Update app.py

app.py

@@ -8,33 +8,40 @@ import requests
 from datetime import datetime
 from transformers import pipeline
 
-# 🎙️ 1) Load detection model
+# 🎙️ Load Hugging Face model for non-verbal vocalization classification
 classifier = pipeline(
     "audio-classification",
     model="padmalcom/wav2vec2-large-nonverbalvocalization-classification"
 )
 
-# === Helpers ===
-
+# === Audio Conversion ===
 def convert_audio(input_path, output_path="input.wav"):
-    cmd = [
-        "ffmpeg", "-i", input_path,
-        "-acodec", "pcm_s16le", "-ar", "16000", "-ac", "1",
-        output_path, "-y"
-    ]
-    subprocess.run(cmd, check=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
+    try:
+        cmd = [
+            "ffmpeg", "-i", input_path,
+            "-acodec", "pcm_s16le", "-ar", "16000", "-ac", "1",
+            output_path, "-y"
+        ]
+        subprocess.run(cmd, check=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
+    except subprocess.CalledProcessError as e:
+        print(f"[ERROR] FFmpeg failed: {e.stderr.decode()}")
+        raise
     return output_path
 
+# === Scream Detection ===
 def detect_scream(audio_path):
     audio, sr = librosa.load(audio_path, sr=16000)
     results = classifier(audio)
+    print(f"[DEBUG] Raw classifier output: {results}")  # For debugging label issues
+
     if not results:
         return {"label": None, "score": 0.0}
     top = results[0]
     return {"label": top["label"].lower(), "score": float(top["score"]) * 100}
 
+# === Salesforce Alerting ===
 def send_salesforce_alert(audio_meta, detection):
-    SF_URL = os.getenv("SF_ALERT_URL")  # e.g. Salesforce REST endpoint
+    SF_URL = os.getenv("SF_ALERT_URL")
     SF_TOKEN = os.getenv("SF_API_TOKEN")
     headers = {"Authorization": f"Bearer {SF_TOKEN}", "Content-Type": "application/json"}
     payload = {
@@ -44,26 +51,36 @@ def send_salesforce_alert(audio_meta, detection):
         "AlertLevel": audio_meta["alert_level"],
         "Timestamp": audio_meta["timestamp"],
     }
-    # POST to Salesforce for alert, metadata and audit
-    resp = requests.post(SF_URL, json=payload, headers=headers, timeout=5)
-    resp.raise_for_status()
-    return resp.json()
 
-# 🎯 2) Main processing function called by Gradio
+    try:
+        resp = requests.post(SF_URL, json=payload, headers=headers, timeout=5)
+        resp.raise_for_status()
+        return resp.json()
+    except requests.exceptions.RequestException as e:
+        print(f"[ERROR] Salesforce alert failed: {e}")
+        if resp is not None:
+            print(f"Response: {resp.status_code} - {resp.text}")
+        raise
+
+# === Main Gradio Processing Function ===
 def process_uploaded(audio_file, start_stop, high_thresh, med_thresh):
     if start_stop != "Start":
         return "System is stopped."
 
-    wav_path = convert_audio(audio_file)
-    detection = detect_scream(wav_path)
+    try:
+        wav_path = convert_audio(audio_file)
+    except Exception as e:
+        return f"Audio conversion failed: {e}"
 
+    detection = detect_scream(wav_path)
     label = detection["label"]
     score = detection["score"]
+    print(f"[INFO] File: {audio_file}, Label: {label}, Score: {score:.2f}%")
 
-    # 3) Determine alert level
-    if label == "scream" and score >= high_thresh:
+    # Determine risk level based on score
+    if label and "scream" in label and score >= high_thresh:
         level = "High-Risk"
-    elif label == "scream" and score >= med_thresh:
+    elif label and "scream" in label and score >= med_thresh:
         level = "Medium-Risk"
     else:
         level = "None"
@@ -74,7 +91,7 @@ def process_uploaded(audio_file, start_stop, high_thresh, med_thresh):
         "alert_level": level
     }
 
-    # 4) Trigger alert if needed
+    # Trigger alert if needed
     if level in ("High-Risk", "Medium-Risk"):
         try:
             sf_resp = send_salesforce_alert(audio_meta, detection)
@@ -84,7 +101,7 @@ def process_uploaded(audio_file, start_stop, high_thresh, med_thresh):
 
     return f"Detection: {label} ({score:.1f}%) — Alert: {level}"
 
-# 🎚️ 5) Setup Gradio Interface with config and control
+# === Gradio Interface ===
 iface = gr.Interface(
     fn=process_uploaded,
     inputs=[
@@ -96,16 +113,15 @@ iface = gr.Interface(
     outputs="text",
     title="📢 Scream Detection & Salesforce Alerts",
     description="""
-- Upload an audio sample or stream via Raspberry Pi integration.  
-- **Start/Stop** toggles detection.  
-- **Threshold sliders** let you configure alert sensitivities.  
-- On detection, alerts are logged and pushed to Salesforce.  
-- All metadata and alerts are stored for dashboards and auditing.
+Upload an audio sample or stream from a device like Raspberry Pi.  
+- **Start/Stop** detection system.  
+- **Adjust thresholds** for sensitivity.  
+- Alerts are sent to Salesforce for audit, dashboard, and real-time response.
 """,
     allow_flagging="never"
 )
 
-# Optional: thread for real-time, device-based listening (e.g., for Pi)
+# === Optional: Real-Time Listener (e.g., for Raspberry Pi) ===
 def pi_listener(high_thresh=80, med_thresh=50, interval=1.0):
     import sounddevice as sd
     import numpy as np
@@ -115,9 +131,9 @@ def pi_listener(high_thresh=80, med_thresh=50, interval=1.0):
         detection = classifier(wav.astype(np.float32))
         lbl, sc = (detection[0]["label"].lower(), detection[0]["score"] * 100)
         level = "None"
-        if lbl == "scream" and sc >= high_thresh:
+        if "scream" in lbl and sc >= high_thresh:
             level = "High-Risk"
-        elif lbl == "scream" and sc >= med_thresh:
+        elif "scream" in lbl and sc >= med_thresh:
             level = "Medium-Risk"
         if level != "None":
             timestamp = datetime.utcnow().isoformat() + "Z"
@@ -126,13 +142,15 @@ def pi_listener(high_thresh=80, med_thresh=50, interval=1.0):
                 {"label": lbl, "score": sc}
             )
             print(f"[{timestamp}] {level} scream detected ({sc:.1f}%) – alert sent.")
+
     with sd.InputStream(channels=1, samplerate=16000, callback=callback):
-        print("Raspberry Pi listener running... Press Ctrl+C to stop.")
+        print("🎧 Raspberry Pi real-time listener running...")
         while True:
             time.sleep(interval)
 
+# === Launch Gradio App ===
 if __name__ == "__main__":
-    # Optional: uncomment next lines to start Pi listener in background
+    # Optional background listener thread (uncomment if needed)
     # pi_thread = threading.Thread(target=pi_listener, daemon=True)
     # pi_thread.start()