This repo assumes you are running Ubuntu 20 LTS or NixOS, but it may work on other versions/distros if you are lucky.

Follow these steps to initialize a fresh environment capable of building and running any Innovation Upstream repo.

1. mkdir $HOME/workspace && cd $HOME/workspace (Optional, see step 4)
2. Clone this repo
3. Run cmd/install.sh
4. Run home-manager switch build the env

(Optional) If you cloned devrc into a different directory, you will need to set the DEVRC_DIR environment variable in ~/.devrc.

$ export DEVRC_DIR="$HOME/workspace-2/devrc"
$ source $DEVRC_DIR/.zshrc

If you need to add/modify your shell profile, you can do so by creating/modifying ~/.devrc. (home-manager will tell zsh to source this file if it exists)

Local speech-to-text using faster-whisper with i3 integration.

1. Run the setup script to create Python venv and install faster-whisper:

   ./scripts/setup-dictation.sh

2. Add keybinding to /etc/nixos/i3config.nix:

   bindsym $mod+s exec --no-startup-id ~/workspace/devrc/scripts/dictate
   

3. Rebuild:

   sudo nixos-rebuild switch
   home-manager switch

With Alt+s:

1. Press once: Start recording (notification shows "Recording...")
2. Auto-stop: Recording stops automatically after 1s of silence
3. Optional: Press Alt+s again to force immediate stop

• First use: Daemon auto-starts (~5-10s to load model on GPU)
• Transcription: ~0.3s on GPU with large-v3-turbo

The script includes optimizations for technical dictation:

• Developer vocabulary prompt: Guides Whisper toward technical terms (NixOS, TypeScript, React, etc.)
• Word replacements: Auto-corrects common misheard terms (e.g., "knicks os" → "NixOS")
• Reduced beam_size: Faster transcription with minimal accuracy loss

Edit WORD_REPLACEMENTS in scripts/dictate.py to add your own corrections.