feat: enhance fuse_scores function with CLAP and FST results integration

app/services/score_fusion.py

@@ -49,22 +49,28 @@ def fuse_scores(
     features: AudioFeatures,
     vocals: Optional[VocalFeatures] = None,
     *,
+    clap_result: Optional[CLAPResult] = None,
+    fst_result: Optional[FSTResult] = None,
     external_music_ai: Optional[dict] = None,
     external_ses_analizi: Optional[dict] = None,
 ) -> FusionResult:
     """
     Fuse all analysis signals into a single detection result.
 
-    Priority hierarchy:
-    1. External production models (music_ai, ses_analizi) if available
-    2. Local feature + vocal analysis (always available)
+    3-Layer architecture:
+      Layer 1: Local feature + vocal analysis (always available)
+      Layer 2: CLAP embedding classifier (optional)
+      Layer 3: FST external API (optional)
+      Legacy:  music_ai / ses_analizi (optional)
 
-    When external models agree with local analysis, confidence is boosted.
+    When multiple layers agree, confidence is boosted.
     When they disagree, confidence is moderated.
 
     Args:
         features: Audio feature extraction results.
         vocals: Vocal analysis results (None if skipped).
+        clap_result: CLAP detector result (Layer 2).
+        fst_result: FST external result (Layer 3).
         external_music_ai: Response from MusicAI service.
         external_ses_analizi: Response from SesAnalizi service.
 
@@ -73,7 +79,7 @@ def fuse_scores(
     """
     indicators: List[str] = []
 
-    # ── Local feature scores ─────────────────────────────────────────
+    # ── Local feature scores (Layer 1) ───────────────────────────────
     local_score = _compute_local_score(features, vocals)
 
     vocal_ai_score = 0.0
@@ -81,20 +87,27 @@ def fuse_scores(
         vocal_ai_score = vocals.vocal_ai_score
     has_vocals = vocals is not None and vocals.has_vocals
 
-    # ── External service scores ──────────────────────────────────────
+    # ── CLAP score (Layer 2) ─────────────────────────────────────────
+    clap_ext = _parse_clap_result(clap_result)
+
+    # ── FST score (Layer 3) ──────────────────────────────────────────
+    fst_ext = _parse_fst_result(fst_result)
+
+    # ── Legacy external service scores ───────────────────────────────
     ext_music_ai = _parse_external_music_ai(external_music_ai)
     ext_ses = _parse_external_ses_analizi(external_ses_analizi)
 
     # ── Fusion logic ─────────────────────────────────────────────────
     confidence, decision_source, mode = _fuse(
-        local_score, ext_music_ai, ext_ses
+        local_score, clap_ext, fst_ext, ext_music_ai, ext_ses
     )
 
     is_ai = confidence > 0.5
 
     # ── Build indicators ─────────────────────────────────────────────
     indicators.extend(_build_indicators(
-        is_ai, confidence, features, vocals, ext_music_ai, ext_ses
+        is_ai, confidence, features, vocals,
+        clap_ext, fst_ext, ext_music_ai, ext_ses,
     ))
 
     return FusionResult(