This repository implements a smart predictor to automatically categorize new beancount transactions with a model learned from historic data. It works by comparing the description of each new transactions with previous ones and assign a predicted category to each new transaction. The description of each transaction is supposed to be imported automatically from your bank.
Please note: this project is currently under development.
Poetry is used to manage the dependencies of this project. To install poetry, please refer to the official documentation.
For the following commands, we expect poetry to be available in your shell.
To use this tool, you should have an existing beancount file where all your historic beancount records are saved, and some new transaction files (can be .csv, .xls, etc depending on your bank) downloaded from your bank portal.
Then, call the beancount importer with BeanBot as a plugin:
poetry run bean-extract tests/data/import.config -e tests/data/main.bean tests/data/raw
You can find currently implemented importers in the [src/beanbot/importer] directory and you are welcomed to submit a PR if the importer for your bank is missing.
Customization can be done by changing the configuration file import.config, where beanbot can be applied as a hook to the beancount importer:
CONFIG = [
apply_hooks(deutsche_bank.Importer('Assets:Checking:DeutscheBank'), [BeanBotPredictionHook()]),
]
This will call the BeanBotPredictionHook after the importer has finished importing the transactions. The hook will then predict the category of each transaction and will return balanced entries.
poetry run pytest tests/Chengxin Wang w@hxdl.org