This commit is contained in:
Space-Banane
2026-04-02 19:47:53 +02:00
parent 9ed4e240c2
commit bdddf602be
30 changed files with 783 additions and 17 deletions

View File

@@ -0,0 +1,25 @@
# Music Discovery Pipeline
Use listening history and taste signals to discover new artists and tracks that match your profile without repeating obvious recommendations.
## Problem
Streaming recommendations often circle familiar content, limiting genuine discovery.
## Core capabilities
- Ingest listening history from Last.fm or Spotify.
- Build a taste profile from genres, eras, and track features.
- Recommend similar but unexplored artists with confidence score.
- Generate themed playlists for focus, workouts, or relaxed sessions.
## MVP scope
- Import history from one service.
- Weekly discovery list with skip/save feedback loop.
- Playlist export to chosen music platform.
## Success criteria
- Increased number of newly liked artists per month.
- Better diversity in listening patterns.
## Stretch ideas
- Event-aware recommendations based on nearby concerts.
- Cross-language discovery with translation-aware metadata.

View File

@@ -0,0 +1,25 @@
# Podcast Chapter Summarizer
Transcribe podcast episodes, split them into chapters, and produce skimmable notes with key ideas and timestamps.
## Problem
Long episodes are hard to review, so valuable insights are often forgotten or never revisited.
## Core capabilities
- Ingest audio from RSS feeds or local files.
- Generate transcript and chapter segmentation.
- Summarize each chapter with key points and action items.
- Export notes to markdown, knowledge bases, or task tools.
## MVP scope
- Process a selected podcast feed.
- Chapter-level summary template with timestamp links.
- Searchable archive of episode notes.
## Success criteria
- Faster content review and better retention.
- Easier sharing of insights from long-form audio.
## Stretch ideas
- Topic-based clipping for personalized highlight reels.
- Question-answer mode over past episode archives.

View File

@@ -0,0 +1,25 @@
# Personal Watchlist Ranker
Rank unwatched movies and shows based on mood, available time, genre preferences, and quality signals.
## Problem
Large watchlists create decision paralysis, especially when available time and mood vary day to day.
## Core capabilities
- Combine ratings, runtime, mood tags, and recency into a ranking score.
- Filter by constraints like time window and platform availability.
- Learn personal preferences from watch history and skips.
- Produce quick-pick recommendations for tonight, weekend, or deep-watch sessions.
## MVP scope
- Import watchlist from one source.
- Score and rank items with editable weighting.
- Weekly shortlist generation.
## Success criteria
- Less time spent choosing content.
- Higher satisfaction with watched items.
## Stretch ideas
- Group recommendation mode for multiple viewers.
- Surprise mode that balances novelty and confidence.