Update app.py

app.py

@@ -5,33 +5,22 @@ from PIL import Image, UnidentifiedImageError, ExifTags
 import requests
 from io import BytesIO
 import wikipedia
-from easygoogletranslate import EasyGoogleTranslate
 from BharatCaptioner import identify_landmark
 from groq import Groq
 import hashlib
+import time  # To simulate character-by-character display
 
-# Initialize EasyGoogleTranslate
-translator = EasyGoogleTranslate(source_language="en", target_language="hi", timeout=10)
-
-# Load configuration for Groq API key
-working_dir = os.path.dirname(os.path.abspath(__file__))
-config_data = json.load(open(f"{working_dir}/config.json"))
-GROQ_API_KEY = config_data["GROQ_API_KEY"]
-os.environ["GROQ_API_KEY"] = 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
-
+# Initialize Groq API client
+os.environ["GROQ_API_KEY"] = "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"
 client = Groq()
 
-# Title of the Streamlit app
 st.title("BharatCaptioner with Conversational Chatbot")
 st.write(
     "A tool to identify/describe Indian Landmarks in Indic Languages and chat about the image."
 )
 
 # Sidebar details
-st.sidebar.title("Developed by Harshal and Harsh Pandey")
-st.sidebar.write(
-    "**For the Model that I trained**: [Mail me here](mailto:harshal19052003@gmail.com)"
-)
+st.sidebar.title("Developed by Harshal")
 st.sidebar.write(
     "**For the Code**: [GitHub Repo](https://github.com/justharshal2023/BharatCaptioner)"
 )
@@ -43,13 +32,20 @@ st.sidebar.write(
 uploaded_file = st.file_uploader("Choose an image...", type=["jpg", "jpeg", "png"])
 url = st.text_input("Or enter a valid image URL...")
 
+# Initialize session state variables
+if "image_hash" not in st.session_state:
+    st.session_state["image_hash"] = None
+if "chat_history" not in st.session_state:
+    st.session_state["chat_history"] = []
+if "chatbot_started" not in st.session_state:
+    st.session_state["chatbot_started"] = False
+
 image = None
 error_message = None
 landmark = None
 summary = None
 caption = None
 
-
 # Function to correct image orientation
 def correct_image_orientation(img):
     try:
@@ -69,17 +65,15 @@ def correct_image_orientation(img):
         pass
     return img
 
-
 # Function to get a unique hash for the image
 def get_image_hash(image):
     img_bytes = image.tobytes()
     return hashlib.md5(img_bytes).hexdigest()
 
-
-# Check if new image or URL is uploaded and reset the chat history
+# Check if new image or URL is uploaded and reset the chat history if necessary
 def reset_chat_if_new_image():
-    if "last_uploaded_hash" not in st.session_state:
-        st.session_state["last_uploaded_hash"] = None
+    global image, landmark, summary, caption
+    new_image_hash = None
 
     # Process the new image or URL
     if uploaded_file:
@@ -102,45 +96,43 @@ def reset_chat_if_new_image():
             st.error(error_message)
     else:
         image = None
-        new_image_hash = None
 
     # If the image is new, reset the chat and session state
-    if new_image_hash and new_image_hash != st.session_state["last_uploaded_hash"]:
-        st.session_state.clear()
-        st.session_state["last_uploaded_hash"] = new_image_hash
-        st.experimental_rerun()
+    if new_image_hash and new_image_hash != st.session_state["image_hash"]:
+        st.session_state["image_hash"] = new_image_hash
+        st.session_state["chat_history"] = []
+        st.session_state["chatbot_started"] = False  # Reset chatbot status
 
     return image
 
-
 # Call the reset function to check for new images or URL
 image = reset_chat_if_new_image()
 
 # If an image is provided
 if image is not None:
-    # Resize image for processing
-    image = image.resize((256, 256))
+    # Keep the original image size for processing
+    original_image = image.copy()  # Create a copy for identification
 
     # Identify the landmark using BharatCaptioner
-    landmark, prob = identify_landmark(image)
+    landmark, prob = identify_landmark(original_image)
     summary = wikipedia.summary(landmark, sentences=3)  # Shortened summary
-    st.write(f"**Landmark Identified:** {landmark} (Confidence: {prob:.2f})")
+    st.write(f"**Landmark Identified:** {landmark}")
 
-    # Display image and landmark name in the sidebar
+    # Display a smaller version of the image in the sidebar
     with st.sidebar:
-        st.image(image, caption="Current Image", use_column_width=True)
-        st.write(f"**Landmark:** {landmark}")
+        small_image = original_image.resize((128, 128))  # Resize for display
+        st.image(small_image, caption=f"Landmark: {landmark}", use_column_width=True)
+        # st.write(f"**Landmark:** {landmark}")
+
+    # Display the original image before the conversation
+    st.image(original_image, caption=f"Image of {landmark}", use_column_width=True)
 
     # Chatbot functionality
     st.write("### Chat with the Chatbot about the Image")
     caption = f"The landmark in the image is {landmark}. {summary}"
 
-    # Initialize chat history in session state if not present
-    if "chat_history" not in st.session_state:
-        st.session_state["chat_history"] = []
-
-    # Chatbot introduction message with bold text for landmark and question
-    if not st.session_state.get("chatbot_started"):
+    # Chatbot introduction message
+    if not st.session_state["chatbot_started"]:
         chatbot_intro = f"Hello! I see the image is of **{landmark}**. {summary} **Would you like to know more** about this landmark?"
         st.session_state["chat_history"].append(
             {"role": "assistant", "content": chatbot_intro}
@@ -156,28 +148,37 @@ if image is not None:
     user_prompt = st.chat_input("Ask the Chatbot about the image...")
 
     if user_prompt:
+        st.session_state["chat_history"].append({"role": "user", "content": user_prompt})
         st.chat_message("user").markdown(user_prompt)
-        st.session_state.chat_history.append({"role": "user", "content": user_prompt})
 
-        # Send the user's message to the LLaMA chatbot
+        # Send the user's message to the chatbot
         messages = [
             {
                 "role": "system",
-                "content": "You are a helpful image conversational assistant. "
+                "content": "You are a helpful image conversational assistant, specialized in explaining about the monuments/landmarks of india. Give answer in points and in detail but dont hallucinate."
                 + f"The caption of the image is: {caption}",
             },
-            *st.session_state.chat_history,
+            *st.session_state["chat_history"],
         ]
 
+        # Simulate character-by-character response
         response = client.chat.completions.create(
             model="llama-3.1-8b-instant", messages=messages
         )
-
         assistant_response = response.choices[0].message.content
-        st.session_state.chat_history.append(
+
+        # Character-by-character output simulation
+        with st.chat_message("assistant"):
+            response_container = st.empty()  # Placeholder for response
+            response_text = ""
+            for char in assistant_response:
+                response_text += char
+                time.sleep(0.005)  # Adjust speed of character display
+                response_container.markdown(response_text)
+
+        # Append full response after display
+        st.session_state["chat_history"].append(
             {"role": "assistant", "content": assistant_response}
         )
 
-        # Display chatbot response
-        with st.chat_message("assistant"):
-            st.markdown(assistant_response)
+