Remove Spotify examples using deprecated tools (#199)
Refactor toolkit implementation examples following the removal of two Spotify tools in [PR#196](https://github.com/ArcadeAI/arcade-ai/pull/196), due to Spotify API deprecation announcement.
This commit is contained in:
Renato Byrro
GitHub
parent
6035cde920
commit
830f6f9f90
2 changed files with 55 additions and 125 deletions
|
|
@ -3,15 +3,15 @@
|
|||
For this example, we are using the prebuilt Spotify toolkit to start playing similar songs to the currently playing song.
|
||||
|
||||
Steps:
|
||||
1. Get the currently playing song
|
||||
2. Get audio features for the currently playing song
|
||||
3. Get song recommendations that are similar to the currently playing song
|
||||
4. Start playing the recommended songs
|
||||
5. Inform the user which recommended song is now playing
|
||||
1. Search for the song
|
||||
2. Start playing the song
|
||||
3. Get info about the currently playing song
|
||||
4. Inform the user which song is now playing
|
||||
"""
|
||||
|
||||
from typing import Any, Optional
|
||||
|
||||
from arcade_spotify.tools.models import SearchType
|
||||
from arcadepy import Arcade # pip install arcade-py
|
||||
|
||||
|
||||
|
|
@ -40,110 +40,55 @@ def call_tool(client: Arcade, tool_name: str, user_id: str, inputs: Optional[dic
|
|||
inputs=inputs,
|
||||
user_id=user_id,
|
||||
)
|
||||
|
||||
if response.output.error:
|
||||
print(response.output.error)
|
||||
|
||||
return response.output.value
|
||||
|
||||
|
||||
def recommend_similar_songs(
|
||||
client: Arcade, provider_to_scopes: dict, tools: list[str], user_id: str, user_country_code: str
|
||||
def search_and_play_song(
|
||||
client: Arcade,
|
||||
provider_to_scopes: dict,
|
||||
tools: list[str],
|
||||
user_id: str,
|
||||
song_name: str,
|
||||
artist_name: str,
|
||||
) -> None:
|
||||
"""Execute the sequence of tools to get recommendations and start playback."""
|
||||
get_permissions(client, provider_to_scopes, user_id)
|
||||
|
||||
recommendation_params = {"seed_genres": []}
|
||||
|
||||
(
|
||||
search_tool,
|
||||
start_playback_tool,
|
||||
get_currently_playing_tool,
|
||||
get_tracks_audio_features_tool,
|
||||
get_recommendations_tool,
|
||||
start_tracks_playback_tool,
|
||||
) = tools
|
||||
|
||||
# Step 1: Get the currently playing song
|
||||
while True:
|
||||
response = call_tool(client, get_currently_playing_tool, user_id)
|
||||
|
||||
if response["is_playing"]:
|
||||
break
|
||||
|
||||
print("Nothing is playing right now. Press Enter once you start playing a song...")
|
||||
input()
|
||||
|
||||
# Step 1.5: Use the previous tool output to construct the inputs for the following tool calls
|
||||
current_track_name = response["track_name"]
|
||||
current_track_artists = response["track_artists"]
|
||||
current_track_id = response["track_id"]
|
||||
current_track_spotify_url = response["track_spotify_url"]
|
||||
current_track_artists_ids = response["track_artists_ids"]
|
||||
print(
|
||||
f"\nYou are currently listening to '{current_track_name}' by {', '.join(current_track_artists)}"
|
||||
)
|
||||
recommendation_params.update({
|
||||
"seed_tracks": [current_track_id],
|
||||
"seed_artists": current_track_artists_ids,
|
||||
})
|
||||
|
||||
# Step 2:Get audio features for the currently playing song
|
||||
# Step 1: search for the song
|
||||
response = call_tool(
|
||||
client,
|
||||
get_tracks_audio_features_tool,
|
||||
user_id,
|
||||
inputs={"track_ids": [current_track_id]},
|
||||
client=client,
|
||||
tool_name=search_tool,
|
||||
user_id=user_id,
|
||||
inputs={
|
||||
"q": f"{song_name} {artist_name}",
|
||||
"types": [SearchType.TRACK],
|
||||
},
|
||||
)
|
||||
|
||||
# Step 2.5: Use the previous tool output to construct the inputs for the following tool calls
|
||||
audio_features = response["audio_features"][0]
|
||||
recommendation_params.update({
|
||||
"target_acousticness": float(audio_features["acousticness"]),
|
||||
"target_danceability": float(audio_features["danceability"]),
|
||||
"target_energy": float(audio_features["energy"]),
|
||||
"target_instrumentalness": float(audio_features["instrumentalness"]),
|
||||
"target_key": int(audio_features["key"]),
|
||||
"target_liveness": float(audio_features["liveness"]),
|
||||
"target_loudness": float(audio_features["loudness"]),
|
||||
"target_mode": int(audio_features["mode"]),
|
||||
"target_speechiness": float(audio_features["speechiness"]),
|
||||
"target_tempo": float(audio_features["tempo"]),
|
||||
"target_time_signature": int(audio_features["time_signature"]),
|
||||
"target_valence": float(audio_features["valence"]),
|
||||
})
|
||||
|
||||
# Step 3: Get song recommendations that are similar to the currently playingsong
|
||||
print(
|
||||
f"Getting recommendations similar to '{current_track_name}' by {', '.join(current_track_artists)} - {current_track_spotify_url}"
|
||||
)
|
||||
response = call_tool(client, get_recommendations_tool, user_id, inputs=recommendation_params)
|
||||
|
||||
# Step 3.5: Use the previous tool output to construct the inputs for the following tool calls
|
||||
# Filter out remixes and the same song from the recommendations
|
||||
tracks = [
|
||||
track for track in response["tracks"] if not track["name"].startswith(current_track_name)
|
||||
]
|
||||
track_ids = [
|
||||
track["id"] for track in tracks if (user_country_code in track["available_markets"])
|
||||
]
|
||||
track_names = [track["name"] for track in tracks]
|
||||
tracks_artists = [[artist["name"] for artist in track["artists"]] for track in tracks]
|
||||
track_spotify_urls = [track["external_urls"]["spotify"] for track in tracks]
|
||||
|
||||
if not track_ids:
|
||||
print("I couldn't find any similar songs that are available in your country.")
|
||||
if not response["tracks"]["items"]:
|
||||
print("Sorry, I couldn't find that song on Spotify.")
|
||||
return
|
||||
|
||||
print("\nHere are some recommendations:")
|
||||
for track_name, track_artists, track_spotify_url in zip(
|
||||
track_names, tracks_artists, track_spotify_urls
|
||||
):
|
||||
print(f"\t{track_name} by {', '.join(track_artists)} - {track_spotify_url}")
|
||||
|
||||
# Step 4: Start playing the recommended songs
|
||||
# Step 2: Start playing the song
|
||||
track_id = response["tracks"]["items"][0]["id"]
|
||||
response = call_tool(
|
||||
client,
|
||||
start_tracks_playback_tool,
|
||||
start_playback_tool,
|
||||
user_id,
|
||||
inputs={"track_ids": track_ids},
|
||||
inputs={"track_ids": [track_id]},
|
||||
)
|
||||
|
||||
# Step 5: Inform the user which recommended song is now playing
|
||||
# Step 3: get currently playing song
|
||||
response = call_tool(client, get_currently_playing_tool, user_id)
|
||||
print(
|
||||
f"\nNow playing: {response['track_name']} by {', '.join(response['track_artists'])} - {response['track_spotify_url']}"
|
||||
|
|
@ -158,21 +103,17 @@ if __name__ == "__main__":
|
|||
"spotify": [
|
||||
"user-read-currently-playing",
|
||||
"user-read-playback-state",
|
||||
"user-modify-playback-state",
|
||||
],
|
||||
}
|
||||
|
||||
tools = [
|
||||
"Spotify.Search", # Search for a song
|
||||
"Spotify.StartTracksPlaybackById", # Start playing the song
|
||||
"Spotify.GetCurrentlyPlaying", # Get info about the current song
|
||||
"Spotify.GetTracksAudioFeatures", # Get audio features for the current song
|
||||
"Spotify.GetRecommendations", # Get recommendations similar to the current song
|
||||
"Spotify.StartTracksPlaybackById", # Start playing the recommended songs
|
||||
]
|
||||
|
||||
user_id = "you@example.com"
|
||||
user_country_code = "US"
|
||||
song_name = input("Enter the song name: ")
|
||||
artist_name = input("Enter the artist name: ")
|
||||
|
||||
while True:
|
||||
recommend_similar_songs(client, provider_to_scopes, tools, user_id, user_country_code)
|
||||
print("\nPress Enter to get more recommendations...")
|
||||
input()
|
||||
search_and_play_song(client, provider_to_scopes, tools, user_id, song_name, artist_name)
|
||||
|
|
|
|||
|
|
@ -1,14 +1,14 @@
|
|||
"""
|
||||
Example script demonstrating how to call multiple tools (sequentially) using an LLM with authentication.
|
||||
|
||||
For this example, we are using the prebuilt Spotify toolkit to start playing similar songs to the currently playing song.
|
||||
For this example, we are using the prebuilt Spotify toolkit to search for a song, start playing it, and
|
||||
get info about the currently playing song.
|
||||
|
||||
Steps:
|
||||
1. Get the currently playing song
|
||||
2. Get audio features for the currently playing song
|
||||
3. Get song recommendations that are similar to the currently playing song
|
||||
4. Start playing the recommended songs
|
||||
5. Inform the user which recommended song is now playing
|
||||
1. Search for the song
|
||||
2. Start playing the song
|
||||
3. Get info about the currently playing song
|
||||
4. Inform the user which song is now playing
|
||||
"""
|
||||
|
||||
import os
|
||||
|
|
@ -31,28 +31,19 @@ def call_tool(client: OpenAI, user_id: str, tool: str, message: dict, history: l
|
|||
return response
|
||||
|
||||
|
||||
def call_tools_with_llm(client: OpenAI, user_id: str, user_country_code: str) -> list[dict]:
|
||||
"""Use an LLM to execute the sequence of tools to get recommendations and start playback."""
|
||||
def call_tools_with_llm(
|
||||
client: OpenAI, user_id: str, song_name: str, artist_name: str
|
||||
) -> list[dict]:
|
||||
"""Use an LLM to execute the sequence of tools to search for a song and start playback."""
|
||||
tools = [
|
||||
"Spotify.GetCurrentlyPlaying",
|
||||
"Spotify.GetTracksAudioFeatures",
|
||||
"Spotify.GetRecommendations",
|
||||
"Spotify.Search",
|
||||
"Spotify.StartTracksPlaybackById",
|
||||
"Spotify.GetCurrentlyPlaying",
|
||||
]
|
||||
|
||||
messages = [
|
||||
{"role": "user", "content": "Get the currently playing song."},
|
||||
{"role": "user", "content": "Retrieve its audio features."},
|
||||
{
|
||||
"role": "user",
|
||||
"content": "Get song recommendations similar to it. "
|
||||
"Do not include the currently playing song in the recommendations "
|
||||
"or any remixed versions of the song. "
|
||||
"Also only include tracks that are available in the user's country. "
|
||||
f"The current user resides in {user_country_code}.",
|
||||
},
|
||||
{"role": "user", "content": "Start playing the recommended songs. Just one tool call."},
|
||||
{"role": "user", "content": f"Search for '{song_name}' by {artist_name}' on Spotify."},
|
||||
{"role": "user", "content": "Start playing the song. Just one tool call."},
|
||||
{"role": "user", "content": "Get the currently playing song."},
|
||||
]
|
||||
|
||||
|
|
@ -82,10 +73,8 @@ if __name__ == "__main__":
|
|||
)
|
||||
|
||||
user_id = "you@example.com"
|
||||
user_country_code = "US"
|
||||
song_name = input("Enter the song name: ")
|
||||
artist_name = input("Enter the artist name: ")
|
||||
|
||||
while True:
|
||||
history = call_tools_with_llm(openai_client, user_id, user_country_code)
|
||||
|
||||
print("\nPress Enter to get more recommendations...")
|
||||
input()
|
||||
history = call_tools_with_llm(openai_client, user_id, song_name, artist_name)
|
||||
print("\n\n", history)
|
||||
|
|
|
|||
Loading…
Reference in a new issue