Added Dynamic HTML form filling - Implemented batch upload and processing with debug output

app.py

@@ -1,5 +1,5 @@
 import gradio as gr
-from pipeline import extract_info_batch, extract_child_fee_info,extract_medical_info
+from pipeline import extract_info_batch, extract_child_fee_info, extract_medical_info, extract_medical_info_batch
 from PIL import Image
 
 
@@ -71,13 +71,12 @@ with gr.Blocks() as demo:
         with gr.Tab("Medical Reimbursement Form"):
             with gr.Row():
                 with gr.Column(scale=2):
-                    medical_img_input = gr.Image(
-                        type="pil",
-                        label="Image Upload",
-                        elem_id="upload-img",
-                        show_label=False,
-                        height=512,
-                        width=512
+                    medical_img_input = gr.File(
+                        label="Upload Medical Form Images (Multiple Allowed)",
+                        file_count="multiple",
+                        file_types=["image"],
+                        elem_id="medical-upload-img-batch",
+                        show_label=True
                     )
                 with gr.Column(scale=2):
                     with gr.Row():
@@ -88,12 +87,12 @@ with gr.Blocks() as demo:
                         med_designation = gr.Textbox(label="Designation")
                         med_ext_code = gr.Textbox(label="Extention No.")
                         med_emp_code = gr.Textbox(label="Employee Code")
-                    medical_upload_btn = gr.Button("Upload and Process")
-                    preview_medical_output = gr.File(label="Download Filled Form")
+                    medical_upload_btn = gr.Button("Upload and Process Batch")
+                    preview_medical_output = gr.File(label="Download Consolidated Form (HTML)")
 
                     
                     medical_upload_btn.click(
-                        fn=extract_medical_info,
+                        fn=extract_medical_info_batch,
                         inputs=[medical_img_input,med_emp_name,med_emp_code,med_department,med_designation,med_company_name,med_ext_code],
                         outputs=preview_medical_output
                     )

pipeline.py

@@ -9,6 +9,7 @@ from models import ReceiptData, ChildFeeForm
 from form_fill import fill_child_fee_pdf, fill_medical_pdf
 from fraud import process_receipt
 from datetime import datetime
+import html
 
 
 load_dotenv()
@@ -68,9 +69,9 @@ medical_form_system_prompt = (
     "    name: str #the patient name\n"
     "    relationship: # self, spouse, parent, child\n"
     "    category: # in-patient, out-patient, maternity(cesarean), maternity(normal)\n"
-    "    detail: # doctor's fee, diagnostic tests, medicines, other hospitalization\n"
-    "    bill_month: Optional[str] = None # Bill Month Field, if not directly stated, find the date and infer the month from that, if not found return null\n"
-    "    amount: float\n"
+    "    detail: # doctor's fee, diagnostic tests, medicines, other hospitalization - only chose from these options, infer from the image which one it is\n"
+    "    bill_month: Optional[str] = None # Bill Month Field, if not directly stated, find the date and infer the month from that, format should be month - year (mm/yy), if not found return null\n"
+    "    amount: float - try your best to extract the exact amount present in the image, sometimes there will be discounts applied, look for the total amount paid\n"
     "class Form(BaseModel):\n"
     "    claims: List[Item]\n"
     "    total: float\n"
@@ -80,6 +81,7 @@ medical_form_system_prompt = (
     "- Only return a valid JSON object matching the model above.\n"
     "- Do not add any explanation or extra text—only the JSON."
     "- Try your very best to extract this information as it is very important that you do so\n"
+    "- Only extract claim items that have an explicitly stated amount next to a discernible service or in an itemized list. Do not infer multiple items if only one amount is clearly listed as a charge.\n"
     "- If you are unable to extract information, return an empty json in the format requested above, never give a response other than a json"
 )
 
@@ -129,7 +131,6 @@ def extract_info(pil_img):
         data = json.loads(raw_output)
         # print(data)
         validated = ReceiptData(**data)
-        # json_block = json.dumps(validated.dict(), indent=2, ensure_ascii=False)
 
         validated_dict = validated.dict()  # This is a Python dict, perfect for fraud check
         print(validated_dict)
@@ -221,7 +222,7 @@ def extract_medical_info(pil_img, emp_name, emp_code, department, designation, c
             {"role": "system", "content": medical_form_system_prompt},
             {"role": "user",
              "content": [
-                {"type": "text", "text": "Here is a child fee bill image:"},
+                {"type": "text", "text": "Here is a medical form image:"},
                 {"type": "image_url", "image_url": {"url": "data:image/png;base64," + img_base64}}
              ]}
         ]
@@ -230,45 +231,223 @@ def extract_medical_info(pil_img, emp_name, emp_code, department, designation, c
     print(raw_output)
     try:
         if raw_output.startswith("```"):
-            raw_output = raw_output.strip("` \n")
+            raw_output = raw_output.strip("` \\n")
             if raw_output.startswith("json"):
                 raw_output = raw_output[4:].strip()
-        data = json.loads(raw_output)
-        print(data)
-        # Validate if needed:
-        # ChildFeeForm(**data)
-
-        claims = data.get("claims", [])
-        bill_month = ""
-        if claims and "bill_month" in claims[0]:
-            bill_month = claims[0]["bill_month"]
-
-        date = datetime.now().strftime("%d-%b-%Y")  # e.g., "10-Jun-2024"
-        total = data.get("total", 0)
-
-        print("bill month:",bill_month)
-
-        print("total:",total)
-        os.makedirs("outputs", exist_ok=True)
-        output_pdf_path = f"outputs/filled_medical_form_{uuid.uuid4().hex}.pdf"
+        data_from_llm = json.loads(raw_output)
+        print("Data from LLM:", data_from_llm)
+
+        # Extract bill_month from LLM data
+        claims_from_llm = data_from_llm.get("claims", [])
+        bill_month_from_llm = "" # Default to empty string
+        if claims_from_llm and isinstance(claims_from_llm, list) and len(claims_from_llm) > 0:
+            first_claim = claims_from_llm[0]
+            if isinstance(first_claim, dict) and "bill_month" in first_claim:
+                bill_month_from_llm = first_claim.get("bill_month", "")
+        print(f"Extracted billing month from LLM: '{bill_month_from_llm}'")
+
+        # Get total from LLM as well
+        total_from_llm = data_from_llm.get("total", 0) # Default to 0 if not found
+        print(f"Extracted total from LLM: {total_from_llm}")
+
+        form_header_data = {
+            "company": company,
+            "employee_name": emp_name,
+            "department": department,
+            "designation": designation,
+            "extension_no": extension_no,
+            "employee_code": emp_code,
+            "date": datetime.now().strftime("%Y-%m-%d"),
+            "billing_month": bill_month_from_llm,
+            "claims": claims_from_llm, # Pass the full claims array
+            "total_amount": total_from_llm # Pass the LLM's total (JS will also calculate)
+        }
+        json_data_for_script = json.dumps(form_header_data)
+
+        html_template_path = os.path.join("templates", "medical_form.html")
+        with open(html_template_path, "r", encoding="utf-8") as f:
+            html_content = f.read()
+
+        # Correctly formatted script to inject
+        script_to_inject = f'''
+<script>
+    document.addEventListener('DOMContentLoaded', function() {{ 
+        const dataToLoad = {json_data_for_script};
+        if (typeof populateMedicalForm === 'function') {{
+            populateMedicalForm(dataToLoad);
+        }} else {{
+            console.error('populateMedicalForm function not defined when trying to load data.');
+        }}
+    }});
+</script>
+</body>'''
+        # Ensure we are replacing the </body> tag correctly
+        if "</body>" in html_content:
+            html_content = html_content.replace("</body>", script_to_inject, 1)
+        else:
+            # Fallback if no </body> tag, append at the end (less ideal)
+            html_content += script_to_inject.replace("</body>","") # remove the body tag we added if appending
+
+
+        output_dir = "outputs"
+        os.makedirs(output_dir, exist_ok=True)
+        output_html_filename = f"filled_medical_form_{uuid.uuid4().hex}.html"
+        output_html_path = os.path.join(output_dir, output_html_filename)
+
+        with open(output_html_path, "w", encoding="utf-8") as f:
+            f.write(html_content)
+        
+        print(f"Populated HTML form saved to: {output_html_path}")
+        return output_html_path
 
-
-        filled_pdf_path = fill_medical_pdf(
-            template_pdf_path="Medical Reim. Form.pdf",
-            output_pdf_path=output_pdf_path,
-            company=company,
-            employee_name=emp_name,
-            employee_code=emp_code,
-            department=department,
-            designation=designation,
-            extension_no=extension_no,
-            billing_month=bill_month,
-            claims=claims,
-            date= date,
-            total=total
-        )
-
-        return filled_pdf_path # Return path to Gradio for download
     except Exception as e:
-        print("ERROR:", e)
-        return None  # or f"Error: {str(e)}"
+        print(f"ERROR in extract_medical_info: {e}")
+        raw_output_escaped = html.escape(raw_output) # Escape raw_output for safe HTML display
+        error_html_content = f"<html><body><h1>Error</h1><p>{html.escape(str(e))}</p><p>Raw LLM Output:</p><pre>{raw_output_escaped}</pre></body></html>"
+        output_dir = "outputs"
+        os.makedirs(output_dir, exist_ok=True)
+        error_html_filename = f"error_medical_form_{uuid.uuid4().hex}.html"
+        error_html_path = os.path.join(output_dir, error_html_filename)
+        with open(error_html_path, "w", encoding="utf-8") as f:
+            f.write(error_html_content)
+        return error_html_path
+
+
+def extract_medical_info_batch(image_file_list, emp_name, emp_code, department, designation, company, extension_no):
+    """
+    Processes a batch of medical form images, consolidates all claims, 
+    generates a single populated HTML file, and returns its path.
+    """
+    if not image_file_list:
+        # Return an error HTML or an empty/default HTML path if no files are provided
+        error_content = "<html><body><h1>No Images Provided</h1><p>Please upload at least one medical form image.</p></body></html>"
+        output_dir = "outputs"
+        os.makedirs(output_dir, exist_ok=True)
+        error_filename = f"error_no_medical_form_images_{uuid.uuid4().hex[:8]}.html"
+        error_path = os.path.join(output_dir, error_filename)
+        with open(error_path, "w", encoding="utf-8") as f:
+            f.write(error_content)
+        return error_path # Gradio expects a list, so we might need to adjust app.py output handling or return [error_path]
+
+    consolidated_claims = []
+    first_billing_month_found = "" # To store the billing month from the first processed image that has one
+    grand_total_from_llm = 0.0
+    processed_file_names = []
+
+    print(f"DEBUG: Starting batch processing for {len(image_file_list)} images.") # DEBUG
+
+    for i, image_file_path_obj in enumerate(image_file_list): # DEBUG: Added enumerate
+        image_name_for_log = image_file_path_obj.name if hasattr(image_file_path_obj, 'name') else str(image_file_path_obj)
+        processed_file_names.append(os.path.basename(image_name_for_log))
+        print(f"DEBUG: --- Iteration {i+1} for image: {image_name_for_log} ---") # DEBUG
+        try:
+            print(f"Processing medical form for consolidation: {image_name_for_log}")
+            pil_img = Image.open(image_file_path_obj.name if hasattr(image_file_path_obj, 'name') else image_file_path_obj)
+
+            processed_image = preprocess_image(pil_img)
+            img_bytes = pil_to_bytes(processed_image)
+            img_base64 = base64.b64encode(img_bytes.getvalue()).decode("utf-8")
+            
+            response = openai.chat.completions.create(
+                model="gpt-4o",
+                messages=[
+                    {"role": "system", "content": medical_form_system_prompt},
+                    {"role": "user",
+                     "content": [
+                        {"type": "text", "text": f"Extract claims from this medical form image ({image_name_for_log}):"},
+                        {"type": "image_url", "image_url": {"url": "data:image/png;base64," + img_base64}}
+                     ]}
+                ]
+            )
+            raw_output = response.choices[0].message.content
+            print(f"DEBUG: Raw LLM output for {image_name_for_log}:\n{raw_output}\n") # DEBUG
+
+            if raw_output.startswith("```"):
+                raw_output = raw_output.strip("` \\n")
+                if raw_output.startswith("json"):
+                    raw_output = raw_output[4:].strip()
+            data_from_llm = json.loads(raw_output)
+            print(f"DEBUG: Parsed LLM data for {image_name_for_log}:\n{json.dumps(data_from_llm, indent=2)}\n") # DEBUG
+            
+            current_claims = data_from_llm.get("claims", [])
+            print(f"DEBUG: Current claims extracted for {image_name_for_log}: {len(current_claims)} items") # DEBUG
+            # print(f"DEBUG: Current claims content for {image_name_for_log}: {json.dumps(current_claims, indent=2)}") # DEBUG - Can be very verbose
+            
+            if current_claims and isinstance(current_claims, list):
+                consolidated_claims.extend(current_claims)
+            print(f"DEBUG: Consolidated claims after {image_name_for_log}: {len(consolidated_claims)} items total") # DEBUG
+            # print(f"DEBUG: Consolidated claims content after {image_name_for_log}: {json.dumps(consolidated_claims, indent=2)}") # DEBUG - Can be very verbose
+            
+            # Get billing month from the first item of the current form's claims, if not already found
+            if not first_billing_month_found and current_claims and isinstance(current_claims, list) and len(current_claims) > 0:
+                first_claim_current_img = current_claims[0]
+                if isinstance(first_claim_current_img, dict) and "bill_month" in first_claim_current_img:
+                    first_billing_month_found = first_claim_current_img.get("bill_month", "")
+            
+            grand_total_from_llm += float(data_from_llm.get("total", 0) or 0) # Ensure float and handle None
+
+        except Exception as e:
+            print(f"ERROR processing medical form image '{image_name_for_log}' for consolidation: {e}")
+            # We can decide to stop or continue. For now, let's log and continue, 
+            # but not add its claims. The final HTML will be generated with claims from successful ones.
+            # Optionally, add an error marker to the final HTML or a separate error report.
+            # For simplicity here, we just skip this file's claims on error.
+
+    print(f"DEBUG: --- End of loop for consolidating claims ---") # DEBUG
+    print(f"DEBUG: Final consolidated_claims before HTML generation: {len(consolidated_claims)} items") # DEBUG
+    # print(f"DEBUG: Final consolidated_claims content: {json.dumps(consolidated_claims, indent=2)}") # DEBUG - Can be very verbose
+
+    # Now, prepare and save the single consolidated HTML file
+    form_header_data = {
+        "company": company,
+        "employee_name": emp_name,
+        "department": department,
+        "designation": designation,
+        "extension_no": extension_no,
+        "employee_code": emp_code,
+        "date": datetime.now().strftime("%Y-%m-%d"),
+        "billing_month": first_billing_month_found,
+        "claims": consolidated_claims, # All claims from all images
+        "total_amount": grand_total_from_llm # Sum of totals from LLM (JS will also recalc)
+    }
+    json_data_for_script = json.dumps(form_header_data)
+
+    html_template_path = os.path.join("templates", "medical_form.html")
+    with open(html_template_path, "r", encoding="utf-8") as f:
+        html_content_template = f.read()
+    
+    current_html_content = str(html_content_template)
+
+    script_to_inject = f'''
+<script>
+    document.addEventListener('DOMContentLoaded', function() {{ 
+        const dataToLoad = {json_data_for_script};
+        if (typeof populateMedicalForm === 'function') {{
+            populateMedicalForm(dataToLoad);
+        }} else {{
+            console.error('populateMedicalForm function not defined when trying to load data.');
+        }}
+    }});
+</script>
+</body>'''
+    if "</body>" in current_html_content:
+        current_html_content = current_html_content.replace("</body>", script_to_inject, 1)
+    else:
+        current_html_content += script_to_inject.replace("</body>","")
+
+    output_dir = "outputs"
+    os.makedirs(output_dir, exist_ok=True)
+    
+    # Create a filename for the consolidated report
+    # You could include employee name or a timestamp if these are consistent per batch.
+    consolidated_filename_भाग = "_-".join(filter(None, processed_file_names))
+    if not consolidated_filename_भाग:
+        consolidated_filename_भाग = "batch"
+    output_html_filename = f"consolidated_medical_form_{consolidated_filename_भाग[:50]}_{uuid.uuid4().hex[:8]}.html"
+    output_html_path = os.path.join(output_dir, output_html_filename)
+
+    with open(output_html_path, "w", encoding="utf-8") as f:
+        f.write(current_html_content)
+    
+    print(f"Consolidated HTML form saved to: {output_html_path}")
+    return output_html_path # Return path to the single consolidated HTML

templates/medical_form.html

@@ -218,8 +218,8 @@
                     <th class="table-claim-col">Detail</th>
                     <th class="table-claim-col">Amount</th>
                 </tr>
-                <tbody>
-                    {{claims_rows}}
+                <tbody id="claims_table_body">
+                    <!-- Claim rows will be dynamically inserted here by JavaScript -->
                 </tbody>
             </table>
             <table style="margin-top: 0;">
@@ -263,7 +263,104 @@
         </div>
     </div>
     <script>
-        // Auto sum the Amount columns
+        // Function to populate the main form fields from a JSON object
+        function populateMedicalForm(jsonData) {
+            if (!jsonData) {
+                console.error("No JSON data provided to populateMedicalForm");
+                return;
+            }
+
+            // Helper to safely set value
+            function setValue(name, value) {
+                const element = document.querySelector(`input[name="${name}"]`);
+                if (element) {
+                    element.value = value !== undefined && value !== null ? value : '';
+                } else {
+                    console.warn(`Element with name "${name}" not found.`);
+                }
+            }
+
+            setValue("company", jsonData.company);
+            setValue("name", jsonData.employee_name); // Assuming JSON key is employee_name for HTML field 'name'
+            setValue("department", jsonData.department);
+            setValue("designation", jsonData.designation);
+            setValue("extension", jsonData.extension_no); // Assuming JSON key is extension_no for HTML field 'extension'
+            setValue("employee_code", jsonData.employee_code);
+            setValue("date", jsonData.date);
+            // billing_month is now being sent, so we uncomment and use it.
+            setValue("billing_month", jsonData.billing_month);
+
+            // Populate the claims table
+            populateClaimsTable(jsonData.claims);
+            // Setup listeners for dynamic amount fields *after* table is populated
+            setupAmountListeners();
+        }
+
+        // Function to populate the claims table dynamically
+        function populateClaimsTable(claimsArray) {
+            const tableBody = document.getElementById('claims_table_body');
+            tableBody.innerHTML = ''; // Clear existing rows
+
+            if (!claimsArray || !Array.isArray(claimsArray)) {
+                console.warn("No claims data provided or data is not an array.");
+                // Add a default empty row if you always want at least one row, or leave empty
+                // const row = tableBody.insertRow();
+                // row.insertCell().innerHTML = '<input type="text" class="table-input" name="claim_name_1">';
+                // row.insertCell().innerHTML = '<input type="text" class="table-input" name="claim_relationship_1">';
+                // row.insertCell().innerHTML = '<input type="text" class="table-input" name="claim_category_1">';
+                // row.insertCell().innerHTML = '<input type="text" class="table-input" name="claim_detail_1">';
+                // row.insertCell().innerHTML = '<input type="number" class="table-input claim-amount" name="claim_amount_1" step="0.01" min="0">';
+                return;
+            }
+
+            claimsArray.forEach((claim, index) => {
+                const row = tableBody.insertRow();
+                const i = index + 1; // For 1-based indexing in names if needed by backend/PDF
+
+                row.insertCell().innerHTML = `<input type="text" class="table-input" name="claim_name_${i}" value="${claim.name || ''}">`;
+                row.insertCell().innerHTML = `<input type="text" class="table-input" name="claim_relationship_${i}" value="${claim.relationship || ''}">`;
+                row.insertCell().innerHTML = `<input type="text" class="table-input" name="claim_category_${i}" value="${claim.category || ''}">`;
+                row.insertCell().innerHTML = `<input type="text" class="table-input" name="claim_detail_${i}" value="${claim.detail || ''}">`;
+                row.insertCell().innerHTML = `<input type="number" class="table-input claim-amount" name="claim_amount_${i}" step="0.01" min="0" value="${claim.amount || ''}">`;
+            });
+             // If no claims, you might want to add a few empty rows for manual input
+            if (claimsArray.length === 0) {
+                for (let i = 1; i <= 4; i++) { // Add 4 empty rows for example
+                    const row = tableBody.insertRow();
+                    row.insertCell().innerHTML = `<input type="text" class="table-input" name="claim_name_${i}">`;
+                    row.insertCell().innerHTML = `<input type="text" class="table-input" name="claim_relationship_${i}">`;
+                    row.insertCell().innerHTML = `<input type="text" class="table-input" name="claim_category_${i}">`;
+                    row.insertCell().innerHTML = `<input type="text" class="table-input" name="claim_detail_${i}">`;
+                    row.insertCell().innerHTML = `<input type="number" class="table-input claim-amount" name="claim_amount_${i}" step="0.01" min="0">`;
+                }
+            }
+        }
+
+        // Function to setup amount listeners and calculate total
+        function setupAmountListeners() {
+            const amountFields = document.querySelectorAll('.claim-amount'); // Select all claim amount fields
+            const totalField = document.getElementById('total_amount');
+
+            amountFields.forEach(f => {
+                f.addEventListener('input', () => {
+                    let total = 0;
+                    amountFields.forEach(ff => {
+                        let val = parseFloat(ff.value);
+                        if (!isNaN(val)) total += val;
+                    });
+                    totalField.value = total > 0 ? total.toFixed(2) : '';
+                });
+                // Trigger input event once to calculate initial total if fields are pre-filled
+                f.dispatchEvent(new Event('input'));
+            });
+             // If there are no amount fields initially, ensure total is 0 or empty
+            if (amountFields.length === 0) {
+                totalField.value = '';
+            }
+        }
+
+        // Auto sum the Amount columns for claims (OLD STATIC SCRIPT - REMOVE OR COMMENT OUT)
+        /*
         const amountFields = [
             document.querySelector('input[name="claim_amount1"]'),
             document.querySelector('input[name="claim_amount2"]'),
@@ -272,15 +369,20 @@
         ];
         const totalField = document.getElementById('total_amount');
         amountFields.forEach(f => {
-            f.addEventListener('input', () => {
-                let total = 0;
-                amountFields.forEach(ff => {
-                    let val = parseFloat(ff.value);
-                    if (!isNaN(val)) total += val;
+            if(f) { // Add a check if the field exists
+                f.addEventListener('input', () => {
+                    let total = 0;
+                    amountFields.forEach(ff => {
+                        if(ff) { // Add a check if the field exists
+                            let val = parseFloat(ff.value);
+                            if (!isNaN(val)) total += val;
+                        }
+                    });
+                    totalField.value = total > 0 ? total.toFixed(2) : '';
                 });
-                totalField.value = total > 0 ? total.toFixed(2) : '';
-            });
+            }
         });
+        */
     </script>
 </body>
 </html>