Telegram Bot BasilAI

Install

• prod environment

  make install-prod

• dev environment

  make install-dev

Generate dataset

make dataset RAW_DATA=<path to telegram dialogs dump json> TARGET=<user_id> DATASET=<path to csv file to save generated dataset>

Generate advanced dataset

make dataset_advanced \
  USER_ID=<user id> \
  RAW_DATASET=<path to raw json dialog file> \
  DIALOG_SECONDS_COOLDOWN=<minimum time distance in seconds between different dialogs> \
  DIALOG_MEMORY=<message number in memory to answer> \
  DAATSET_OUTPUT=<path to save generated dataset folder>

Fit tokenizer

make fit_tokenizer \
  TOKENIZER_DATASET=<path to the dataset> \
  TOKENIZER=<tokenizer class> \
  TOKENIZER_SIZE=<tokenizer vocab size> \
  TOKENIZER_OUTPUT=<path to save tokenizer>

Train model

make train TRAIN_CONFIG=<path to config file>

Run telegram bot locally

make bot TOKEN=<telegram bot auth token>

Heroku hosting

Getting started with heroku

• authenticate on heroku

  heroku login

• add git heroku remote

  heroku git:remote -a basilai-bot

Pushing new version to heroku

• push local master branch to heroku main branch

  git push heroku master:main

  you can push any local branch - just change master to branch name

• heroku restarts application with new version automatically

Scaling heroku application

• check heroku running apps

  heroku ps

• scale heroku apps

  heroku ps:scale worker=1

• stop heroku apps

  heroku ps:stop worker.1

Get logs

heroku logs -t