Music Information Retrieval API

Documentation complète

Overview

This project is a FastAPI-based web service that combines natural language processing with music database querying to provide detailed answers to music-related questions.

Features

• Natural language question processing
• Music information extraction from questions
• Discogs database querying
• AI-powered response generation
• Confidence scoring for answers

Prerequisites

• Python 3.8+
• FastAPI
• Anthropic API key
• Discogs API key and secret

Installation

1. Clone the repository:

   git clone https://github.com/yourusername/music-info-retrieval-api.git
   cd music-info-retrieval-api
   

2. Install dependencies:

   pip install -r requirements.txt
   

3. Set up environment variables: Create a .env file in the project root and add the following:

   ANTHROPIC_API_KEY=your_anthropic_api_key
   DISCOGS_API_KEY=your_discogs_api_key
   DISCOGS_API_SECRET=your_discogs_api_secret
   

Usage

1. Start the FastAPI server:

   uvicorn backend.main:app --reload
   

2. The API will be available at http://localhost:8000

3. Use the /ask endpoint to submit questions:

   POST /ask
   {
     "text": "Who released the album 'Thriller' in 1982?"
   }
   

API Endpoints

POST /ask

Submit a music-related question.

Request body:

json { "answer": "string", "confidence": "float", "discogs_info": { "artist": "string", "title": "string", "year": "integer", "genre": "string" } }

Architecture

1. Question Analysis: The input question is processed by Claude AI to extract relevant music information.
2. Discogs Query: The extracted information is used to query the Discogs database.
3. Response Generation: Claude AI generates a final response based on the original question and Discogs data.
4. Confidence Scoring: A confidence score is assigned based on the quality of the Discogs results.

Error Handling

The API uses extensive logging to track the flow of execution and capture any errors. Errors are logged and appropriate HTTP exceptions are raised with descriptive messages.