<?xml version="1.0" encoding="UTF-8"?>
<evaluations>
<description>
Comprehensive evaluation scenarios for Spotify Playlist MCP Server.
Tests core playlist management, track discovery, similarity engine,
and complex multi-step workflows.
</description>
<evaluation id="eval_001">
<name>Basic Playlist Creation and Population</name>
<description>Test creating a new playlist and adding tracks from search results</description>
<scenario>
You want to create a new workout playlist called "Morning Energy" and add some
high-energy tracks. First search for "Eye of the Tiger" by Survivor, then create
the playlist, and finally add the track to it.
</scenario>
<expected_steps>
1. Search for tracks using spotify_search_tracks with query "Eye of the Tiger Survivor"
2. Create a new playlist using spotify_create_playlist with name "Morning Energy"
3. Add the track to the playlist using spotify_add_tracks_to_playlist with the track URI
</expected_steps>
<success_criteria>
- Playlist is created successfully with the correct name
- Track is found in search results
- Track is added to the playlist
- All operations return success messages
</success_criteria>
</evaluation>
<evaluation id="eval_002">
<name>Similarity-Based Playlist Creation with Euclidean Strategy</name>
<description>Create a playlist of similar tracks using the similarity engine</description>
<scenario>
A user loves the song "Bohemian Rhapsody" by Queen and wants to create a playlist
of 25 similar tracks from the Spotify catalog. Use the default euclidean similarity
strategy to find tracks with similar audio characteristics and automatically create
a playlist called "Like Bohemian Rhapsody".
</scenario>
<expected_steps>
1. Search for "Bohemian Rhapsody" by Queen to get the track ID
2. Use spotify_find_similar_tracks with:
- track_id from search
- strategy: euclidean
- scope: catalog
- limit: 25
- action: create_playlist
- playlist_name: "Like Bohemian Rhapsody"
</expected_steps>
<success_criteria>
- Track is found successfully
- Similarity search returns 25 tracks
- Playlist is created with all 25 tracks
- Response includes playlist URL
</success_criteria>
</evaluation>
<evaluation id="eval_003">
<name>High-Energy Workout Playlist with Energy Match Strategy</name>
<description>Create a workout playlist using energy-focused similarity</description>
<scenario>
A user wants to create a high-energy workout playlist based on the song "Thunderstruck"
by AC/DC. They want 30 tracks that match the energy and danceability of this song,
using the energy_match strategy which prioritizes those specific audio features.
</scenario>
<expected_steps>
1. Search for "Thunderstruck AC/DC" to get track ID
2. Use spotify_find_similar_tracks with:
- track_id from search
- strategy: energy_match
- scope: catalog
- limit: 30
- action: create_playlist
- playlist_name: "High Energy Workout"
</expected_steps>
<success_criteria>
- Track is found successfully
- 30 similar tracks are found using energy_match strategy
- Playlist is created automatically
- Tracks should have similar energy and danceability values
</success_criteria>
</evaluation>
<evaluation id="eval_004">
<name>Genre Filtering Within Existing Playlist</name>
<description>Filter a playlist to find tracks matching a specific genre</description>
<scenario>
A user has a "Discover Weekly" playlist and wants to find all tracks in it that
match the genre of "indie rock". They want to create a new playlist with just
those genre-matching tracks.
</scenario>
<expected_steps>
1. Get user's playlists using spotify_get_user_playlists to find Discover Weekly
2. Use spotify_find_similar_tracks with:
- track_id of a known indie rock track (or search for one first)
- strategy: genre_match
- scope: playlist
- scope_id: Discover Weekly playlist ID
- limit: 20
- action: create_playlist
- playlist_name: "Indie Rock from Discover Weekly"
</expected_steps>
<success_criteria>
- Discover Weekly playlist is found
- Genre matching finds tracks successfully
- New playlist is created with genre-filtered tracks
- Tracks share similar genres
</success_criteria>
</evaluation>
<evaluation id="eval_005">
<name>Mood-Based Playlist with Custom Weights</name>
<description>Create a playlist using weighted similarity emphasizing specific features</description>
<scenario>
A user wants to create a calm, acoustic playlist for studying. They want tracks
similar to "Skinny Love" by Bon Iver, but they want to heavily prioritize
valence (mood) and acousticness, with less emphasis on other features.
Use custom weights: valence=5.0, acousticness=5.0, energy=0.5, others=1.0.
</scenario>
<expected_steps>
1. Search for "Skinny Love Bon Iver" to get track ID
2. Use spotify_find_similar_tracks with:
- track_id from search
- strategy: weighted
- weights: {valence: 5.0, acousticness: 5.0, energy: 0.5}
- scope: catalog
- limit: 25
- action: create_playlist
- playlist_name: "Calm Acoustic Study Mix"
</expected_steps>
<success_criteria>
- Track is found successfully
- Weighted similarity calculation is applied correctly
- 25 tracks are found and added to new playlist
- Tracks should have similar valence and acousticness values
</success_criteria>
</evaluation>
<evaluation id="eval_006">
<name>Audio Feature Analysis and Comparison</name>
<description>Analyze and compare audio features of multiple tracks</description>
<scenario>
A user wants to understand the audio characteristics of three different tracks:
"Happy" by Pharrell Williams, "Someone Like You" by Adele, and "Smells Like Teen Spirit"
by Nirvana. Get the audio features for all three tracks to compare their energy,
valence, and tempo.
</scenario>
<expected_steps>
1. Search for each track to get their track IDs
2. Use spotify_get_audio_features with the list of all three track IDs
3. Compare the returned audio features
</expected_steps>
<success_criteria>
- All three tracks are found successfully
- Audio features are retrieved for all tracks
- Features include energy, valence, tempo, danceability, etc.
- Response format allows for comparison
</success_criteria>
</evaluation>
<evaluation id="eval_007">
<name>Recommendations with Audio Feature Tuning</name>
<description>Get recommendations with specific audio feature constraints</description>
<scenario>
A user wants to discover new high-energy, happy dance music. They want recommendations
that have energy between 0.7-1.0, danceability between 0.6-1.0, and valence (happiness)
above 0.7. Use electronic and pop as seed genres, and limit to 20 tracks.
</scenario>
<expected_steps>
1. Use spotify_get_recommendations with:
- seed_genres: ["electronic", "pop"]
- min_energy: 0.7
- min_danceability: 0.6
- min_valence: 0.7
- limit: 20
</expected_steps>
<success_criteria>
- 20 recommendations are returned
- All tracks match the specified audio feature constraints
- Tracks are from electronic/pop genres
- Response includes track details
</success_criteria>
</evaluation>
<evaluation id="eval_008">
<name>Finding Similar Tracks Within User's Saved Library</name>
<description>Search for similar tracks within the user's own saved music</description>
<scenario>
A user just discovered they love "Mr. Brightside" by The Killers and wants to find
similar tracks from their own saved music library. They want to add these similar
tracks to an existing playlist called "My Favorites".
</scenario>
<expected_steps>
1. Search for "Mr. Brightside The Killers" to get track ID
2. Get user's playlists to find "My Favorites" playlist ID
3. Use spotify_find_similar_tracks with:
- track_id from search
- strategy: euclidean
- scope: saved_tracks
- limit: 15
- action: add_to_playlist
- target_playlist_id: My Favorites playlist ID
</expected_steps>
<success_criteria>
- Track is found successfully
- Similar tracks are found in user's saved library
- Tracks are added to existing "My Favorites" playlist
- Response confirms number of tracks added
</success_criteria>
</evaluation>
<evaluation id="eval_009">
<name>Rhythm-Based Running Playlist</name>
<description>Create a running playlist with consistent tempo using rhythm_match strategy</description>
<scenario>
A user wants to create a running playlist with songs around 180 BPM (matching their
running cadence). They want to use "Don't Stop Me Now" by Queen as the reference
track and find 40 songs with similar tempo using the rhythm_match strategy.
</scenario>
<expected_steps>
1. Search for "Don't Stop Me Now Queen" to get track ID
2. Optional: Get audio features to verify tempo is around 180 BPM
3. Use spotify_find_similar_tracks with:
- track_id from search
- strategy: rhythm_match
- scope: catalog
- limit: 40
- action: create_playlist
- playlist_name: "180 BPM Running Mix"
</expected_steps>
<success_criteria>
- Track is found successfully
- Rhythm matching finds 40 tracks with similar tempo
- Playlist is created automatically
- Tracks should have tempo values close to the reference track
</success_criteria>
</evaluation>
<evaluation id="eval_010">
<name>Complex Multi-Step Playlist Curation Workflow</name>
<description>Complete workflow: browse, analyze, filter, and create curated playlist</description>
<scenario>
A user wants to create the ultimate chill evening playlist. The workflow:
1. Browse their existing playlists to find a "Chill Vibes" playlist
2. Get tracks from that playlist to see what's in it
3. Pick a favorite track from the list
4. Get audio features for that track to understand its characteristics
5. Find similar tracks using mood_match strategy (focusing on valence and acousticness)
6. Create a new playlist called "Ultimate Chill Evening" with 30 similar tracks
7. Verify the playlist was created by listing user playlists again
</scenario>
<expected_steps>
1. spotify_get_user_playlists to browse available playlists
2. spotify_get_playlist_tracks to see tracks in "Chill Vibes"
3. spotify_get_audio_features for a selected track to analyze it
4. spotify_find_similar_tracks with:
- track_id from selected track
- strategy: mood_match
- scope: catalog
- limit: 30
- action: create_playlist
- playlist_name: "Ultimate Chill Evening"
5. spotify_get_user_playlists again to verify new playlist appears
</expected_steps>
<success_criteria>
- User can browse and find existing playlists
- Track list is retrieved successfully
- Audio features provide insight into track characteristics
- Mood-based similarity finds appropriate tracks
- New playlist is created with 30 tracks
- New playlist appears in user's playlist list
- All operations complete successfully in sequence
</success_criteria>
</evaluation>
<evaluation id="eval_011">
<name>Artist-Style Based Discovery</name>
<description>Create a playlist based on an artist's overall style</description>
<scenario>
A user loves Radiohead's style and wants to discover 25 tracks that sound similar
to their overall artistic style (not just one song). Use artist-based similarity
to analyze Radiohead's catalog and find similar tracks.
</scenario>
<expected_steps>
1. Search for "Radiohead" to find artist (may need track search then extract artist ID)
2. Use spotify_find_similar_tracks with:
- artist_id for Radiohead
- strategy: cosine
- scope: catalog
- limit: 25
- action: create_playlist
- playlist_name: "Artists Like Radiohead"
</expected_steps>
<success_criteria>
- Radiohead artist is identified
- Similarity engine analyzes artist's style
- 25 similar tracks are found
- Playlist is created successfully
</success_criteria>
</evaluation>
<evaluation id="eval_012">
<name>Pagination and Large Result Handling</name>
<description>Test pagination when working with large playlists or search results</description>
<scenario>
A user has a very large playlist with 200+ tracks and wants to browse through all
of them in batches. They want to see tracks 50-100 (using offset and limit).
Then they want to search for "love songs" which returns hundreds of results,
and page through the first 3 pages (20 tracks per page).
</scenario>
<expected_steps>
1. spotify_get_user_playlists to find the large playlist
2. spotify_get_playlist_tracks with offset=50, limit=50
3. spotify_search_tracks with query="love songs", limit=20, offset=0 (page 1)
4. spotify_search_tracks with query="love songs", limit=20, offset=20 (page 2)
5. spotify_search_tracks with query="love songs", limit=20, offset=40 (page 3)
</expected_steps>
<success_criteria>
- Pagination works correctly for playlist tracks
- Offset and limit parameters are respected
- Search pagination returns different results for each page
- has_more flag indicates if more results available
- No duplicate tracks across pages
</success_criteria>
</evaluation>
</evaluations>
