Markov chain IRC bot and PoC exercise in golang and Redis

• Background
  • Markov Chains
• Usage
  • Standalone
  • Docker
  • Populating Corpus
• License

Meowkov is a simple Markov chain IRC bot. Created as a learning exercise in golang and Redis.

• Markov Chains at Wikipedia
• Markov Chain Explained Visually

To start your own instance:

1. Clone the repo: git clone https://github.com/lidel/meowkov.git
2. Copy meowkov.conf.template to meowkov.conf and change at least BotName, Channels and RedisServer
3. Run make build to build meowkov binary
4. Run ./meowkov
5. That is all: meowkov bot will join specified room after a few seconds

• make build builds the app
• make dev-updatedeps updates dependencies to latest versions
• ./meowkov runs the app against meowkov.conf in current directory
• ./meowkov -c /some/path/meowkov.conf runs the app with specified config file
• echo "some text" | ./meowkov -import=true -purge=false adds piped strings to the corpus
• echo "some text" | ./meowkov -import=true -purge=true replaces corpus with piped data (destructive, remember to backup Redis database before executing this)

To start dockerized instance with latest Redis:

1. Clone the repo: git clone https://github.com/lidel/meowkov.git
2. Copy meowkov.conf.template to meowkov.conf and change at least BotName and Channels (RedisServer will be set automatically by Docker)
3. Run make docker-rebuild to build (in foreground) and run (in background) via Docker container
4. That is all: meowkov bot will join specified room after a few seconds

Makefile provides commands that simplify build and deployment.
Commands assume meowkov.conf to be present in the same directory as Makefile.
Redis data with corpus is persisted to data/dump.rpd.

• make docker-rebuild builds the app and runs it in a container
• make docker-update same as docker-rebuild but also checks for updates of golang and redis images
• make docker-stop stops already existing container
• make docker-start starts already existing container
• make docker-logs tails the output (runs in debug by default)
• make docker-clean removes meowkov containers and a custom image
• echo "some text" | make docker-corpus-add adds piped strings to the corpus
• echo "some text" | make docker-corpus-replace replaces corpus with piped data (destructive, remember to backup data/dump.rpd before executing this)

The bot is as good as its corpus.
Running it with empty one will not produce any meaningful results for a long time.
It is a good idea to bootstrap the corpus using old IRC logs, news articles, etc.

Text can be loaded into the corpus (which is backed by Redis) like this:

echo "line one\nline two with more text" | make docker-corpus-add

or if run in standalone mode:

echo "line one\nline two with more text" | ./meowkov -import=true -purge=false

One may also want to generate input on a different machine, for example from weechat logs:

find ~/.weechat/logs -name "*#foo*" -type f -exec sh -c "grep -vP '^.+\t(</-|-->|--|.*\*)\t' {} | cut -f3 -d$'\t'" \; > corpus.txt

Then transfer the file to the box with meowkov and perform import there:

cat corpus.txt | make docker-corpus-add

Changes are instantaneous: corpus import can be performed while bot is running, no restart is required.
The same text can be imported multiple times: Markov chains are kept in Redis Sets which provide automatic deduplication.

CC0 Public Domain Dedication