What changed

• Reworked the Streamlit dashboard into an insight-first local Yandex Music self-analytics product.
• Added a useful Focus controls sidebar with quick lenses, genre/liked/search/year/repeat/playlist-coverage filters.
• Added chart-first screens: Story, Taste Map, Mix Shift, Rediscovery, Playlists, Explorer, Actions, and Data Quality.
• Reduced visible tables by moving exact rows into audit expanders while keeping source-backed data available.
• Added token-help and real-compose smoke improvements from the local Yandex Music flow.

Why

The previous dashboard read like a BI table dump. This makes it more useful as a personal music analytics product: discovery backlog, genre shifts, release-era mix, artist affinity, repeat signals, playlist health, and cleanup candidates are visible without digging through tables.

Validation

• python -m compileall -q dashboard scripts tests
• python scripts/smoke_dashboard_content.py
• python scripts/smoke_dashboard.py
• python scripts/validate_yamusic_local.py
• PYTHONPATH=. pytest tests/test_dashboard_actions.py tests/test_dashboard_filters.py tests/test_yamusic_ingest.py -q
• make product-answers-smoke
• Browser QA on http://localhost:8501 for expanded filters, Story, Mix Shift, Playlists, and horizontal overflow.

Notes

• Uses only Yandex Music metadata; no audio is downloaded or stored.
• Data Analytics/Product Design saved plugin contexts are not configured, so local DuckDB remains the source of truth for this implementation.