UI provided by Github to add webhooks is good
• 2. Create a server with at least a post route (ex: /webhooks) which will get the payloads send by the webhooks. I've manage to get this quickly done with [SimpleServer]. (https:/SimonLab/simpleServer) and ngrok (to expose my localhost server to the Github webhook instead of using Heroku)
     
• 3. create the logic of the server. This is the main work to be done. The logic of the rules from Actions that can trigger the dwylbot #5 needs to be implemented and triggered depending of the action type received on the payload of the post request.
• 4. Use Github API to post comment, change assignee, change label... to indicate how to follow the contributing guide

So Phoenix Server + Heroku + Github webhooks for the first version.

Once the main parts of the application are done (server, tests, doc, heroku app), the idea is to have a process or architecture that allow us to create new rules quickly and easily. I will focus on this once I finish #3 as it will give me a better idea on how to abstract rules.

@iteles @nelsonic @markwilliamfirth any comments or more details you would like to see (I will add more issue when I start on #3 that will explain the implementation)?

[nelsonic]

@SimonLab this is an excellent start! (well done for spiking the web hook so quickly!) 🐎
Keep going! it sounds like you should have something useful soon! 👍
I will suggest that you start storing some of the GitHub web hook request payloads somewhere (your choice. e.g: .json on your computer or in a DB if you can) so that you can re-play them as "mock" data for tests. 😉
Thanks for doing this! You're a hero! ❤️ ✅ 🚀

[iteles]

@SimonLab I'd love to have an issue in learn-phoenix about using APIs and the need for tentacat ❤️

[nelsonic]

@SimonLab do we have a JSON file with a sample GitHub Webhook request payload?

[SimonLab]

I've create a few json files in the project using the issue test

[nelsonic]

@SimonLab yeah, sorry I saw the https:/dwyl/dwylbot/tree/master/github_webhooks folder shortly after asking the question and forgot to update this issue to inform you that I knew.
Thanks for your great work on this! ❤️ ✅

[ghost]

@SimonLab what's the next steps with this issue? The implementation documentation?

[ghost]

@SimonLab can this be closed?

[SimonLab]

We are now using Github App which give us webhooks and a way to install dwylbot via the Github marketplace