5.5 KiB
Misskey-ebooks-bot
Misskey eBooks Bot with Markov Chain
Introduction
This small python script is a Markov Chain eBooks bot based on the framework of mi.py
It can only read and write from and to Misskey. Reading from Mastodon or Pleroma is not (yet) implemented.
It posts every hour on his own and reacts to mention. Every 12 hours the bot reloads the notes and recalculates the Markov Chain.
Operating mode
On the first start up the bot loads a given number of posts into his database and calculates the Markov Chain out of it. After this he only updates the database with new posts. The upgrading is threaded so the bot itself isn't interrupted while the new markov chain is calculated.
Installation
Host Installation
To run mi.py you must install python3.9 and python3.9-dev onto your system. (Please be aware of the requirements for mi.py!)
mi.py is still under development and a lot of things change there quickly so please be aware that there could be chances that something changed, that I haven't implemented in the bot at the moment.
to install mi.pyplease use the following command.
pip install git+https://github.com/yupix/Mi.py.git
For the bot to run you also need two additional packages
markovify
configparser
regex
or just use the command pip install -r requirements.txt in the local folder where you cloned the repo.
Before starting the bot, please copy example-bot.cfg to bot.cfg and
configure it according to the configuration section below.
The best way to run it would be a systemd unit file and run it as a daemon.
Just to test it you can use nohup python3.9 rdbot.py & in the directory the bot is located in.
Docker
To host this image with docker, copy the docker-compose.yml file to the directory that you want to host it from.
Next, you'll need to copy the contents of example-bot.cfg to bot.cfg in the
same directory and configure it according to the configuration section below.
Run touch markov.json roboduck.db in order to create the markov and database
files before starting the docker container. These files must already exist
before starting the docker container.
Then, simply run docker-compose up to start the app, or docker-compose up -d
to start the bot in detached mode!
You can find the bot on Dockerhub just use docker pull fotoente/misskey-ebooks-bot:latest
Configuration
Following things can be edited:
| Name | Values | Explanation |
|---|---|---|
| users | @username1@domain1.tld;@username2@domain2.tld |
The users you want to read the notes from, separated by semicolon (;). For single user just provide one username@domain.tld |
| instance_write | domain.tld | Put here the domain of the Misskey instance your bot is running. Only domain name and TLD, no /,: or https |
| token | String |
The token from your bot. Needs right to write notes and read notification |
| cw | String or None |
If the markov bot post should be posted with a content warning. Any String given here will show up as CW text. "none" to deactivate. |
| min_notes | interger |
How many posts should be read at the initial start. Please state a number in 100 increments. Higher number means more variety but also more load when loading those and a bigger database and json file. 5000 notes resulted in ~3 MB disk space used. Default 5000 |
| max_notes | interger |
How many posts should be stored in the database. Everything over this number will be deleted during an update cycle Default 5000 |
| includeReplies | boolean |
Should reply included into the markov chain? Default True |
| includeMyRenotes | boolean |
Should the notes you renoted be included? So your bot will use sentences for the markov chain you haven't written. Default false |
| excludeNsfw | boolean |
Should be Notes included that are behind a CW Tag? Default False (Use with caution! The bot not CW any post he makes!) |
| test_output | boolean |
Should be the created sentence be tested against the following statements? Default true (Highly recommended, otherwise sentences could repeat itself, could be very short or very long.) |
| tries | integer |
How many times the bot tries to make a sentence that meets the given criteria |
| max_overlap_ratio | float |
How many percent of the created sentence is allowed to be the same as the source text. Lower value means more gibberish, higher value to more exact copies of source material. Can be between 0and 1. Default 0.6 |
| max_overlap_total | integer |
How many words are allowed to be the same? Default 15 |
| min_words | integer |
How many words the sentence must have at least. Default None |
| max_words | integer |
How many words the sentence must have at maximum. Default None |
Changes to the Markov chain section of the .cfg-file will have immediate effect. Changes to the misskey part of the *.cfg-file, requires a restart of the bot.
Known Quirks
- The startup needs quite some time. On my system about 10 seconds. You know that everything runs well when the first Note is posted.
- When the bot is started, he runs in a timeout in the first 60 seconds. To prevent that, just mention the bot, and he will stay in a loop.
Contributors
Shibao - Docker support and bug hunting
Yupix - Mi.py framework and clean code
Nullobsi - Added multi-user support
ThatOneCalculator - Option to CW the posts
Thank you very much! Without your this project wouldn't be on this level!