arcade-mcp/toolkits/spotify/evals/eval_search.py
Eric Gustin efee9589fa
More Spotify Tools (#140)
# PR Description
This PR adds three new spotify tools that are natural language friendly.

1. `search` - Search Spotify Catalog information
2. `play_artist_by_name` - Gets 5 songs by the specified artist and
plays them. Uses `search`, and `start_tracks_playback_by_id` under the
hood
3. `play_track_by_name` - Plays the specified song, optionally provide
the artist name who plays the song. Uses `search`, and
`start_tracks_playback_by_id` under the hood
2024-11-01 13:15:43 -07:00

52 lines
1.4 KiB
Python

from arcade_spotify.tools.models import SearchType
from arcade_spotify.tools.search import search
from arcade.sdk import ToolCatalog
from arcade.sdk.eval import (
EvalRubric,
EvalSuite,
tool_eval,
)
from arcade.sdk.eval.critic import BinaryCritic, SimilarityCritic
# Evaluation rubric
rubric = EvalRubric(
fail_threshold=0.9,
warn_threshold=0.95,
)
catalog = ToolCatalog()
catalog.add_tool(search, "Spotify")
@tool_eval()
def spotify_search_eval_suite() -> EvalSuite:
"""Create an evaluation suite for Spotify "player" tools."""
suite = EvalSuite(
name="Spotify Tools Evaluation",
system_message="You are an AI assistant that can manage Spotify using the provided tools.",
catalog=catalog,
rubric=rubric,
)
suite.add_case(
name="Search Spotify catalog",
user_message="search for 3 songs in the the album 'American IV: The Man Comes Around' by Johnny Cash",
expected_tool_calls=[
(
search,
{
"q": "album:American IV: The Man Comes Around artist:Johnny Cash",
"types": [SearchType.TRACK],
"limit": 3,
},
)
],
critics=[
SimilarityCritic(critic_field="q", weight=0.5),
BinaryCritic(critic_field="limit", weight=0.25),
BinaryCritic(critic_field="types", weight=0.25),
],
)
return suite