Upload app.py with huggingface_hub

app.py

@@ -46,16 +46,21 @@ def run_all_agents(img: Image.Image) -> Tuple[List[AgentEvidence], ForensicVerdi
         ratio = max_dim / max(w, h)
         img = img.resize((int(w * ratio), int(h * ratio)), Image.LANCZOS)
 
-    # Signal processing agents (fast, run in parallel)
+    # ── Capture Modality Detection (runs BEFORE agents) ───────────────
+    from agents.modality_detector import detect_modality
+    modality = detect_modality(img)
+    adj = modality.score_adjustments
+
+    # Signal processing agents (fast, run in parallel with modality adjustments)
     signal_agents = [
-        ("optical", run_optical_agent),
-        ("sensor", run_sensor_agent),
-        ("model", run_model_agent),
-        ("statistical", run_statistical_agent),
-        ("metadata", run_metadata_agent),
+        ("optical", lambda i: run_optical_agent(i, adj)),
+        ("sensor", lambda i: run_sensor_agent(i, adj)),
+        ("model", lambda i: run_model_agent(i, adj)),
+        ("statistical", lambda i: run_statistical_agent(i, adj)),
+        ("metadata", lambda i: run_metadata_agent(i, adj)),
     ]
 
-    # VLM agents (slower, also parallel)
+    # VLM agents (no modality adjustment needed — VLM sees the image directly)
     vlm_agents = [
         ("semantic", run_semantic_agent),
         ("text", run_text_agent),
@@ -73,7 +78,6 @@ def run_all_agents(img: Image.Image) -> Tuple[List[AgentEvidence], ForensicVerdi
             try:
                 results[name] = future.result()
             except Exception as e:
-                # Create a failed agent evidence
                 results[name] = AgentEvidence(
                     agent_name=f"{name.title()} Agent (Error)",
                     violation_score=0.0,
@@ -97,6 +101,14 @@ def run_all_agents(img: Image.Image) -> Tuple[List[AgentEvidence], ForensicVerdi
     # Bayesian synthesis
     verdict = bayesian_synthesis(ordered)
 
+    # Attach modality info to verdict for reporting
+    verdict.reasoning_tree["modality"] = {
+        "detected": modality.modality,
+        "confidence": modality.confidence,
+        "indicators": {k: v for k, v in modality.indicators.items() if not isinstance(v, np.ndarray)},
+        "adjustments_applied": len(modality.score_adjustments),
+    }
+
     # Generate explanations
     verdict.forensic_report = generate_forensic_report(verdict)
     reasoning = generate_reasoning_tree(verdict)