Update app.py

app.py

@@ -8,6 +8,15 @@ import matplotlib.pyplot as plt
 import io
 from PIL import Image
 from datetime import datetime
+from transformers import pipeline as hf_pipeline  # prevent name collision with gradio pipeline
+
+# Emotion model (no retraining needed)
+emotion_pipeline = hf_pipeline(
+    "text-classification",
+    model="j-hartmann/emotion-english-distilroberta-base",
+    top_k=None,
+    truncation=True
+)
 
 # --- Timeline Visualization Function ---
 def generate_abuse_score_chart(dates, scores, labels):
@@ -130,6 +139,9 @@ DARVO_MOTIFS = [
     "You’re the one who’s always making me feel like I’m the one who’s abusive.",
     "You’re the one who’s always making me feel like I’m the one who’s toxic."
 ]
+def get_emotion_profile(text):
+    emotions = emotion_pipeline(text)
+    return {e['label'].lower(): round(e['score'], 3) for e in emotions}
 def detect_contradiction(message):
     patterns = [
         (r"\b(i love you).{0,15}(i hate you|you ruin everything)", re.IGNORECASE),
@@ -319,7 +331,11 @@ def analyze_single_message(text, thresholds):
         abuse_score_raw = min(abuse_score_raw + 25, 100)
 
     abuse_score = min(abuse_score_raw, 100 if "threat" in threshold_labels or "control" in threshold_labels else 95)
-
+    # New: Emotion Profile
+    emotion_profile = get_emotion_profile(text)
+    print("Emotion Profile:")
+    for emotion, score in emotion_profile.items():
+        print(f"  {emotion.capitalize():10}: {score}")
     print("\n--- Debug Info ---")
     print(f"Text: {text}")
     print(f"Sentiment: {sentiment} (raw: {result['label']}, score: {result['score']:.3f})")