Files
derp/plugins/_musicbrainz.py
user da9ed51c74 feat: auto-discover similar tracks during autoplay via Last.fm/MusicBrainz
Every Nth autoplay pick (configurable via discover_ratio), query Last.fm
for similar tracks. When Last.fm has no key or returns nothing, fall back
to MusicBrainz tag-based recording search (no API key needed). Discovered
tracks are resolved via yt-dlp and deduplicated within the session. If
discovery fails, the kept-deck shuffle continues as before.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-23 21:19:41 +01:00

119 lines
3.6 KiB
Python

"""MusicBrainz API helper for music discovery fallback.
Private module (underscore prefix) -- plugin loader skips it.
All functions are blocking; callers should run them in an executor.
"""
from __future__ import annotations
import json
import logging
import time
from urllib.request import Request
log = logging.getLogger(__name__)
_BASE = "https://musicbrainz.org/ws/2"
_UA = "derp-bot/2.0.0 (https://git.mymx.me/username/derp)"
# Rate limit: MusicBrainz requires max 1 request/second.
# We use 1.1s between calls to stay well within limits.
_RATE_INTERVAL = 1.1
_last_request: float = 0.0
def _mb_request(path: str, params: dict | None = None) -> dict:
"""Rate-limited GET to MusicBrainz API. Blocking."""
global _last_request
from derp.http import urlopen
elapsed = time.monotonic() - _last_request
if elapsed < _RATE_INTERVAL:
time.sleep(_RATE_INTERVAL - elapsed)
qs = "&".join(f"{k}={v}" for k, v in (params or {}).items())
url = f"{_BASE}/{path}?fmt=json&{qs}" if qs else f"{_BASE}/{path}?fmt=json"
req = Request(url, headers={"User-Agent": _UA})
try:
resp = urlopen(req, timeout=10, proxy=False)
_last_request = time.monotonic()
return json.loads(resp.read().decode())
except Exception:
_last_request = time.monotonic()
log.warning("musicbrainz: request failed: %s", path, exc_info=True)
return {}
def mb_search_artist(name: str) -> str | None:
"""Search for an artist by name, return MBID or None."""
from urllib.parse import quote
data = _mb_request("artist", {"query": quote(name), "limit": "1"})
artists = data.get("artists", [])
if not artists:
return None
# Require a reasonable score to avoid false matches
score = artists[0].get("score", 0)
if score < 50:
return None
return artists[0].get("id")
def mb_artist_tags(mbid: str) -> list[str]:
"""Fetch top 5 tags for an artist by MBID."""
data = _mb_request(f"artist/{mbid}", {"inc": "tags"})
tags = data.get("tags", [])
if not tags:
return []
# Sort by count descending, take top 5
sorted_tags = sorted(tags, key=lambda t: t.get("count", 0), reverse=True)
return [t["name"] for t in sorted_tags[:5] if t.get("name")]
def mb_find_similar_recordings(artist: str, tags: list[str],
limit: int = 10) -> list[dict]:
"""Find recordings by other artists sharing top tags.
Searches MusicBrainz for recordings tagged with the top 2 tags,
excluding the original artist. Returns [{"artist": str, "title": str}].
"""
from urllib.parse import quote
if not tags:
return []
# Use top 2 tags for the query
tag_query = " AND ".join(f'tag:"{t}"' for t in tags[:2])
query = f'({tag_query}) AND NOT artist:"{artist}"'
data = _mb_request("recording", {
"query": quote(query),
"limit": str(limit),
})
recordings = data.get("recordings", [])
if not recordings:
return []
seen = set()
results = []
for rec in recordings:
title = rec.get("title", "")
credits = rec.get("artist-credit", [])
if not credits or not title:
continue
rec_artist = credits[0].get("name", "") if credits else ""
if not rec_artist:
continue
# Skip the original artist (case-insensitive)
if rec_artist.lower() == artist.lower():
continue
# Deduplicate by artist+title
key = f"{rec_artist.lower()}:{title.lower()}"
if key in seen:
continue
seen.add(key)
results.append({"artist": rec_artist, "title": title})
return results