This is a work-in-progress docker-compose setup for running Scripty, a speech-to-text transcription bot for Discord.
The setup here will currently only work with CUDA (Nvidia) GPUs.
First, ensure that Git and Docker desktop are installed.
# Clone this repository, making sure to include submodules
git clone --recurse-submodules -j8 https://github.com/wilt00/scripty-stack.git
cd scripty-stackCreate a file called scripty.config.toml with your configuration settings; see scripty.config.example.toml for more on which values need to be changed.
# This temporary database is required to build scripty; you can delete the container once the bot is running
docker run --name db-temp -p "5432:5432" -e POSTGRES_USER=scripty -e POSTGRES_PASSWORD=scripty -d postgres
docker-compose up -dIf the STT service and the bot service don't start up in exactly the right order, the bot service may fail to start; in this case, restarting the bot container should fix it.
To enable slash commands, run ~register_cmds in a Discord channel where the bot is listening. This will not work if certain languages are enabled, as translations for these commands are missing.