Update app.py

app.py

@@ -1,237 +1,59 @@
 import gradio as gr
-import torch
-from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
-from threading import Thread
-import os
-import pytesseract
-from pdf2image import convert_from_path
-from PIL import Image
-import trafilatura
-import datetime
-import json
-import random
-import re
-
-# ==========================================
-# 1. CONFIGURATION & MOCK DATABASE (SOP A.2)
-# ==========================================
-MODEL_ID = "Qwen/Qwen2.5-0.5B-Instruct"
-CONFIDENCE_THRESHOLD = 0.85
-ADMIN_EMAIL = "admin@company.com"
-
-# Simulated "Simple Table" for Vendor Lookups (SOP A.2.a)
-# In production, this would be a SQL DB or Zoho API Fetch
-VENDOR_DB = {
-    "AMAZON WEB SERVICES": {"id": "VEN-001", "status": "Active", "currency": "INR"},
-    "ZOHO CORPORATION": {"id": "VEN-002", "status": "Active", "currency": "INR"},
-    "MICROSOFT AZURE": {"id": "VEN-003", "status": "Active", "currency": "INR"}
-}
-
-# ==========================================
-# 2. SYSTEM UTILITIES (The "Hands")
-# ==========================================
-
-def mock_email_admin(subject, body):
-    """SOP C.3: Email Admin on Failure"""
-    log_entry = f"""
-    [MOCK EMAIL SERVER]
-    TO: {ADMIN_EMAIL}
-    SUBJECT: {subject}
-    BODY: {body}
-    ------------------------------------------------
-    """
-    print(log_entry)
-    return "📧 Alert Sent to Admin"
-
-def perform_ocr_with_confidence(file_obj):
-    """
-    Runs OCR and calculates a heuristic confidence score.
-    Since Tesseract raw confidence is tricky, we use text density/clarity as a proxy for this demo.
-    """
-    if file_obj is None: return None, 0.0, "No File"
+import requests
+
+# --- YOUR CREDENTIALS (HARDCODED FOR SPEED) ---
+CLIENT_ID = "1000.SIMKGAO5719K0TQ0QZQ31ZU57RLFNQ"
+CLIENT_SECRET = "60b329b4fe51930abee900cba6524ec7332cd67e06"
+# IMPORTANT: This must match EXACTLY what you set in Zoho Console
+# If you used "http://www.google.com" in console, keep it here.
+REDIRECT_URI = "http://www.google.com" 
+
+def get_zoho_token(auth_code, region):
+    """Exchanges the 10-min code for a permanent Refresh Token"""
+    
+    # Select Data Center (India accounts use .in, US/Global use .com)
+    domain = "zoho.in" if region == "India (.in)" else "zoho.com"
+    url = f"https://accounts.{domain}/oauth/v2/token"
+    
+    payload = {
+        "code": auth_code,
+        "client_id": CLIENT_ID,
+        "client_secret": CLIENT_SECRET,
+        "redirect_uri": REDIRECT_URI,
+        "grant_type": "authorization_code"
+    }
     
     try:
-        filename = os.path.basename(file_obj)
-        if filename.lower().endswith(".pdf"):
-            images = convert_from_path(file_obj, first_page=1, last_page=1)
-            image = images[0]
-        else:
-            image = Image.open(file_obj).convert("RGB")
-        
-        text = pytesseract.image_to_string(image)
+        print(f"Attempting connection to {url}...")
+        response = requests.post(url, data=payload)
         
-        # Heuristic Confidence Calculation (Demo Version)
-        # Real production uses bounding box confidence averages
-        word_count = len(text.split())
-        has_digits = bool(re.search(r'\d', text))
-        
-        if word_count > 10 and has_digits:
-            score = 0.92 # High confidence simulation
-        elif word_count > 5:
-            score = 0.70 # Low confidence
+        # Check for success
+        if response.status_code == 200 and "refresh_token" in response.text:
+            data = response.json()
+            return (
+                f"✅ SUCCESS! \n\nREFRESH TOKEN:\n{data['refresh_token']}\n\n"
+                f"ACCESS TOKEN:\n{data['access_token']}\n\n"
+                "👉 Save the 'Refresh Token' securely. It never expires."
+            )
         else:
-            score = 0.10 # Garbage
+            return f"❌ FAILED:\n{response.text}\n\nHint: Did you generate a NEW code? The old one expires in 10 mins."
             
-        return image, score, text
     except Exception as e:
-        return None, 0.0, str(e)
-
-def normalize_date(extracted_date):
-    """SOP B.1: Force Today's Date if missing."""
-    if extracted_date and "null" not in extracted_date.lower():
-        return extracted_date
-    return datetime.datetime.now().strftime("%Y-%m-%d") + " (Auto-Filled)"
+        return f"System Error: {str(e)}"
 
-def manage_vendor(extracted_vendor_name):
-    """SOP A.2: Lookup or Create Vendor."""
-    if not extracted_vendor_name or extracted_vendor_name == "null":
-        return "UNKNOWN_VENDOR", "Needs Review"
+# --- UI ---
+with gr.Blocks(title="Zoho Token Generator") as demo:
+    gr.Markdown("## 🔑 Zoho Token Generator")
+    gr.Markdown("1. Go to Zoho Console -> Self Client -> Generate Code.\n2. Paste the code below immediately.")
     
-    # Simple Normalization (Upper case match)
-    key = extracted_vendor_name.upper().strip()
-    
-    if key in VENDOR_DB:
-        return VENDOR_DB[key]["id"], "Active"
-    else:
-        # SOP A.2.b: Create new and flag
-        new_id = f"VEN-{random.randint(1000, 9999)}"
-        VENDOR_DB[key] = {"id": new_id, "status": "Pending_Review", "currency": "INR"}
-        return new_id, "Created (Pending Review)"
-
-# ==========================================
-# 3. AI BRAIN (Qwen Loading)
-# ==========================================
-print(f">>> SYSTEM: Initializing {MODEL_ID}...")
-try:
-    tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
-    model = AutoModelForCausalLM.from_pretrained(MODEL_ID, device_map="cpu", torch_dtype=torch.float32, low_cpu_mem_usage=True)
-    print(">>> SYSTEM: AI Online.")
-except Exception as e:
-    print(f"CRITICAL: {e}")
-    model = None
-
-# ==========================================
-# 4. PROMPTING STRATEGY (SOPs & Guardrails)
-# ==========================================
-SYSTEM_PROMPT = """
-<|im_start|>system
-You are an **Invoice Assistant** strictly following these SOPs:
-1. **Source of Truth:** The Invoice text is the absolute authority for amounts and dates.
-2. **Extraction:** Extract 'Vendor', 'Date', 'Total', and 'Tax'. If missing, output 'null'.
-3. **Currency:** Always assume INR (Indian Rupee) for now.
-4. **Output:** Only raw JSON or bullet points. No chit-chat.
-<|im_end|>
-"""
-
-def query_llm(prompt, history=[]):
-    if model is None: return "System Error: AI Offline"
-    
-    full_text = f"{SYSTEM_PROMPT}"
-    for u, a in history:
-        full_text += f"<|im_start|>user\n{u}<|im_end|>\n<|im_start|>assistant\n{a}<|im_end|>\n"
-    full_text += f"<|im_start|>user\n{prompt}<|im_end|>\n<|im_start|>assistant\n"
-    
-    inputs = tokenizer(full_text, return_tensors="pt")
-    streamer = TextIteratorStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
-    
-    thread = Thread(target=model.generate, kwargs=dict(inputs, streamer=streamer, max_new_tokens=256, temperature=0.3))
-    thread.start()
-    
-    response = ""
-    for t in streamer: response += t
-    return response
-
-# ==========================================
-# 5. THE ORCHESTRATOR (Main Logic Flow)
-# ==========================================
-def orchestrate_process(file_obj):
-    logs = "🔵 Process Started...\n"
-    
-    # Step 1: OCR
-    img, score, text = perform_ocr_with_confidence(file_obj)
-    logs += f"👁️ OCR Completed. Confidence Score: {score}\n"
-    
-    # Step 2: Confidence Check (SOP A.1)
-    workflow_status = "APPROVED"
-    if score < CONFIDENCE_THRESHOLD:
-        workflow_status = "NEEDS_REVIEW"
-        logs += f"⚠️ Low Confidence (<{CONFIDENCE_THRESHOLD}). Workflow flagged for Manual Review.\n"
-    
-    # Step 3: AI Extraction
-    extraction_prompt = f"""
-    Extract the following fields from this invoice text. Return valid JSON only.
-    Fields: vendor_name, invoice_date (YYYY-MM-DD), total_amount, line_items_summary.
-    
-    INVOICE TEXT:
-    {text[:2000]}
-    """
-    ai_json_str = query_llm(extraction_prompt)
-    
-    # Attempt to parse AI JSON (Simple fallback parsing)
-    try:
-        # Cleanup markdown code blocks if AI adds them
-        clean_json = ai_json_str.replace("```json", "").replace("```", "").strip()
-        data = json.loads(clean_json)
-    except:
-        data = {"vendor_name": "null", "invoice_date": "null", "total_amount": "0"}
-        logs += "⚠️ AI JSON Parsing Failed. Using raw fallbacks.\n"
-
-    # Step 4: Data Normalization (SOP B.1 & A.2)
-    final_date = normalize_date(data.get("invoice_date"))
-    vendor_id, vendor_status = manage_vendor(data.get("vendor_name"))
-    
-    if vendor_status == "Created (Pending Review)":
-        logs += f"🆕 New Vendor Detected: {data.get('vendor_name')}. Created ID: {vendor_id}\n"
-
-    # Step 5: Zoho Payload Construction (SOP A.3)
-    zoho_payload = {
-        "customer_id": vendor_id,
-        "date": final_date,
-        "line_items": [
-            {
-                "name": data.get("line_items_summary", "Services Rendered"),
-                "rate": data.get("total_amount", 0),
-                "quantity": 1
-            }
-        ],
-        "currency_id": "INR", # SOP A.5
-        "status": "draft" if workflow_status == "NEEDS_REVIEW" else "sent"
-    }
+    with gr.Row():
+        code_input = gr.Textbox(label="Paste New Authorization Code (1000.xxxx...)", placeholder="1000.726...")
+        region_input = gr.Radio(["India (.in)", "Global (.com)"], label="Zoho Account Region", value="India (.in)")
     
-    # Step 6: Error Handling Simulation (SOP C.3)
-    # We mock an API failure if the amount is huge just to show the logic
-    try:
-        total_val = float(str(data.get("total_amount", "0")).replace(",",""))
-        if total_val > 1000000:
-            raise Exception("Simulated Zoho API Timeout")
-    except Exception as e:
-        alert_msg = mock_email_admin("Zoho Sync Failed", str(e))
-        logs += f"❌ {alert_msg}\n"
-
-    return img, json.dumps(zoho_payload, indent=2), logs, workflow_status
-
-# ==========================================
-# 6. UI LAUNCHER
-# ==========================================
-with gr.Blocks(theme=gr.themes.Soft(), title="Zoho Invoice Orchestrator") as demo:
-    gr.Markdown("## 🧠 Zoho Invoice Orchestrator (SOP Compliant)")
+    btn = gr.Button("Generate Permanent Token", variant="primary")
+    output = gr.Textbox(label="Result (Copy this)", lines=10)
     
-    with gr.Row():
-        with gr.Column():
-            file_in = gr.File(label="Upload Invoice")
-            process_btn = gr.Button("🚀 Process Invoice", variant="primary")
-            img_out = gr.Image(label="OCR Preview", height=300)
-        
-        with gr.Column():
-            status_box = gr.Textbox(label="Workflow Status (SOP A.1)")
-            log_box = gr.Textbox(label="Orchestrator Logs", lines=6)
-            json_out = gr.JSON(label="Zoho V3 Payload")
-
-    process_btn.click(
-        fn=orchestrate_process,
-        inputs=[file_in],
-        outputs=[img_out, json_out, log_box, status_box]
-    )
+    btn.click(get_zoho_token, inputs=[code_input, region_input], outputs=output)
 
 if __name__ == "__main__":
     demo.launch()