Update app.py

app.py

@@ -1,63 +1,56 @@
-import streamlit as st
-from PIL import Image
-import easyocr
-import os
-import datetime
-
-# Initialize OCR reader
-reader = easyocr.Reader(['en', 'hi', 'te', 'ta', 'bn', 'ml', 'gu', 'mr'], gpu=False)
-
-# Create a directory to save uploads
-UPLOAD_DIR = "uploaded_data"
-os.makedirs(UPLOAD_DIR, exist_ok=True)
-
-st.set_page_config(page_title="Indian Landmark Mapper", layout="centered")
-
-st.title("📍 Indian Landmark Mapper")
-st.write("Help preserve India's heritage. Upload a local landmark photo and describe it in your native language.")
-
-# File uploader
-image_file = st.file_uploader("Upload a photo of the landmark", type=['jpg', 'jpeg', 'png'])
-
-# Text input
-description = st.text_area("Write a short description in your local language")
-
-# Optional: OCR toggle
-run_ocr = st.checkbox("Run OCR on uploaded image (optional)")
-
-# Optional: Location
-location = st.text_input("Enter location (optional: village, town, district)")
-
-# Submission button
-if st.button("Submit"):
-    if image_file is None or not description.strip():
-        st.warning("Please upload an image and enter a description.")
-    else:
-        # Save image
-        timestamp = datetime.datetime.now().strftime("%Y%m%d%H%M%S")
-        image_path = os.path.join(UPLOAD_DIR, f"{timestamp}_{image_file.name}")
-        with open(image_path, "wb") as f:
-            f.write(image_file.getbuffer())
-
-        # Save metadata
-        metadata = {
-            "description": description.strip(),
-            "location": location,
-            "image_filename": image_path,
-            "timestamp": timestamp
-        }
-
-        # In real deployment, replace this with saving to Hugging Face dataset or DB
-        st.success("✅ Submission received! Thank you for preserving your local heritage.")
-        st.json(metadata)
-
-# OCR Section
-if image_file and run_ocr:
-    st.subheader("🧠 OCR Result")
-    img = Image.open(image_file)
-    st.image(img, caption="Uploaded Image", use_column_width=True)
-
-    with st.spinner("Running OCR..."):
-        result = reader.readtext(np.array(img))
-        extracted_text = "\n".join([line[1] for line in result])
-        st.code(extracted_text)
+import streamlit as st
+from PIL import Image
+import pytesseract
+from transformers import RagRetriever, RagTokenForGeneration, RagTokenizer
+from datasets import load_dataset
+
+# --- App Title ---
+st.set_page_config(page_title="Landmark Mapper", layout="wide")
+st.title("🗺️ Landmark Mapper - Discover, Describe & Contribute")
+
+# --- Image Upload ---
+uploaded_file = st.file_uploader("Upload a landmark image", type=["jpg", "jpeg", "png"])
+
+# --- OCR + Description Input ---
+description = ""
+if uploaded_file:
+    image = Image.open(uploaded_file)
+    st.image(image, caption="Uploaded Image", use_column_width=True)
+
+    if st.checkbox("Run OCR to extract text from image"):
+        with st.spinner("Extracting text..."):
+            ocr_text = pytesseract.image_to_string(image)
+        st.text_area("Extracted Text", ocr_text, height=100)
+
+    description = st.text_area("Enter a description in your local language", height=150)
+
+# --- RAG Integration ---
+if st.button("Analyze with AI") and description:
+    with st.spinner("Running RAG model..."):
+        # Load RAG model components
+        tokenizer = RagTokenizer.from_pretrained("facebook/rag-token-base")
+        model = RagTokenForGeneration.from_pretrained("facebook/rag-token-base")
+        retriever = RagRetriever.from_pretrained("facebook/rag-token-base", index_name="legacy")
+
+        # Encode and retrieve
+        input_dict = tokenizer.prepare_seq2seq_batch(description, return_tensors="pt")
+        input_dict["input_ids"] = input_dict["input_ids"][:, :128]  # limit input length
+        input_dict["retrieval_kwargs"] = {"n_docs": 5}
+
+        generated = model.generate(
+            input_ids=input_dict["input_ids"],
+            context_input_ids=None,
+            context_attention_mask=None,
+            num_beams=2,
+            min_length=30,
+            max_length=128
+        )
+
+        output = tokenizer.batch_decode(generated, skip_special_tokens=True)
+        st.subheader("📘 AI-Enhanced Landmark Info")
+        st.write(output[0])
+
+# --- Corpus Contribution ---
+if description:
+    st.success("✅ Thank you! Your description is now part of the landmark language corpus.")
+    st.markdown("Help us map Indian culture, one landmark at a time.")