How gregorytravis/rhythmr works
Project Type and Purpose
• The project is an interactive audio loop creation system. It processes randomly acquired audio loops and lets users vote (swipe left/right) to iteratively refine a model that helps generate a complete song.
• It’s built in Haskell (note the abundance of .hs files in src) and leverages a combination of audio analysis, user-interaction, model-building, and musical score insertion.
Key Components from the File Structure
a. Core Haskell Modules (located in the src directory):
• Modules like Main.hs, Analysis.hs, Loop.hs, Looper.hs, Project.hs, and Gui.hs indicate the engine for managing loops, processing audio, interfacing with the user, and handling business logic.
• Other modules (e.g., AesonHelpers.hs, State.hs, SaveLoad.hs) suggest data modeling, serialization, state management, and perhaps interactions with external data sources or config files.
b. Build and Packaging:
• The build directory contains scripts, templates, and helper utilities (such as generate-distribution, generate-icns, shell scripts, and even a packaged macOS app template under Rhythmr.app-template) for generating a distributable version of the application.
c. Audio Input/Resources:
• The wavs directory and scripts like get-mp3s.py and musicinformationretrieval.py imply components for sourcing or preprocessing audio.
d. Documentation and Presentation:
• The doc and presentation directories indicate extensive documentation, installation instructions, and feature descriptions essential for onboarding and maintenance.
• From “User Interface” to “Core Orchestrator” (user actions) and vice versa for presenting analysis results.
• Between Audio Processing and Model Training, showing iterative improvements.
4. A separate section for build/deployment workflows, drawing connections from the core system to packaging scripts.
5. Component with the relevant technologies (e.g., “Written in Haskell”, “Packaging scripts in Bash/Python”).