I need a small, self-contained Python backend that ingests local MP3 files and automatically extracts musical information. For every track the service must detect • BPM and time signature • Overall tempo as well as integrated loudness (LUFS or a comparable metric) • Phrase boundaries and sections of complete silence Once a file is processed the results have to be written to a single CSV file—one row per track, neatly formatted with column headers for each metric. No database or JSON output is required. A minimal REST or command-line interface that lets me submit a path to an MP3 and returns the processing status is enough; the core value is the signal-processing code itself. Please rely on well-known Python libraries such as librosa, Essentia, pydub or similar so the solution remains transparent and easy to maintain. All third-party dependencies should be listed in a requirements.txt. Deliverables 1. Clean, well-commented source code (Python 3.x) 2. requirements.txt with exact versions 3. Example CSV produced from at least two sample MP3s 4. Short README explaining how to install, run, and extend the service I will consider the project complete when I can install the environment with a single pip command, feed an MP3 into the script or endpoint, and see a correctly populated CSV that matches the reported BPM, tempo, loudness, phrase start/end times, and silent regions.