BankBot-AI / utils.py
mohsin-devs's picture
Add OCR text extraction fallback with UI improvements
461d803
import re
import requests
from datetime import datetime
import uuid
import json
import os
import random
import hashlib
import streamlit as st
import PyPDF2
import io
from ollama_integration import (
get_ollama_response,
stream_ollama_response,
get_ai_response,
stream_ai_response,
get_active_backend,
is_banking_query
)
USER_FILE = "users.json"
SESSION_FILE = "session.json"
HISTORY_FILE = "chat_history.json"
INTENTS_FILE = os.path.join("data", "intents.json")
@st.cache_data
def load_intents():
if not os.path.exists(INTENTS_FILE):
return {"intents": []}
try:
with open(INTENTS_FILE, "r") as f:
return json.load(f)
except Exception as e:
print(f"Error loading intents: {e}")
return {"intents": []}
# Global intents data, initialized from cached function
intents_data = load_intents()
def persist_user(username, email, password):
users = get_persisted_users()
users[username] = {"email": email, "password": password}
with open(USER_FILE, "w") as f:
json.dump(users, f)
def get_persisted_users():
if not os.path.exists(USER_FILE):
return {}
try:
with open(USER_FILE, "r") as f:
return json.load(f)
except:
return {}
def save_active_session(username):
with open(SESSION_FILE, "w") as f:
json.dump({"username": username}, f)
def get_active_session():
if not os.path.exists(SESSION_FILE):
return None
try:
with open(SESSION_FILE, "r") as f:
data = json.load(f)
return data.get("username")
except:
return None
# ─── Password Security ────────────────────────────────────────────────────────
def hash_password(password):
"""Hashes a password using SHA-256."""
return hashlib.sha256(password.encode()).hexdigest()
def verify_password(stored_password, provided_password):
"""Verifies a password against its hash."""
return stored_password == hash_password(provided_password)
def migrate_plaintext_passwords():
"""Migrates any legacy plaintext passwords to SHA-256 hashes."""
users = get_persisted_users()
changed = False
for username in users:
password = users[username]["password"]
# Check if it looks like a SHA-256 hash (64 hex chars)
if not (len(password) == 64 and all(c in "0123456789abcdef" for c in password.lower())):
users[username]["password"] = hash_password(password)
changed = True
if changed:
with open(USER_FILE, "w") as f:
json.dump(users, f, indent=4)
# ─── User Management ──────────────────────────────────────────────────────────
def is_admin(username):
users = get_persisted_users()
return users.get(username, {}).get("is_admin", False)
def create_admin_account(password):
users = get_persisted_users()
users["admin"] = {
"email": "admin@centralbank.ai",
"password": hash_password(password),
"is_admin": True,
"created_at": get_timestamp(),
"balance": 1000000.0,
"transactions": [],
"language": "English"
}
with open(USER_FILE, "w") as f:
json.dump(users, f, indent=4)
def persist_user(username, email, password, is_admin=False):
users = get_persisted_users()
users[username] = {
"email": email,
"password": hash_password(password),
"is_admin": is_admin,
"created_at": get_timestamp(),
"balance": 1000.0, # Starting balance
"transactions": [],
"language": "English"
}
with open(USER_FILE, "w") as f:
json.dump(users, f, indent=4)
def get_user_data(username):
users = get_persisted_users()
return users.get(username, {})
def update_user_data(username, data):
users = get_persisted_users()
if username in users:
users[username].update(data)
with open(USER_FILE, "w") as f:
json.dump(users, f, indent=4)
return True
return False
# ─── Banking Simulation ───────────────────────────────────────────────────────
def get_balance(username):
return get_user_data(username).get("balance", 0.0)
def update_balance(username, amount):
user_data = get_user_data(username)
if user_data:
user_data["balance"] = amount
update_user_data(username, user_data)
return True
return False
def add_transaction(username, type, amount, category, details=""):
user_data = get_user_data(username)
if user_data:
transaction = {
"id": str(uuid.uuid4()),
"date": get_timestamp(),
"type": type,
"amount": amount,
"category": category,
"details": details
}
if "transactions" not in user_data:
user_data["transactions"] = []
user_data["transactions"].insert(0, transaction)
update_user_data(username, user_data)
return True
return False
def get_transactions(username):
return get_user_data(username).get("transactions", [])
def transfer_funds(sender, receiver_username, amount, category="Transfer", details=""):
users = get_persisted_users()
if receiver_username not in users:
return False, "Receiver not found"
sender_balance = get_balance(sender)
if sender_balance < amount:
return False, "Insufficient funds"
# Deduct from sender
update_balance(sender, sender_balance - amount)
add_transaction(sender, "debit", amount, category, f"To: {receiver_username}")
# Add to receiver
receiver_balance = get_balance(receiver_username)
update_balance(receiver_username, receiver_balance + amount)
add_transaction(receiver_username, "credit", amount, category, f"From: {sender}")
return True, "Transfer successful"
# ─── Fraud Detection ──────────────────────────────────────────────────────────
def check_fraud_alerts(username):
"""Analyzes transactions for suspicious activity."""
transactions = get_transactions(username)
alerts = []
# 1. High amount transfer
for txn in transactions:
if txn["type"] == "debit" and txn["amount"] >= 50000:
alerts.append({
"severity": "high",
"message": f"Large transaction of {format_currency(txn['amount'])} detected",
"timestamp": txn["date"],
"details": f"Category: {txn['category']}"
})
# 2. Rapid transactions (more than 3 in 1 hour - simplified check)
# This is a mock implementation
if len(transactions) >= 3:
alerts.append({
"severity": "medium",
"message": "Multiple transactions in a short period",
"timestamp": get_timestamp(),
"details": "Please verify if these were initiated by you"
})
return alerts
def get_fraud_alerts_summary(username):
alerts = check_fraud_alerts(username)
return {
"total": len(alerts),
"high": len([a for a in alerts if a["severity"] == "high"]),
"medium": len([a for a in alerts if a["severity"] == "medium"]),
"alerts": alerts
}
# ─── Data & File Utilities ────────────────────────────────────────────────────
def save_intents(data):
"""Saves updated intents to the JSON file."""
try:
os.makedirs(os.path.dirname(INTENTS_FILE), exist_ok=True)
with open(INTENTS_FILE, "w") as f:
json.dump(data, f, indent=4)
return True
except Exception as e:
print(f"Error saving intents: {e}")
return False
def extract_text_with_ocr(pdf_file):
"""Fallback OCR extraction for scanned or image-based PDFs."""
try:
import pytesseract
import cv2
import numpy as np
from pdf2image import convert_from_bytes
import os
import platform
if platform.system() == 'Windows':
# Hardcode path for local Windows testing
pytesseract.pytesseract.tesseract_cmd = r'C:\Program Files\Tesseract-OCR\tesseract.exe'
poppler_path = os.path.join(os.path.dirname(__file__), 'poppler-24.02.0', 'Library', 'bin')
else:
poppler_path = None
except ImportError as e:
raise Exception(f"OCR Python packages missing: {e}. Please install pdf2image, pytesseract, opencv-python-headless, numpy.")
try:
if hasattr(pdf_file, 'seek'):
pdf_file.seek(0)
pdf_bytes = pdf_file.read()
if platform.system() == 'Windows':
images = convert_from_bytes(pdf_bytes, poppler_path=poppler_path)
else:
images = convert_from_bytes(pdf_bytes)
text = ""
for img in images:
img_cv = cv2.cvtColor(np.array(img), cv2.COLOR_RGB2BGR)
gray = cv2.cvtColor(img_cv, cv2.COLOR_BGR2GRAY)
thresh = cv2.threshold(gray, 150, 255, cv2.THRESH_BINARY)[1]
page_text = pytesseract.image_to_string(thresh)
text += page_text + "\n"
text = text.replace('β‚Ή', 'Rs.')
text = re.sub(r'\n+', '\n', text)
extracted = text.strip()
if not extracted:
raise Exception("OCR completed but no text was found in the images.")
return extracted
except Exception as e:
raise Exception(f"OCR System dependencies missing or failed: {e}. Make sure Tesseract OCR and Poppler are installed on your OS and added to PATH.")
def extract_text_from_pdf(pdf_file):
"""Extracts text from an uploaded PDF file with OCR fallback. Returns (text, error)."""
try:
if hasattr(pdf_file, 'seek'):
pdf_file.seek(0)
reader = PyPDF2.PdfReader(pdf_file)
text = ""
for page in reader.pages:
page_text = page.extract_text()
if page_text:
text += page_text
extracted = text.strip()
if extracted:
return extracted, None
# Fallback to OCR if empty
return extract_text_with_ocr(pdf_file), None
except Exception as e:
try:
return extract_text_with_ocr(pdf_file), None
except Exception as ocr_error:
return None, str(ocr_error)
def clear_active_session():
if os.path.exists(SESSION_FILE):
os.remove(SESSION_FILE)
def validate_email(email):
pattern = r'^[a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\.[a-zA-Z]{2,}$'
return re.match(pattern, email) is not None
def validate_password_strength(password):
if len(password) < 8:
return False, "Password must be at least 8 characters long"
if not re.search(r'[A-Z]', password):
return False, "Password must contain at least one uppercase letter"
if not re.search(r'[a-z]', password):
return False, "Password must contain at least one lowercase letter"
if not re.search(r'\d', password):
return False, "Password must contain at least one number"
if not re.search(r'[!@#$%^&*(),.?":{}|<>]', password):
return False, "Password must contain at least one special character"
return True, "Password is strong"
def format_currency(amount):
return f"β‚Ή{amount:,.2f}"
def get_timestamp():
return datetime.now().strftime("%Y-%m-%d %H:%M:%S")
def generate_session_id():
return str(uuid.uuid4())
def get_chat_preview(messages, max_length=50):
if not messages:
return "Empty chat"
for msg in messages:
if msg["role"] == "user":
content = msg["content"]
if len(content) > max_length:
return content[:max_length] + "..."
return content
return "No user messages"
@st.cache_data(ttl=30)
def load_history_file():
if not os.path.exists(HISTORY_FILE):
return {}
try:
with open(HISTORY_FILE, "r") as f:
return json.load(f)
except:
return {}
def save_history_file(history):
with open(HISTORY_FILE, "w") as f:
json.dump(history, f, indent=4)
def get_all_chat_sessions(username):
history = load_history_file()
return history.get(username, [])
def save_chat_session(username, session_state, messages, session_id=None):
if not messages or len(messages) == 0:
return None
history = load_history_file()
user_sessions = history.get(username, [])
if session_id:
# Update existing session
found = False
for session in user_sessions:
if session["session_id"] == session_id:
session["messages"] = messages
session["preview"] = get_chat_preview(messages)
session["timestamp"] = get_timestamp()
found = True
break
# Also update in-memory session_state for immediate UI feedback
for session in session_state.chat_sessions:
if session["session_id"] == session_id:
session["messages"] = messages
session["preview"] = get_chat_preview(messages)
session["timestamp"] = get_timestamp()
break
else:
# Create new session
session_id = generate_session_id()
new_session = {
"session_id": session_id,
"timestamp": get_timestamp(),
"messages": messages,
"preview": get_chat_preview(messages)
}
user_sessions.insert(0, new_session)
if "chat_sessions" not in session_state:
session_state.chat_sessions = []
session_state.chat_sessions.insert(0, new_session)
history[username] = user_sessions
save_history_file(history)
return session_id
def load_chat_session(username, session_id):
user_sessions = get_all_chat_sessions(username)
for session in user_sessions:
if session["session_id"] == session_id:
return session["messages"]
return None
def delete_chat_session(username, session_state, session_id):
history = load_history_file()
user_sessions = history.get(username, [])
user_sessions = [s for s in user_sessions if s["session_id"] != session_id]
history[username] = user_sessions
save_history_file(history)
if "chat_sessions" in session_state:
session_state.chat_sessions = [s for s in session_state.chat_sessions if s["session_id"] != session_id]
return True
def clear_all_chat_history(username, session_state):
history = load_history_file()
history[username] = []
save_history_file(history)
session_state.chat_sessions = []
return True
@st.cache_data(ttl=10)
def check_ollama_connection():
from ollama_integration import check_ollama_connection as _check
return _check()
def get_faq_response(prompt, language="English"):
"""
Checks if the user's prompt matches any common frequently asked questions
using the structured intents.json data.
"""
prompt_lower = prompt.lower().strip()
if not intents_data or "intents" not in intents_data:
return None
# Iterate through intents to find a matching pattern
for intent in intents_data["intents"]:
for pattern in intent["patterns"]:
p_lower = pattern.lower()
# For short patterns (like 'hi'), use word boundary check
if len(p_lower) <= 3:
if re.search(rf"\b{re.escape(p_lower)}\b", prompt_lower):
return get_localized_response(intent, language)
# For longer patterns, substring match is usually fine and more flexible
elif p_lower in prompt_lower:
return get_localized_response(intent, language)
return None
def get_localized_response(intent, language):
"""Helper to pick a response in the requested language."""
if language == "Hindi":
responses = intent.get("responses_hi", intent.get("responses"))
elif language == "Marathi":
responses = intent.get("responses_mr", intent.get("responses"))
else:
responses = intent.get("responses")
return random.choice(responses)
def calculate_loan_eligibility(monthly_income, existing_emis, tenure_years):
"""
Calculates loan eligibility based on FOIR (Fixed Obligation to Income Ratio).
Standard FOIR is usually 50% for most banks.
"""
# Max EMI allowed (50% of income)
max_emi_allowed = monthly_income * 0.5
# Available EMI for new loan
available_emi = max_emi_allowed - existing_emis
if available_emi <= 0:
return 0, 0
# Reverse EMI calculation to find principal
# EMI = [P x R x (1+R)^N]/[(1+R)^N-1]
# P = EMI * [(1+R)^N-1] / [R * (1+R)^N]
rate_annual = 0.09 # Assume 9% interest for eligibility check
r = (rate_annual / 12)
n = tenure_years * 12
principal = available_emi * ((1 + r)**n - 1) / (r * (1 + r)**n)
return round(principal, 2), round(available_emi, 2)