Add application file

app.py

@@ -245,58 +245,146 @@ class AadhaarBackData(BaseModel):
 
 # ── Endpoints ─────────────────────────────────────────────────────────────────
 
-@app.post("/extract-front", response_model=AadhaarFrontData)
+@app.post("/extract-front")
 async def extract_front(file: UploadFile = File(...)):
     """
-    Upload the FRONT side of an Aadhaar card image.
+    Upload the FRONT side of an Aadhaar card (image or PDF).
     Pipeline:
-      1. nemotron-ocr-v1  →  raw OCR text
-      2. nvidia-nemotron-nano-9b-v2  →  structured JSON
-    Returns: name, dob, gender, aadhaar_no
+      1. If PDF: extract images from PDF
+      2. For each image: nemotron-ocr-v1  →  raw OCR text
+      3. nvidia-nemotron-nano-9b-v2  →  structured JSON
+    Returns: extracted front data (for single image) or list (for PDF with multiple pages)
     """
-    ocr_text = await run_ocr(file)
-
-    if not ocr_text.strip():
-        raise HTTPException(status_code=422, detail="OCR produced no text. Check the image quality.")
-
-    parsed = call_llm(ocr_text, FRONT_SYSTEM_PROMPT)
-
-    raw_aadhaar = str(parsed.get("aadhaar_no", ""))
-    aadhaar_digits = re.sub(r"\D", "", raw_aadhaar)
-
-    return AadhaarFrontData(
-        name=str(parsed.get("name", "")).strip()[:100],
-        dob=str(parsed.get("dob", "")).strip()[:12],
-        gender=str(parsed.get("gender", "")).strip()[:20],
-        aadhaar_no=aadhaar_digits[:12],
-    )
+    # Check if it's a PDF
+    if file.filename.lower().endswith('.pdf'):
+        images = await extract_images_from_pdf(file)
+        if not images:
+            raise HTTPException(status_code=422, detail="No images found in PDF")
+        
+        results = []
+        for page_num, pil_image in enumerate(images):
+            page_result = {"page": page_num + 1}
+            try:
+                img_byte_arr = BytesIO()
+                pil_image.save(img_byte_arr, format='PNG')
+                img_byte_arr.seek(0)
+                image_b64 = base64.b64encode(img_byte_arr.getvalue()).decode()
+                
+                if len(image_b64) >= 2_600_000:
+                    page_result["error"] = "Image too large. Resize and try again."
+                    results.append(page_result)
+                    continue
+                
+                # Create a temporary upload file from the image
+                temp_file = UploadFile(filename=f"page_{page_num}.png", file=BytesIO(img_byte_arr.getvalue()))
+                
+                ocr_text = await run_ocr(temp_file)
+                if not ocr_text.strip():
+                    page_result["error"] = "OCR produced no text. Check the image quality."
+                    results.append(page_result)
+                    continue
+                
+                parsed = call_llm(ocr_text, FRONT_SYSTEM_PROMPT)
+                raw_aadhaar = str(parsed.get("aadhaar_no", ""))
+                aadhaar_digits = re.sub(r"\D", "", raw_aadhaar)
+                
+                page_result["name"] = str(parsed.get("name", "")).strip()[:100]
+                page_result["dob"] = str(parsed.get("dob", "")).strip()[:12]
+                page_result["gender"] = str(parsed.get("gender", "")).strip()[:20]
+                page_result["aadhaar_no"] = aadhaar_digits[:12]
+                results.append(page_result)
+            except Exception as e:
+                page_result["error"] = str(e)
+                results.append(page_result)
+        
+        return results
+    else:
+        # Handle single image
+        ocr_text = await run_ocr(file)
+        if not ocr_text.strip():
+            raise HTTPException(status_code=422, detail="OCR produced no text. Check the image quality.")
+        
+        parsed = call_llm(ocr_text, FRONT_SYSTEM_PROMPT)
+        raw_aadhaar = str(parsed.get("aadhaar_no", ""))
+        aadhaar_digits = re.sub(r"\D", "", raw_aadhaar)
+        
+        return AadhaarFrontData(
+            name=str(parsed.get("name", "")).strip()[:100],
+            dob=str(parsed.get("dob", "")).strip()[:12],
+            gender=str(parsed.get("gender", "")).strip()[:20],
+            aadhaar_no=aadhaar_digits[:12],
+        )
 
 
-@app.post("/extract-back", response_model=AadhaarBackData)
+@app.post("/extract-back")
 async def extract_back(file: UploadFile = File(...)):
     """
-    Upload the BACK side of an Aadhaar card image.
+    Upload the BACK side of an Aadhaar card (image or PDF).
     Pipeline:
-      1. nemotron-ocr-v1  →  raw OCR text
-      2. nvidia-nemotron-nano-9b-v2  →  structured JSON
-    Returns: address, village_city, state, pincode
+      1. If PDF: extract images from PDF
+      2. For each image: nemotron-ocr-v1  →  raw OCR text
+      3. nvidia-nemotron-nano-9b-v2  →  structured JSON
+    Returns: extracted back data (for single image) or list (for PDF with multiple pages)
     """
-    ocr_text = await run_ocr(file)
-
-    if not ocr_text.strip():
-        raise HTTPException(status_code=422, detail="OCR produced no text. Check the image quality.")
-
-    parsed = call_llm(ocr_text, BACK_SYSTEM_PROMPT)
-
-    raw_pin = str(parsed.get("pincode", ""))
-    pin_digits = re.sub(r"\D", "", raw_pin)[:6]
-
-    return AadhaarBackData(
-        address=str(parsed.get("address", "")).strip()[:200],
-        village_city=str(parsed.get("village_city", "")).strip()[:100],
-        state=str(parsed.get("state", "")).strip()[:60],
-        pincode=pin_digits,
-    )
+    # Check if it's a PDF
+    if file.filename.lower().endswith('.pdf'):
+        images = await extract_images_from_pdf(file)
+        if not images:
+            raise HTTPException(status_code=422, detail="No images found in PDF")
+        
+        results = []
+        for page_num, pil_image in enumerate(images):
+            page_result = {"page": page_num + 1}
+            try:
+                img_byte_arr = BytesIO()
+                pil_image.save(img_byte_arr, format='PNG')
+                img_byte_arr.seek(0)
+                image_b64 = base64.b64encode(img_byte_arr.getvalue()).decode()
+                
+                if len(image_b64) >= 2_600_000:
+                    page_result["error"] = "Image too large. Resize and try again."
+                    results.append(page_result)
+                    continue
+                
+                # Create a temporary upload file from the image
+                temp_file = UploadFile(filename=f"page_{page_num}.png", file=BytesIO(img_byte_arr.getvalue()))
+                
+                ocr_text = await run_ocr(temp_file)
+                if not ocr_text.strip():
+                    page_result["error"] = "OCR produced no text. Check the image quality."
+                    results.append(page_result)
+                    continue
+                
+                parsed = call_llm(ocr_text, BACK_SYSTEM_PROMPT)
+                raw_pin = str(parsed.get("pincode", ""))
+                pin_digits = re.sub(r"\D", "", raw_pin)[:6]
+                
+                page_result["address"] = str(parsed.get("address", "")).strip()[:200]
+                page_result["village_city"] = str(parsed.get("village_city", "")).strip()[:100]
+                page_result["state"] = str(parsed.get("state", "")).strip()[:60]
+                page_result["pincode"] = pin_digits
+                results.append(page_result)
+            except Exception as e:
+                page_result["error"] = str(e)
+                results.append(page_result)
+        
+        return results
+    else:
+        # Handle single image
+        ocr_text = await run_ocr(file)
+        if not ocr_text.strip():
+            raise HTTPException(status_code=422, detail="OCR produced no text. Check the image quality.")
+        
+        parsed = call_llm(ocr_text, BACK_SYSTEM_PROMPT)
+        raw_pin = str(parsed.get("pincode", ""))
+        pin_digits = re.sub(r"\D", "", raw_pin)[:6]
+        
+        return AadhaarBackData(
+            address=str(parsed.get("address", "")).strip()[:200],
+            village_city=str(parsed.get("village_city", "")).strip()[:100],
+            state=str(parsed.get("state", "")).strip()[:60],
+            pincode=pin_digits,
+        )
 
 
 @app.post("/extract-pdf")

index.html

@@ -253,11 +253,11 @@
         <div class="grid">
             <!-- Image Upload - Front -->
             <div class="card">
-                <h2>📸 Extract Front (Image)</h2>
+                <h2>📸 Extract Front (Image or PDF)</h2>
                 <form id="frontImageForm">
                     <div class="form-group">
-                        <label for="frontImageFile">Upload Front Side Image</label>
-                        <input type="file" id="frontImageFile" accept="image/*" required>
+                        <label for="frontImageFile">Upload Front Side Image or PDF</label>
+                        <input type="file" id="frontImageFile" accept="image/*,.pdf" required>
                     </div>
                     <button type="submit">Extract Front Data</button>
                     <div class="loading" id="frontImageLoading">
@@ -268,11 +268,11 @@
 
             <!-- Image Upload - Back -->
             <div class="card">
-                <h2>📸 Extract Back (Image)</h2>
+                <h2>📸 Extract Back (Image or PDF)</h2>
                 <form id="backImageForm">
                     <div class="form-group">
-                        <label for="backImageFile">Upload Back Side Image</label>
-                        <input type="file" id="backImageFile" accept="image/*" required>
+                        <label for="backImageFile">Upload Back Side Image or PDF</label>
+                        <input type="file" id="backImageFile" accept="image/*,.pdf" required>
                     </div>
                     <button type="submit">Extract Back Data</button>
                     <div class="loading" id="backImageLoading">