Update app.py

app.py

@@ -14,16 +14,11 @@ import matplotlib.pyplot as plt
 from huggingface_hub import login
 import os
 from dotenv import load_dotenv
-import threading
-
 
 SPREADSHEET_ID = "1CsBub3Jlwyo7WHMQty6SDnBShIZMjl5XTVSoOKrxZhc"
 RANGE_NAME = 'Sheet1!A1:E'
 SERVICE_ACCOUNT_FILE = r"C:\Users\bhagy\AI\credentials.json"
-
-
-csv_file_path = r"C:\Users\bhagy\Downloads\900_products_dataset.csv"
-
+csv_file_path = r"C:\Users\bhagy\OneDrive\Desktop\INFOSYS PROJECT\900_products_dataset.csv"
 
 class CustomEmbeddingFunction:
     def __init__(self, model_name="sentence-transformers/all-MiniLM-L6-v2"):
@@ -37,14 +32,8 @@ class CustomEmbeddingFunction:
         embeddings = outputs.last_hidden_state.mean(dim=1).squeeze().numpy()
         return embeddings
 
-
-sentiment_pipeline = pipeline("sentiment-analysis")
-# persist_directory = "chromadb_storage"
-# chroma_client = PersistentClient(path=persist_directory)
 persist_directory = "chromadb_storage"
-os.makedirs(persist_directory, exist_ok=True)  # Ensure the directory exists
 chroma_client = PersistentClient(path=persist_directory)
-
 collection_name = "crm_data"
 
 try:
@@ -92,7 +81,7 @@ hf_token= os.getenv("HUGGINGFACE_TOKEN")
 login(token=hf_token)
 if not hf_token:
     raise ValueError("Hugging Face API key not found! Please set the HUGGINGFACE_TOKEN variable.")
-print(f"API Key Loaded: {hf_token[:5]}****")
+print(f"API Key Loaded: {hf_token[:5]}")
 
 model_name = "tabularisai/multilingual-sentiment-analysis"
 model = AutoModelForSequenceClassification.from_pretrained(model_name)
@@ -110,8 +99,6 @@ def analyze_sentiment(text):
         result = sentiment_analyzer(processed_text)[0]
         
         print(f"Sentiment Analysis Result: {result}")
-        
-        
         sentiment_map = {
             'Very Negative': "NEGATIVE",
             'Negative': "NEGATIVE",
@@ -134,17 +121,13 @@ def load_csv(file_path):
         if data is not None:
             st.session_state.crm_data = data  
             print("CRM data loaded successfully!")
-            return data
-        else:
-            st.error("Failed to load CRM data: File is empty or invalid.")
+        return data
     except Exception as e:
-        st.error(f"Error loading CSV: {e}")
         print(f"Error loading CSV: {e}")
         return None
 
 data = load_csv(csv_file_path)
 
-    
 def process_crm_data(data):
     try:
         chunks = [str(row) for row in data.to_dict(orient="records")]
@@ -186,8 +169,6 @@ def query_crm_data_with_context(prompt, top_k=3):
         st.error(f"Error querying CRM data: {e}")
         return ["Error in querying recommendations."]
 
-
-
 sentence_model = SentenceTransformer('sentence-transformers/all-MiniLM-L6-v2')
 faiss_index = faiss.IndexFlatL2(384)
 
@@ -202,13 +183,8 @@ def load_objection_responses(csv_file_path):
 
 objection_response_pairs = load_objection_responses(r"C:\Users\bhagy\OneDrive\Desktop\INFOSYS PROJECT\objections_responses.csv")
 objections = list(objection_response_pairs.keys())
-# objection_embeddings = sentence_model.encode(objections)
-# faiss_index.add(np.array(objection_embeddings, dtype="float32"))
-
 objection_embeddings = sentence_model.encode(objections)
-objection_embeddings = objection_embeddings.reshape(-1, 384)  # Reshape to 2D array
-faiss_index.add(objection_embeddings.astype("float32"))
-
+faiss_index.add(np.array(objection_embeddings, dtype="float32"))
 
 def find_closest_objection(query):
     query_embedding = sentence_model.encode([query])
@@ -300,7 +276,7 @@ def generate_comprehensive_summary(chunks):
     summary += f"• Neutral Interactions: {neutral_count}\n"
     
     summary += "\nKey Conversation Points:\n"
-    for interaction in key_interactions[:3]:  # Limit to top 3 key points
+    for interaction in key_interactions[:3]: 
         summary += f"• {interaction}\n"
     
     if positive_count > negative_count:
@@ -325,8 +301,8 @@ def show_help():
 
     st.header("1. Introduction to the AI Assistant")
     st.write("""
-        - **What It Does**: The assistant analyzes live sales calls in real-time. It detects sentiment shifts, provides product recommendations, and suggests dynamic question handling techniques.
-        - **Key Features**:
+        - *What It Does*: The assistant analyzes live sales calls in real-time. It detects sentiment shifts, provides product recommendations, and suggests dynamic question handling techniques.
+        - *Key Features*:
             - Real-time speech-to-text conversion and sentiment analysis.
             - Product recommendations based on customer context.
             - Dynamic question prompt generator.
@@ -336,58 +312,58 @@ def show_help():
 
     st.header("2. Getting Started")
     st.write("""
-        - **How to Start a Call**: To start a sales call, Click on Start Listening. Once connected, initiate the call, and the assistant will begin analyzing.
-        - **What to Expect**: During the call, the assistant will provide real-time feedback, such as sentiment scores, product recommendations, and objection handling tips.
+        - *How to Start a Call*: To start a sales call, Click on Start Listening. Once connected, initiate the call, and the assistant will begin analyzing.
+        - *What to Expect*: During the call, the assistant will provide real-time feedback, such as sentiment scores, product recommendations, and objection handling tips.
     """)
 
     st.header("3. Using the Assistant During Sales Calls")
     st.write("""
-        - **Speech-to-Text Instructions**: Speak clearly into your microphone for the assistant to accurately capture and analyze your speech.
-        - **Real-time Feedback**: The assistant will display real-time feedback on the sentiment of the conversation, suggest responses for objections, and provide product recommendations.
+        - *Speech-to-Text Instructions*: Speak clearly into your microphone for the assistant to accurately capture and analyze your speech.
+        - *Real-time Feedback*: The assistant will display real-time feedback on the sentiment of the conversation, suggest responses for objections, and provide product recommendations.
     """)
 
 
     st.header("4. Understanding the Interface")
     st.write("""
-        - **Tabs Navigation**: The interface has different tabs:
-            - **Call Summary**: After the call, review the summary, which highlights conversation key points.
-            - **Sentiment Analysis**: See how the sentiment changed throughout the conversation.
-            - **Product Recommendations**: View the recommended products based on customer intent and conversation context.
+        - *Tabs Navigation*: The interface has different tabs:
+            - *Call Summary*: After the call, review the summary, which highlights conversation key points.
+            - *Sentiment Analysis*: See how the sentiment changed throughout the conversation.
+            - *Product Recommendations*: View the recommended products based on customer intent and conversation context.
     """)
 
 
     st.header("5. FAQs and Troubleshooting")
     st.write("""
-        - **Sentiment Detection Accuracy**: If the assistant's sentiment analysis isn't accurate, ensure you speak clearly and avoid background noise.
-        - **Speech Recognition Issues**: Rephrase unclear statements and ensure the microphone is working well.
-        - **Context Handling**: If the assistant misses some context, remind it of the product or the customer’s intent.
+        - *Sentiment Detection Accuracy*: If the assistant's sentiment analysis isn't accurate, ensure you speak clearly and avoid background noise.
+        - *Speech Recognition Issues*: Rephrase unclear statements and ensure the microphone is working well.
+        - *Context Handling*: If the assistant misses some context, remind it of the product or the customer’s intent.
     """)
 
 
     st.header("6. Support and Contact Information")
     st.write("""
-        - **Live Chat Support**: Chat with us in real-time by clicking the support icon in the bottom right.
-        - **Email and Phone Support**: You can also reach us at [email protected] or call us at +1-800-555-1234.
-        - **Feedback**: Please provide feedback to help us improve the assistant.
+        - *Live Chat Support*: Chat with us in real-time by clicking the support icon in the bottom right.
+        - *Email and Phone Support*: You can also reach us at [email protected] or call us at +1-800-555-1234.
+        - *Feedback*: Please provide feedback to help us improve the assistant.
     """)
 
     st.header("7. Advanced Features")
     st.write("""
-        - **Integration with CRM and Google Sheets**: Sync with CRM systems and Google Sheets to enhance product recommendations.
-        - **Customization Options**: Customize the assistant’s tone, product categories, and question prompts through the settings tab.
+        - *Integration with CRM and Google Sheets*: Sync with CRM systems and Google Sheets to enhance product recommendations.
+        - *Customization Options*: Customize the assistant’s tone, product categories, and question prompts through the settings tab.
     """)
 
     st.header("8. Privacy and Security")
     st.write("""
-        - **Data Privacy**: All conversations are anonymized for analysis purposes. We ensure compliance with privacy regulations.
-        - **Security Protocols**: All data is encrypted and stored securely.
+        - *Data Privacy*: All conversations are anonymized for analysis purposes. We ensure compliance with privacy regulations.
+        - *Security Protocols*: All data is encrypted and stored securely.
     """)
 
 
     st.header("9. Updates and New Features")
     st.write("""
-        - **Changelog**: We release regular updates to improve performance. Please refer to the changelog for new features and improvements.
-        - **How to Update**: If an update is available, follow the instructions in the settings tab to install the latest version.
+        - *Changelog*: We release regular updates to improve performance. Please refer to the changelog for new features and improvements.
+        - *How to Update*: If an update is available, follow the instructions in the settings tab to install the latest version.
     """)
 def calculate_overall_sentiment(sentiment_scores):
     if sentiment_scores:
@@ -418,14 +394,14 @@ def process_real_time_audio():
 
 
             st.write("Transcribing audio...")
-            transcribed_text = recognizer.recognize_sphinx(audio)
+            transcribed_text = recognizer.recognize_google(audio)
             st.write(f"You said: {transcribed_text}")
 
             if 'stop' in transcribed_text.lower():
                 st.warning("Stopping the speech recognition process.")
                 break
 
-            st.markdown("### **Sentiment Analysis**")
+            st.markdown("### *Sentiment Analysis*")
             sentiment_label, sentiment_score = analyze_sentiment(transcribed_text)
             st.write(f"Sentiment: {sentiment_label}")
             st.write(f"Sentiment Score: {sentiment_score}")
@@ -434,15 +410,15 @@ def process_real_time_audio():
             response = None
 
             add_to_sentiment_history(transcribed_text, sentiment_label, sentiment_score, closest_objection, response)
-            st.markdown("### **Recommendations**")
+            st.markdown("### *Recommendations*")
             recommendations = query_crm_data_with_context(transcribed_text)
             for i, rec in enumerate(recommendations, start=1):
                 if isinstance(rec, dict) and 'Product' in rec and 'Recommendations' in rec:
-                    st.markdown(f"- **{rec['Product']}**: {rec['Recommendations']}")
+                    st.markdown(f"- *{rec['Product']}*: {rec['Recommendations']}")
                 else:
                     st.markdown(f"- {rec}")
 
-            st.markdown("### **Objection Handling**")
+            st.markdown("### *Objection Handling*")
             closest_objection, response = find_closest_objection(transcribed_text)
             st.write(f"Objection: {closest_objection}")
             st.write(f" Response: {response}")
@@ -460,8 +436,6 @@ def process_real_time_audio():
         except Exception as e:
             st.error(f"Error: {e}")
             break
-    speech_thread = threading.Thread(target=recognize_speech, daemon=True)
-    speech_thread.start()
   
 def generate_sentiment_pie_chart(sentiment_history):
     if not sentiment_history:
@@ -490,7 +464,7 @@ def generate_sentiment_pie_chart(sentiment_history):
     ax.pie(sizes, labels=labels, autopct='%1.1f%%', startangle=90, colors=colors,textprops={'fontsize':12, 'color':'white'})
     fig.patch.set_facecolor('none')
     ax.axis('equal')  
-    st.markdown("### *Sentiment Distribution*")
+    st.markdown("### Sentiment Distribution")
     st.pyplot(fig)
 
 def generate_post_call_summary(sentiment_history, recommendations=[]): 
@@ -503,7 +477,6 @@ def generate_post_call_summary(sentiment_history, recommendations=[]):
     summarizer = pipeline("summarization", model="facebook/bart-large-cnn")
     combined_text = " ".join([item["Text"] for item in sentiment_history])
 
-    # summary = summarizer(combined_text, max_length=100, min_length=30, do_sample=False)[0]["summary_text"]
     scores = [item["Score"] for item in sentiment_history]
 
     st.markdown("## Summary of the Call")
@@ -511,7 +484,7 @@ def generate_post_call_summary(sentiment_history, recommendations=[]):
     summary = generate_comprehensive_summary(chunks)
     st.write(summary)
 
-    st.markdown("### **Overall Sentiment for the Call**")
+    st.markdown("### *Overall Sentiment for the Call*")
     sentiment_scores = [entry["Score"] for entry in sentiment_history]
     overall_sentiment = calculate_overall_sentiment(sentiment_scores)
     st.write(f"Overall Sentiment: {overall_sentiment}")
@@ -522,7 +495,7 @@ def generate_post_call_summary(sentiment_history, recommendations=[]):
         plt.figure(figsize=(10, 6))
         plt.bar(range(len(sentiment_scores)), sentiment_scores, color=colors)
         plt.axhline(0, color='black', linestyle='--', linewidth=1, label='Neutral')
-        st.markdown("### **Sentiment Trend Bar Chart**")
+        st.markdown("### *Sentiment Trend Bar Chart*")
         plt.title("Sentiment Trend Throughout the Call")
         plt.xlabel("Segment")
         plt.ylabel("Sentiment Score")
@@ -533,7 +506,7 @@ def generate_post_call_summary(sentiment_history, recommendations=[]):
     with col2:
         generate_sentiment_pie_chart(sentiment_history)
 
-    st.markdown("### **Future Insights**")
+    st.markdown("### *Future Insights*")
     
    
     if overall_sentiment == "Negative":
@@ -545,28 +518,26 @@ def generate_post_call_summary(sentiment_history, recommendations=[]):
 
    
     if recommendations:
-        st.write("### **Product Recommendations**")
+        st.write("### *Product Recommendations*")
         for rec in recommendations:
             st.write(f"- {rec}")
 
     if sentiment_history:
-        st.write("### **Sentiment Breakdown by Segment**")
+        st.write("### *Sentiment Breakdown by Segment*")
         for idx, entry in enumerate(sentiment_history, 1):
             st.write(f"Segment {idx}: Sentiment = {entry['Sentiment']}, Score = {entry['Score']:.2f}")
 
-# Main
 def main():
     st.title("🤖 RealTime AI-Powered Sales Assistant For Enhanced Conversation")
     st.markdown(
         "An intelligent assistant to analyze speech, handle objections, and recommend products in real-time."
     )
 
-    # Tabs for navigation
-    tabs = st.tabs(["🎙️ Real-Time Audio", "📊 Text Search ", "📋 Visualization","🕘 Query History","❓Help","💬 Feedback"])
+    tabs = st.tabs(["🎙 Real-Time Audio", "📊 Text Search ", "📋 Visualization","🕘 Query History","❓Help","💬 Feedback"])
 
     
     with tabs[0]:
-        st.header("🎙️ Real-Time Audio Analysis")
+        st.header("🎙 Real-Time Audio Analysis")
         st.write(
             "Use this feature to analyze live speech, perform sentiment analysis, and get product recommendations."
         )
@@ -610,7 +581,6 @@ def main():
             st.dataframe(st.session_state["crm_history"])
 
     with tabs[4]:
-        # st.subheader("❓Help")
         show_help()
     
     with tabs[5]:
@@ -623,7 +593,6 @@ def main():
             st.session_state["app_feedback"].append(feedback)
             st.success("Thank you for your feedback!")
         
-        # Display previous feedback
         if st.session_state["app_feedback"]:
             st.write("### Previous Feedback:")
             for idx, feedback_entry in enumerate(st.session_state["app_feedback"], 1):