Update page1.py

page1.py

@@ -12,18 +12,15 @@ import io
 import time
 from PIL import Image
 import os
+
 # Set your Google API key here
 GOOGLE_API_KEY = os.environ.get("api_key")
 
 
 def convert_to_base64(uploaded_file):
-    """Convert uploaded image to Base64 format (supports JPEG and PNG)"""
     image = Image.open(uploaded_file)
     buffered = io.BytesIO()
-
-    # Preserve format (default to PNG if unknown)
     format = image.format if image.format in ["JPEG", "PNG"] else "PNG"
-
     image.save(buffered, format=format)
     return base64.b64encode(buffered.getvalue()).decode("utf-8")
 
@@ -31,14 +28,12 @@ def convert_to_base64(uploaded_file):
 def text():
     st.title("Gemini 2.0 Thinking Experimental")
     st.sidebar.title("Capabilities:")
-
-    # Add bullet points
     st.sidebar.markdown("""
-            - **Text Queries**
-            - **Visual Queries**
-            - **PDF Support**
-           
-            """)
+        - **Text Queries**
+        - **Visual Queries**
+        - **PDF Support**
+    """)
+
     st.markdown("""
     <style>
         .anim-typewriter {
@@ -73,7 +68,6 @@ def text():
     </style>
     """, unsafe_allow_html=True)
 
-    # Initialize session state
     if "messages" not in st.session_state:
         st.session_state.messages = []
         st.session_state.chat_history = StreamlitChatMessageHistory()
@@ -82,13 +76,13 @@ def text():
             memory_key="history",
             chat_memory=st.session_state.chat_history
         )
-    system_prompt = (
-        "You are a compassionate and emotionally intelligent AI assistant trained in cognitive behavioral therapy (CBT), "
-        "mindfulness, and active listening. You provide supportive, empathetic responses without making medical diagnoses. "
-        "Use a warm tone and guide users to explore their feelings, reframe thoughts, and reflect gently."
-    )    
-    st.session_state.chat_history.add_message(SystemMessage(content=system_prompt))
-    # Initialize Gemini model
+        system_prompt = (
+            "You are a compassionate and emotionally intelligent AI assistant trained in cognitive behavioral therapy (CBT), "
+            "mindfulness, and active listening. You provide supportive, empathetic responses without making medical diagnoses. "
+            "Use a warm tone and guide users to explore their feelings, reframe thoughts, and reflect gently."
+        )
+        st.session_state.chat_history.add_message(SystemMessage(content=system_prompt))
+
     llm = ChatGoogleGenerativeAI(
         model="gemini-2.5-flash",
         google_api_key=GOOGLE_API_KEY,
@@ -96,24 +90,15 @@ def text():
         streaming=True,
         timeout=120,
         max_retries=6
-
     )
-        # Add system prompt (psychological assistant)
-    
-    
-    # Add SystemMessage to chat history
-    #st.session_state.chat_history.add_message(SystemMessage(content=system_prompt))
-    # Display chat messages
+
     chat_container = st.container()
     with chat_container:
-        # Show initial bot message
         if len(st.session_state.messages) == 0:
             animated_text = '<div class="anim-typewriter">Hello 👋, how may I assist you today?</div>'
-            # st.chat_message("assistant").markdown(animated_text, unsafe_allow_html=True)
             st.session_state.messages.append({"role": "assistant", "content": "Hello 👋, how may I assist you today?"})
 
-        # Display historical messages
-        for message in st.session_state.messages[0:]:  # Skip first static message
+        for message in st.session_state.messages:
             if message["role"] == "user":
                 if message.get("image"):
                     st.chat_message("user", avatar="🧑").markdown(
@@ -125,7 +110,6 @@ def text():
             else:
                 st.chat_message("assistant", avatar="🤖").markdown(message["content"])
 
-    # Chat input with multimodal support
     user_input = st.chat_input("Say something", accept_file=True, file_type=["png", "jpg", "jpeg", "pdf"])
 
     if user_input:
@@ -133,22 +117,18 @@ def text():
         file_name = ""
         image_base64 = convert_to_base64("pdf_icon.png")
         image_url = f"data:image/jpeg;base64,{image_base64}"
-        # Process user input
-        #image_url = ""
         message_content = [{"type": "text", "text": user_input.text}]
         files = user_input["files"]
 
-
         if files:
             file_type = files[0].type
-            
 
         if file_type in ["image/png", "image/jpg", "image/jpeg"]:
             uploaded_file = user_input["files"][0]
             image_base64 = convert_to_base64(uploaded_file)
             image_url = f"data:image/jpeg;base64,{image_base64}"
-
             message_content.append({"type": "image_url", "image_url": image_url})
+
         text = ""
         if file_type == "application/pdf":
             uploaded_file = user_input["files"][0]
@@ -156,12 +136,10 @@ def text():
             pdf_reader = PdfReader(uploaded_file)
             for page in pdf_reader.pages:
                 text += page.extract_text()
-            #st.sidebar.write(text)
-            prompt = "this is pdf data: \n"+text +"this is user asking about pdf:"+user_input.text
+            prompt = "this is pdf data: \n" + text + "this is user asking about pdf:" + user_input.text
             message_content = [{"type": "text", "text": prompt}]
             message_content.append({"type": "text", "text": file_name})
-            #message_content.append({"type": "image_url", "image_url": image_url})
-        # Add user message to UI
+
         with chat_container:
             if file_type:
                 st.chat_message("user", avatar="🧑").markdown(
@@ -174,34 +152,30 @@ def text():
                     """,
                     unsafe_allow_html=True
                 )
-
             else:
                 st.chat_message("user", avatar="🧑").markdown(user_input.text)
 
-        # Store in session state
         st.session_state.messages.append({
             "role": "user",
             "content": user_input.text,
             "image": image_url if user_input["files"] else "",
-            "file_name" : file_name,
-            "file_type" : file_type
+            "file_name": file_name,
+            "file_type": file_type
         })
 
-        # Create LangChain message
         user_message = HumanMessage(content=message_content)
         st.session_state.chat_history.add_message(user_message)
 
-        # Generate streaming response
+        # Ensure valid message history (SystemMessage only at index 0)
         history = st.session_state.chat_history.messages
+        valid_history = [msg for msg in history if not isinstance(msg, SystemMessage)]
+        valid_history = [history[0]] + valid_history  # Keep the first SystemMessage only
+
         typing_container = st.empty()
 
-        def stream_generator(history, user_message):
-            # Placeholder for "Thinking..." and "Typing..."
+        def stream_generator(valid_history, user_message):
             typing_container = st.empty()
-
-            # Show "Thinking..." first
             typing_container.markdown('<p class="fade-text">Thinking...</p>', unsafe_allow_html=True)
-
             st.markdown("""
                 <style>
                     @keyframes fade {
@@ -218,17 +192,11 @@ def text():
                 </style>
             """, unsafe_allow_html=True)
 
-            response = llm.stream(history + [user_message])
-
-            # Buffer for partial words
+            response = llm.stream(valid_history + [user_message])
             buffer = ""
-
-            # Flag to change message
             first_chunk_received = False
-
-            # Pause settings
             PAUSE_AFTER = {".", "!", "?", ",", ";", ":"}
-            PAUSE_MULTIPLIER = 2.5  # Pause longer for punctuation
+            PAUSE_MULTIPLIER = 2.5
 
             for chunk in response:
                 if not first_chunk_received:
@@ -239,51 +207,37 @@ def text():
                 content = buffer + chunk.content
                 words = content.split(' ')
 
-                # Check if last word is complete
                 if not content.endswith(' '):
                     buffer = words.pop()
                 else:
                     buffer = ""
 
                 for word in words:
-                    yield word + ' '  # Stream word-by-word
-
-                    # Add delay for natural pauses
+                    yield word + ' '
                     base_delay = 0.03
                     last_char = word[-1] if word else ''
                     time.sleep(base_delay * PAUSE_MULTIPLIER if last_char in PAUSE_AFTER else base_delay)
 
-            # Yield any remaining content in buffer
             if buffer:
                 yield buffer
                 time.sleep(0.03)
 
-            # Clear "Typing..." message after response finishes
             typing_container.empty()
 
-        # Generate streaming response
         with st.chat_message("assistant", avatar="🤖"):
-
             full_response = st.write_stream(
-                stream_generator(
-                    st.session_state.chat_history.messages,
-                    user_message
-                )
+                stream_generator(valid_history, user_message)
             )
+            typing_container.empty()
 
-            typing_container.empty()  # Remove status message
-
-        # Update session state
         st.session_state.messages.append({
             "role": "assistant",
             "content": full_response
         })
 
-        # Update conversation memory
         ai_message = AIMessage(content=full_response)
         st.session_state.chat_history.add_message(ai_message)
         st.session_state.memory.save_context(
             {"input": user_message.content},
             {"output": ai_message.content}
         )
-        #st.sidebar.write(user_message)