Skip to content

Latest commit

 

History

History
53 lines (39 loc) · 2.79 KB

creating-custom-actions.md

File metadata and controls

53 lines (39 loc) · 2.79 KB
Raw

Create Your Own Bot Action

The initial version of the bot ships with actions that allow you to monitor the status of Metaflow runs, past and present. Imagine starting a training run that lasts for hours - you can now monitor it anywhere using Slack on your mobile device! You can converse with the bot over direct messages or or invite the bot to a channel.

With the diversity of machine learning and data science use cases, we have wanted to make it easy to extend the bot with new actions. As an example, we ship a simple jokes command 🤡. You can use it as a template to create custom actions which can be enabled just by doing a pip install. We would love to see people contributing actions of all kinds - please share with us (http://slack.outerbounds.co) if you have any ideas or prototypes!

How To Create Your Own Bot Action

Create your own custom action by creating a Python package with the following folder structure -

your_bot_action/ # the name of this dir doesn't matter
├ setup.py
├ metaflowbot/ # namespace package name
│  └ __init__.py       # special pkgutil namespace __init__.py
│  └ action/      # namespace sub package name
│      ├__init__.py    # special pkgutil namespace __init__.py
│      └ your-special-acton/            # dir name must match the package name in `setup.py`
│        └ __init__.py  # Contains a prespecified code block given below
│        └ rules.yml.   # This mandatory to create rules
│        └ commands.py. # This create main commands from click
.

Every module must contain a rules.yml, a __init__.py, and a module that contains click commands imported from metaflowbot.cli.actions. Every rule in the rules.yml should follow this structure

The __init__.py inside an installable action must contain the following code to integrate with metaflowbot's actions

import pkgutil

from metaflowbot.rules import MFBRules

data = pkgutil.get_data(__name__, "rules.yml")
RULES = MFBRules.make_subpackage_rules(data)
from . import commands

Locally Development

Export the slack tokens as environment variables :

export SLACK_APP_TOKEN=xapp-1-AAAAAAAAAAA-2222222222222-AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
export SLACK_BOT_TOKEN=xoxb-2222222222222-2222222222222-AAAAAAAAAAAAAAAAAAAAAAAA
  • Install metaflowbot repository and install your custom development repo; ensure you have set the PYTHONPATH correctly so that metaflowbot can be resolved.
pip install metaflowbot
pip install -e ./<PATH_TO_metaflowbot_action_directory>
  • If you are running the bot locally with a local metadata provider, then run the above command inside the directory where the .metaflow folder is present.