import streamlit as st from streamlit_chat import message as st_message from streamlit_option_menu import option_menu import os import plotly.express as px from io import StringIO from langchain.schema import HumanMessage, SystemMessage, AIMessage from langchain.chat_models import AzureChatOpenAI, ChatOpenAI from langchain.memory import ConversationBufferWindowMemory from langchain.prompts import PromptTemplate import warnings import time from sqlalchemy import create_engine, Column, Integer, String, Text, Table, MetaData from sqlalchemy.orm import sessionmaker import matplotlib.pyplot as plt from langchain_groq import ChatGroq import pandas as pd import numpy as np from dotenv import load_dotenv import re warnings.filterwarnings("ignore", category=DeprecationWarning) load_dotenv() os.environ['GROQ_API_KEY'] = os.getenv("GROQ_API_KEY") llm = ChatGroq(model="llama-3.1-70b-versatile") # Streamlit page configuration st.set_page_config( page_title="TraffiTrack", page_icon="", layout="wide", initial_sidebar_state="expanded", ) # Initialize session state for messages and banned users if 'messages' not in st.session_state: st.session_state.messages = [{"message": "Hi! How can I assist you today?", "is_user": False}] if 'banned_users' not in st.session_state: st.session_state.banned_users = [] if 'flowmessages' not in st.session_state: st.session_state.flowmessages = [] # Function to handle registration def registration(): st.title("User Registration") # Ensure session state is initialized if "user_data" not in st.session_state: st.session_state.user_data = [] name = st.text_input("Enter your name") phone_number = st.text_input("Enter your phone number") if st.button("Register"): if name and phone_number: # Append user data to session state as a dictionary st.session_state.user_data.append({"name": name, "phone_number": phone_number}) st.success("Registration successful!") else: st.warning("Please fill in all fields.") # Function to simulate drug tracking data def generate_sample_data(): data = { "Drug Name": ["MDMA", "LSD", "Mephedrone", "Cocaine", "Heroin"], "Detected Instances": [10, 15, 7, 12, 5], "Flagged Users": [5, 10, 4, 7, 3], "IP Addresses": [3, 8, 2, 6, 2] } return pd.DataFrame(data) # Function to check for drug-related content and extract info def check_for_drug_content(input_text): drug_keywords = ["MDMA", "LSD", "Mephedrone", "Cocaine", "Heroin"] pattern = r'(\+?\d{1,3}[-. ]?)?\(?\d{1,4}?\)?[-. ]?\d{1,4}[-. ]?\d{1,4}[-. ]?\d{1,9}' # Regex for phone numbers ip_pattern = r'(?:(?:25[0-5]|2[0-4][0-9]|[01]?[0-9][0-9]?)\.){3}(?:25[0-5]|2[0-4][0-9]|[01]?[0-9][0-9]?)' # Regex for IP addresses found_drugs = [keyword for keyword in drug_keywords if keyword.lower() in input_text.lower()] phone_numbers = re.findall(pattern, input_text) ip_addresses = re.findall(ip_pattern, input_text) return found_drugs, phone_numbers, ip_addresses # Sidebar with options selected = option_menu( "Main Menu", ["Home", "Registration","Chat"], icons=['house', 'person','chat-dots'], menu_icon="cast", default_index=0, orientation="horizontal", styles={ "container": {"padding": "5px", "background-color": "#DEF9C4"}, "icon": {"color": "#468585", "font-size": "25px"}, "nav-link": { "font-size": "16px", "text-align": "left", "margin": "0px", "--hover-color": "#9CDBA6" }, "nav-link-selected": {"background-color": "#50B498"}, } ) # Function to get a response from the chat model def get_chatmodel_response(user_message): # Ensure user_message is a string if "user_data" in st.session_state and st.session_state.user_data: user_info = st.session_state.user_data[-1] # Get the most recent registered user name = user_info.get("name", "Unknown") phone_number = user_info.get("phone_number", "Unknown") else: name = "Unknown" phone_number = "Unknown" #better the prompt more formatting #chat types -> human message #response -> no drug message detected #from registration #remove debugging PROMPT_TEMPLATE = """ You are an expert assistant specializing in detecting drug-related messages for monitoring illegal drug activities. Your role is to analyze user messages carefully to identify mentions of illegal substances or any activity related to drug trafficking, sales, or use. Your task is not just to detect specific drug names but to understand the context of the conversation, even if indirect or slang terms are used. Key substances and related slang to focus on: - MDMA (Molly, Mandy) - LSD (Acid, Tabs) - Mephedrone (Meow Meow) - Cocaine (Coke, Snow, Blow, White, Charlie) - Heroin (Smack, H, Brown) - Marijuana (Weed, Pot, Mary Jane, Ganja, Bud) - Ecstasy (X, XTC, E) - Crack (Rock, Freebase) Common phrases to consider: - “Looking for Molly” (MDMA reference) - “I need some snow” (Cocaine reference) - “Got any tabs?” (LSD reference) - “Can you hook me up with some blow?” (Cocaine reference) - “I don’t want to mess with Charlie” (Cocaine reference in a negative context) Instructions: 1. If you detect any mention of the above substances, slang, or any activities related to drug sales, trafficking, or use, respond with a simple confirmation of detection, including the drug name or slang term detected. 2. Be aware of **negations** or sentences where the user may **refuse** the drugs (e.g., "I don’t want cocaine"). Do not flag these as positive detections. 3. **Do not** include the original user message in your response. 4. Ensure the format includes the following fields: sender's name, phone number, and the detected drug word/slang. However, if there is **no drug-related content**, respond with only the message "drug word detected: none." Below is the required format for your response: If a drug word is detected: {name} {phone_number} drug word detected: {{drug_word_detected}} If no drug word is detected: drug word detected: none Below is the user message User message: {user_message} Response: """ memory = ConversationBufferWindowMemory(k=5, return_messages=True) user_message = str(user_message) # Use the parameter user_message to format the prompt formatted_prompt = PROMPT_TEMPLATE.format( user_message=user_message, name=name, phone_number=phone_number ) # Add the formatted prompt to the conversation history st.session_state.flowmessages.append(HumanMessage(content=user_message)) # Generate a response from the model response = llm([SystemMessage(content=formatted_prompt)]) # Ensure the response.content is a string response_content = str(response.content) # Add the AI response to the conversation history st.session_state.flowmessages.append(AIMessage(content=response_content)) # Save the conversation context memory.save_context({"input": user_message}, {"output": response_content}) return response_content # User input for query # Button to send the message # if st.button("Send"): # if user_input: # response = get_chatmodel_response(user_input) # st.session_state.messages.append({"message": response, "is_user": False}) # st.experimental_rerun() # else: # st.warning("Please enter a message.") # Display the conversation history if "flowmessages" in st.session_state: st.subheader("Chat") for message in st.session_state.flowmessages: if isinstance(message, HumanMessage): st_message(message.content, is_user=True) elif isinstance(message, AIMessage): st_message(message.content, is_user=False) def display_home_info(): # Set background color st.markdown( """ """, unsafe_allow_html=True ) # Title with emoji st.title("🏠 Welcome to the Drug-Related Content Detector") # Section for description st.markdown( """

Our software solution helps identify drug-related content across multiple platforms.

""", unsafe_allow_html=True ) # Features list st.write("### Features include:") st.markdown( """ """, unsafe_allow_html=True ) if selected == "Registration": registration() elif selected == "Home": display_home_info() elif selected == "Chat": def traffitrack_chatbot(): st.title('TraffiTrack 💬') # Dropdown to select platform platform = st.selectbox( "Choose a platform", ["Live 💁‍♀️", "WhatsApp 📱", "Instagram 📸", "Telegram ✉️"], index=0 ) if platform == "Telegram ✉️": # Hardcoded CSV content csv_content = """sender_name,sender_id,phone_number,message_text Shruti,1580593004,917304814120,But I would prefer blowing a bag of Charlie Shruti,1580593004,917304814120,I want to eat ice cream i am bored Shruti,1580593004,917304814120,He’s heavily into smack Shruti,1580593004,917304814120,There was a bag of snow in the car Shruti,1580593004,917304814120,Did you bring the Mary Jane for the party tonight? Shruti,1580593004,917304814120,Mary Jane Ritika,1065437474,918828000465,I WANT A BAG OF CHARLIE Ritika,1065437474,918828000465,Okayy Preeyaj,6649015430,,Haa bhej cocain thoda Ritika,1065437474,918828000465,Maal chahiye? Preeyaj,6649015430,,Llm Ritika,1065437474,918828000465,Kya kar rahe ho? Ritika,1065437474,918828000465,Hey""" # Read the CSV content into a DataFrame messages_df = pd.read_csv(StringIO(csv_content)) # Reverse the DataFrame to display messages from first to last for idx, row in messages_df[::-1].iterrows(): # Reverse the DataFrame here sender_name = row['sender_name'] message_text = row['message_text'] # Display each message with its corresponding sender name st_message(f"{sender_name}: {message_text}", is_user=False, key=f"telegram_message_{idx}") if st.button("Analyze 🚨"): # Initialize count and list for drug-related messages drug_count = 0 # Initialize drug_count here drug_messages = [] user_data = {} # Initialize user data dictionary # Analyze each message for drug-related content for idx, row in messages_df.iterrows(): message_text = row['message_text'] sender_name = row['sender_name'] sender_id = row['sender_id'] phone_number = row['phone_number'] # Get response from the chat model response_content = get_chatmodel_response(message_text) # Check for drug word detected in the response if "drug word detected" in response_content and "none" not in response_content: drug_word = response_content.split("drug word detected: ")[1].strip() drug_count += 1 drug_messages.append({ "sender_name": sender_name, "sender_id": sender_id, "phone_number": phone_number, "message_text": message_text, "drug_word": drug_word }) # Aggregate data by user if sender_name not in user_data: user_data[sender_name] = { "phone_number": phone_number, "message_count": 0, "drug_words": [] } user_data[sender_name]["message_count"] += 1 user_data[sender_name]["drug_words"].append(drug_word) # Display statistics st.subheader("Analysis Results 📊") st.write(f"Total drug-related messages detected: {drug_count}") if drug_count > 0: # st.write("Details of detected messages:") # for message in drug_messages: # st.markdown(f"**Phone Number**: {message['phone_number']} \ # **Sender ID**: {message['sender_id']} \ # **Message**: {message['message_text']} \ # **Drug Detected**: {message['drug_word']}") # Prepare data for visualization user_names = list(user_data.keys()) message_counts = [data["message_count"] for data in user_data.values()] phone_numbers = [data["phone_number"] for data in user_data.values()] # 1. Bar chart: Messages per user st.markdown("### Number of Messages per User 📊") fig = px.bar( x=user_names, y=message_counts, labels={'x': 'User Name', 'y': 'Message Count'}, title="Messages Detected per User" ) st.plotly_chart(fig) # 2. Pie chart: Distribution of drug-related messages st.markdown("### Drug Distribution Among Users 🍰") drugs_detected = [drug for user in user_data.values() for drug in user["drug_words"]] fig = px.pie( names=drugs_detected, title="Distribution of Detected Drugs" ) st.plotly_chart(fig) # 3. Horizontal bar chart: Number of drug-related messages per user st.markdown("### Drug-related Messages per User 📊") fig = px.bar( y=user_names, x=message_counts, orientation='h', labels={'y': 'User Name', 'x': 'Drug-related Messages Count'}, title="Drug-related Messages per User" ) st.plotly_chart(fig) # 4. Display user details in a table st.markdown("### User Details Table 📋") user_df = pd.DataFrame({ "User Name": user_names, "Phone Number": phone_numbers, "Message_id" : sender_id, "Messages Detected": message_counts }) st.dataframe(user_df) # Optionally: Link to the statistics page st.markdown("[View Statistics Page](#)") else: st.write("No drug-related messages detected.") else: # Display chat messages for other platforms with unique keys for idx, msg in enumerate(st.session_state.messages): st_message(msg["message"], is_user=msg["is_user"], key=f"message_{idx}") # Input for user query input_text = st.text_input("Enter your text", key="user_input") if st.button("Send"): if input_text: # Append the user's message to session state st.session_state.messages.append({"message": input_text, "is_user": True}) # Get the response from the model response = get_chatmodel_response(input_text) # Append the response from the model st.session_state.messages.append({"message": response, "is_user": False}) # Rerun to refresh the UI with new messages st.experimental_rerun() else: st.warning("Please enter a message.") # Call the chatbot function traffitrack_chatbot() # elif selected == "Statistics": # st.title('Drug Trafficking Statistics 📊') # # Generate sample data # data = generate_sample_data() # # Display data # st.subheader("Overview of Detected Drugs") # st.dataframe(data) # # Plotting the data # st.subheader("Detected Instances of Drugs") # fig, ax = plt.subplots(figsize=(8, 5)) # ax.bar(data["Drug Name"], data["Detected Instances"], color="#50B498") # plt.title("Detected Instances of Drugs") # plt.xlabel("Drug Name") # plt.ylabel("Detected Instances") # st.pyplot(fig) # # Plotting flagged users # st.subheader("Flagged Users") # fig, ax = plt.subplots(figsize=(8, 5)) # ax.bar(data["Drug Name"], data["Flagged Users"], color="#468585") # plt.title("Flagged Users") # plt.xlabel("Drug Name") # plt.ylabel("Flagged Users") # st.pyplot(fig) # # Plotting IP addresses # st.subheader("Detected IP Addresses") # fig, ax = plt.subplots(figsize=(8, 5)) # ax.bar(data["Drug Name"], data["IP Addresses"], color="#9CDBA6") # plt.title("Detected IP Addresses") # plt.xlabel("Drug Name") # plt.ylabel("Detected IP Addresses") # st.pyplot(fig) # Custom CSS for a better user interface st.markdown(f""" """, unsafe_allow_html=True)