Update app.py

app.py

@@ -1,4 +1,5 @@
 import streamlit as st
+from sentence_transformers import SentenceTransformer
 from PIL import Image
 import pinecone
 import pinecone_text
@@ -48,8 +49,6 @@ index_name = "canopy--document-uploader"  # Replace with your chosen index name
 
 index = pc.Index(name=index_name)
 
-# Initialize Pinecone Text Client for embedding
-encoder = pinecone_text.OpenAIEncoder()
 
 # Step 1: Clone the Dataset Repository
 repo = Repository(
@@ -254,10 +253,15 @@ def query_pinecone(vector, index, top_k=5):
     query_results = index.query(vector=vector, top_k=top_k)
     return query_results["matches"]
 
+from sentence_transformers import SentenceTransformer
+
+# Initialize the Sentence Transformer model
+model = SentenceTransformer('all-MiniLM-L6-v2')
+
 def text_to_vector(text):
-    # Convert input text to vector using Pinecone Text Client
-    embedding = encoder.encode([text])[0]  # Assume single text input; adjust accordingly
-    return embedding
+    # Convert input text to vector
+    embedding = model.encode(text)
+    return embedding  # No need to convert to list, depending on how you use it later
 
 
 
@@ -505,6 +509,7 @@ def page2():
 
 def page3():
     try:
+        # Style adjustments for the Streamlit page
         st.markdown("""
             <style>
             #MainMenu {visibility: hidden;}
@@ -512,6 +517,7 @@ def page3():
             </style>
             """, unsafe_allow_html=True)
     
+        # Layout configuration
         col1, col2 = st.columns([3, 1])
         with col1:
             st.title("Kosten- und Strukturdaten der Krankenhäuser")
@@ -521,32 +527,38 @@ def page3():
                 image = Image.open(image_path)
                 st.image(image, use_column_width='always')
 
+        # Display existing chat history
         display_chat_history(st.session_state['chat_history_page3'])
 
+        # Input for new user query
         query = st.text_input("Geben Sie hier Ihre Frage ein / Enter your question here:")
 
         if query:
-            # Convert the query text to a vector using Pinecone Text Client
-            query_vector = encoder.encode([query])[0]
-            matches = query_pinecone(query_vector, top_k=5)
+            # Convert the query text to a vector
+            query_vector = text_to_vector(query)
+
+            # Query the Pinecone index with the vector
+            matches = index.query(queries=[query_vector], top_k=5)["matches"][0]
 
-            response_messages = []
+            # Display the results and update chat history
             for match in matches:
                 matched_text = match["metadata"].get("summary", "Detailed information not available.")
                 similarity_score = match["score"]
-                response_messages.append((f"Pinecone Match - Score: {similarity_score:.2f}", matched_text))
+                response_message = f"Matched Text: {matched_text} - Score: {similarity_score:.2f}"
+                st.write(response_message)
+                # Append the response to the chat history
+                st.session_state['chat_history_page3'].append(("Eve", response_message))
 
-            # Update session state with the new chat and response
-            st.session_state['chat_history_page3'].extend([(query, "User Query")] + response_messages)
-            
-            # Display the new responses
-            display_chat_history([(query, "User Query")] + response_messages)
+            # Append the user query to the chat history
+            st.session_state['chat_history_page3'].append(("User", query))
 
-            # Save the updated conversation to a file
-            save_conversation(st.session_state['chat_history_page3'], st.session_state['session_id'])
+            # Save the updated chat history to a session or external storage as needed
+            # This is a placeholder for where you might save the chat history
+            # save_conversation(st.session_state['chat_history_page3'], st.session_state['session_id'])
 
     except Exception as e:
         st.error(f"An unexpected error occurred: {e}")
+