feat(uploads): match release groups with MusicBrainz and AcoustID

services/api/internal/uploads/release_audio.go

@@ -0,0 +1,76 @@
+package uploads
+
+import (
+	"context"
+	"errors"
+	"sort"
+
+	"github.com/google/uuid"
+	"pipu/api/internal/acoustid"
+)
+
+const maxAudioFiles = 50
+
+func chooseAudioMatch(id uuid.UUID, matches []acoustid.Match) AudioRecognition {
+	// Multiple AcoustIDs can point to the same recording. Compare distinct recordings only.
+	best := map[string]float64{}
+	for _, m := range matches {
+		if parsed, err := uuid.Parse(m.RecordingID); err == nil && parsed != uuid.Nil && m.Score > best[parsed.String()] && m.Score <= 1 {
+			best[parsed.String()] = m.Score
+		}
+	}
+	unique := []acoustid.Match{}
+	for id, score := range best {
+		unique = append(unique, acoustid.Match{RecordingID: id, Score: score})
+	}
+	sort.Slice(unique, func(i, j int) bool { return unique[i].Score > unique[j].Score })
+	r := AudioRecognition{FileID: id, State: "not_found"}
+	if len(unique) == 0 {
+		return r
+	}
+	r.Score = unique[0].Score
+	if unique[0].Score < .9 || len(unique) > 1 && unique[0].Score-unique[1].Score < .05 {
+		r.State = "ambiguous"
+		return r
+	}
+	r.State, r.RecordingID = "matched", unique[0].RecordingID
+	return r
+}
+
+// identifyAudio runs only after an explicit reviewer request. It never modifies file tags.
+func (s *Service) identifyAudio(ctx context.Context, files []File) ([]AudioRecognition, map[uuid.UUID]AudioRecognition, []string, error) {
+	results := []AudioRecognition{}
+	byFile := map[uuid.UUID]AudioRecognition{}
+	ids := []string{}
+	if s.identifier == nil || s.encodedFingerprint == nil {
+		return results, byFile, ids, errors.New("acoustid: not configured")
+	}
+	for _, f := range releaseFiles(files) {
+		if f.Status != StatusPendingReview && f.Status != StatusImportFailed {
+			continue
+		}
+		if len(results) >= maxAudioFiles {
+			return results, byFile, ids, errors.New("acoustid: batch too large")
+		}
+		r := AudioRecognition{FileID: f.ID, State: "unavailable"}
+		if f.DurationMs != nil && *f.DurationMs > 0 && f.StagingPath != "" {
+			fingerprint, err := s.encodedFingerprint(ctx, s.stagingAbs(f.StagingPath))
+			if err == nil {
+				matches, err := s.identifier.Lookup(ctx, fingerprint, (*f.DurationMs+500)/1000)
+				if err != nil {
+					return results, byFile, ids, err
+				}
+				r = chooseAudioMatch(f.ID, matches)
+			}
+		}
+		if err := ctx.Err(); err != nil {
+			return results, byFile, ids, err
+		}
+		results = append(results, r)
+		byFile[f.ID] = r
+		if r.State == "matched" {
+			ids = append(ids, r.RecordingID)
+		}
+	}
+	return results, byFile, ids, nil
+}