This is the officially support Keybase Python library. It is an unopinionated, simple wrapper around the Keybase CLI API for creating an interactive bot or general scripting. This library does not attempt to do intent parsing or manage state at all. You'll have to build that yourself, but with the examples, this library will hopefully make whatever you want to build much much easier
This library is very far from exhaustively covering the complete Keybase API, but it is our hope that it will be easy to add to (see chat_client.py for the pattern). It currently does reading from channels and writing messages/reactions pretty well. That's enough for the vast majority of basic functionality. Future work can add teams behavior, more wallet functionality (e.g. sending money), ...
pip install pykeybasebot
Python 3.7 or greater, please. And it's all async, so you'll need to call into it with that in mind.
Generally speaking, here's what you need to do:
- Create a handler function that takes event objects and does something with them. This function will get called with your bot instance (described below) and the KbEvent instance.
- Create a bot. You must initialize this with the handler function to call with each event. You may optionally pass in: (1) the username and/or paperkey for the bot's identity (it'll default to the currently logged-in user otherwise), (1) the event loop that you want new tasks to be sent to (this is necessary if you want to lock on async behavior -- see the examples), (2) the location of the running keybase app (defaults to
keybasewhich is fine if it's in your PATH), your user's home directory, or pid_file. These three are more useful for complicated local development with multiple accounts and less useful if you're running in a docker container or as the only user on your system.
- If you are not already running on a logged-in device, you need to do that. We recommend doing this with the
oneshotcommand. It's in the examples.
- start the bot inside the asyncio event loop. This bot command wraps
keybase chat api-listen, (and it takes basically the same exact options) and fires off events to your handler function.
Definitely definitely check out the examples. We're really counting on them to make it clear how to use this library.
PRs are extremely welcome. To start:
git clone https://github.com/keybase/pykeybasebot cd pykeybasebot
We use Poetry to handle our packaging. Go check out their website for installation instructions. To start Poetry, you'll need the
python executable in your path to link to Python 3.7. We recommend using pyenv to handle different versions of Python on your machine. With pyenv installed, it should automatically set
python to 3.7 when you
cd into this repo.
Once you have the right Python version, you can run:
pip install poetry poetry install
This will set up a virtualenv for you and install all the dependencies needed into it!
Static code analysis tools
We use a few different static analysis tools to perform linting, type-checking, formatting, etc. The correct versions should be install when you run
poetry install, but you'll probably want to configure your editor to work with:
We check all git commits with the above tools with pre-commit.com pre-commit hooks. To enable use of these pre-commit hooks:
- Install the
- Remove any existing pre-commit hooks via
- Configure via
Then proceed as normal.
To run tests, type
poetry run python -m pytest
Tests are admittedly weak. You could change that!
Most of the types the bot uses are generated from definitions defined in the
protocol/ directory inside the Keybase client repo. This ensures that the types that the bot uses are consistent across bots and always up to date with the output of the API.
go get github.com/keybase/client/go/keybase
cd client/protocol yarn install
Then you can generate the types by using the provided Makefile in this repo.
cd path/to/keybase-bot make
Should you need to remove all the types for some reason, you can run
Poetry can build and publish packages to PyPI. We've run into some issues with uploading to PyPI and Poetry, though, so for now we're recommending building with Poetry and uploading with Twine.
poetry build # Upload to Test PyPi. You only need to run the first command once poetry config repositories.testpypi https://test.pypi.org/legacy/ poetry publish -r testpypi # Upload to real PyPi poetry publish