fixed emotions issues

app.py

@@ -1,5 +1,5 @@
 import spaces
-from transformers import AutoTokenizer, AutoModelForCausalLM
+from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
 import torch
 import gradio as gr
 from PIL import Image
@@ -33,6 +33,7 @@ If you encounter unclear input, politely ask for clarification instead of guessi
 
 Respond to the following input:
 """
+sentiment_analyzer = pipeline("sentiment-analysis")
 
 # Emotional states
 emotions = {
@@ -51,32 +52,52 @@ emotions.update({
 })
 
 # Add keywords for new emotions in `analyze_history`
-
+emotion_keywords = {
+    "happy": ["happy", "joy", "excited", "awesome"],
+    "playful": ["fun", "play", "joke", "quirky"],
+    "curious": ["curious", "wonder", "question", "thinking"],
+    "thoughtful": ["sad", "thoughtful", "hmm", "ponder"],
+    "concerned": ["error", "wrong", "problem", "issue", "stuck"],
+    "mischievous": ["trouble", "mischief", "sneaky", "prank"]
+}
     
 current_emotion = "happy"
 
 # Analyze history for emotional state
 def analyze_history(history):
-    # Analyze recent history for emotion keywords
-    recent_messages = " ".join(history[-5:])  # Check the last 5 messages
-    error_count = recent_messages.lower().count("error")
-    happy_count = recent_messages.lower().count("happy")
-    sad_count = recent_messages.lower().count("sad")
-    excited_count = recent_messages.lower().count("excited")
-    trouble_count = recent_messages.lower().count("trouble")
-
-    if excited_count > 0:
-        return "excited"
-    elif trouble_count > 0:
-        return "mischievous"
-    elif error_count > 2:
-        return "concerned"
-    elif happy_count > sad_count:
-        return "happy"
-    elif sad_count > happy_count:
-        return "thoughtful"
+    recent_messages = " ".join(history[-5:]).lower()
+    print(f"Analyzing history: {recent_messages}")  # Debug log
+
+    # Count keyword matches
+    keyword_counts = {emotion: sum(recent_messages.count(keyword) for keyword in keywords)
+                      for emotion, keywords in emotion_keywords.items()}
+    print(f"Keyword counts: {keyword_counts}")
+
+    # Perform sentiment analysis
+    sentiment_result = sentiment_analyzer(recent_messages)
+    sentiment = sentiment_result[0]["label"]
+    sentiment_score = sentiment_result[0]["score"]
+    print(f"Sentiment analysis: {sentiment}, Score: {sentiment_score}")
+
+    # Determine sentiment-based emotion
+    if sentiment == "POSITIVE":
+        sentiment_emotion = "happy"
+    elif sentiment == "NEGATIVE":
+        sentiment_emotion = "thoughtful"
     else:
-        return "curious"
+        sentiment_emotion = "curious"
+
+    # Combine results using weights
+    combined_scores = {emotion: keyword_counts.get(emotion, 0) for emotion in emotion_keywords}
+    combined_scores[sentiment_emotion] += sentiment_score * 2  # Give more weight to sentiment analysis
+    print(f"Combined scores: {combinsed_scores}")
+
+    # Choose the emotion with the highest score
+    detected_emotion = max(combined_scores, key=combined_scores.get)
+    print(f"Detected emotion: {detected_emotion}")
+
+    return detected_emotions
+