Plexamp

[michaelrubloff]

This repo is awesome! Is there any chance it might work with Plexamp too?

[kirarpit]

It can definitely get music library stats from tautulli relatively easily and use that to create cards.

[kirarpit]

@claude can you take a look?

[claude]

Claude finished @kirarpit's task in 1m 47s —— View job

---

Plexamp Support Analysis Complete ✓

---

Current State

How plexwrap works:

• Fetches viewing data from Tautulli (which connects to Plex)
• Analyzes TV shows and movies viewing history
• Explicitly filters out music tracks (media_type == "track") in two places:
  • analyzer.py:168 - during history analysis
  • analyzer.py:966 - when getting top content
• Creates personalized year-in-review cards with AI-generated content

What is Plexamp:
Plexamp is Plex's dedicated music player. Since Tautulli tracks all Plex activity (including music played through Plexamp), the data is already available!

---

Good News! 🎵

The data is already flowing through Tautulli - we just need to stop filtering it out and add music-specific analysis.

Implementation Options

Option A: Unified Wrap (Music + Video)

• Single wrap showing both watching and listening stats
• Mixed cards for holistic media consumption view
• More complex to display coherently

Option B: Separate Music Wraps (Recommended ⭐)

• Separate music-focused wraps for Plexamp users
• Clean separation of concerns
• Easier to iterate and test
• Can reuse most existing infrastructure

Option C: Configurable Mode

• Config toggle to generate video-only, music-only, or combined wraps
• Maximum flexibility
• More code complexity

---

Required Changes

1. Remove music filtering (analyzer.py)

   • Remove media_type == "track" filters (lines 168, 966)
   • Add music-specific fields (artist, album, play count)

2. Add music-specific analysis (analyzer.py)

   • Top artists (similar to actors/directors)
   • Top albums (similar to top content)
   • Top songs with play counts
   • Music genres (already partially handled)
   • Listening patterns (time of day, mood analysis)

3. Create music-specific card types (frontend)

   • Artist spotlight cards
   • Album collection cards
   • Music genre breakdown
   • Listening personality cards
   • "On repeat" cards for most-played songs

4. Update data models (models.py)

   • Add music-specific stats structures
   • Artist/Album models (similar to ActorStat)
   • Track model with play count

5. Configuration options (config.yaml)

   • Add include_music: true/false toggle
   • Add music_only_mode: true/false for pure music wraps

---

My Recommendation

Option B (separate music wraps) is the cleanest path forward:

1. Keep existing video filtering as default
2. Add new --music-mode flag to pregenerate.py
3. Create music-specific analyzer methods
4. Reuse LLM card generation with music-focused prompts
5. Add music-themed card components

This allows incremental development without breaking existing functionality.

---

Next Steps

I'm ready to implement this if you'd like! Would you prefer:

1. Option B - Separate music-focused wrap generator (recommended)
2. Option C - Configurable music/video/both modes
3. Just the analysis - Let you implement it yourself

Let me know and I can start coding! 🚀

---

[kirarpit]

let's do option A -- the unified approach. pull music data. give it to the text llm client and let it include cards on music when there is enough data to do so @claude

[claude]

Claude finished @kirarpit's task in 4m 14s —— View job • claude/issue-3-20260107-0359 • Create PR ➔

---

Unified Music + Video Wrap Implementation Complete! ✓

Successfully implemented Option A: Unified approach that pulls both music and video data, letting the LLM decide when to include music cards based on available data.

Todo List

• Read CLAUDE.md for repo-specific guidelines
• Analyze current code structure (analyzer.py, models.py)
• Check LLM client to understand prompt generation
• Remove music filtering from analyzer.py
• Add music-specific data extraction (artists, albums, tracks)
• Update data models to support music statistics
• Update LLM context to include music data
• Check for syntax errors
• Commit and push changes
• Create PR

Implementation Summary

Changes made:

1. analyzer.py - Modified analyze_history():

   • Added music-specific counters (tracks_listened, artists, albums, tracks, music_genres)
   • Changed music handling from filtering to tracking
   • Music tracks now populate separate statistics
   • Added _extract_genres() helper method to reduce code duplication
   • Return dict now includes: tracks_listened, music_listen_time, top_artists, top_albums, top_tracks, music_genres

2. analyzer.py - Modified analyze_user_raw_data():

   • Added all music statistics to raw_data dict
   • LLM now receives complete music data alongside video data

3. clients/llm_client.py - Updated system prompt:

   • Added music-specific card ideas (top artists, albums, tracks, genres)
   • LLM instructed to only include music cards when there's significant data
   • Maintains backward compatibility for video-only users

How it works:

• Tautulli data includes both video and music (from Plexamp)
• Both are now analyzed and included in raw_data
• LLM receives unified data and decides what cards to generate
• If user has no/little music data → video-only cards
• If user has significant music data → mixed video + music cards
• Fully automatic, no configuration needed!

Testing recommendations:

1. Test with a user who has only video data (should work as before)
2. Test with a user who has only music data (should generate music-focused wrap)
3. Test with a user who has both (should generate unified wrap with best content from both)

---