About the project

A focused tool for
exploring music research.

Music Research Assistant is a Python web application by Avery Wang, associated with Melodies 4 Medicine (M4M), and hosted on Hugging Face Spaces.

What the application does.

The application combines a searchable collection of PDF passages with an external language model. It retrieves material related to a user’s question, then uses that material to support a concise answer with source references.

Its subject areas include music, health, neurodevelopment, cognition, and brain structure. The collection determines what it can retrieve; the application is not a comprehensive, live search of all published research.

A web interface

Gradio provides the question field, progress feedback, answer display, retrieval settings, and source presentation.

A local search index

Prepared passage text and numerical embeddings support semantic search over the collection.

External answer generation

OpenRouter receives the question and selected passages for answer generation.

The research collection

2,000+ publications.
A foundation for exploration.

2,000+ publicationsProject research collection
Multiple research areasMusic, health & the brain
Passage-based searchFind relevant source material

The project brings together more than 2,000 research publications spanning music, health, neurodevelopment, cognition, and brain structure.

Documents are prepared as searchable passages, allowing questions to be matched with related material. Source references give readers a route back to the publications for closer review.

Publication counts and passage counts describe different things: a single document may contribute many passages to the search index. Retrieval depends on the indexed material available to the application.

What this resource is for.

Use the assistant for educational exploration, focused questions, and identifying passages for further reading. Check references before citing them. The presence of a source excerpt does not by itself establish that an answer represents the paper accurately.

The site does not claim clinical validation or a measured accuracy rate. Individual diagnosis and treatment decisions should be discussed with a qualified healthcare professional.

Read the public project documentation for deployment information.

Read the project motivation →