ApnaLawyer / app.py
adarsh-maurya's picture
Update app.py
6f7abd6 verified
# Standard libraries
import os
import re
import io
import json
import time
import base64
import hashlib
import tempfile
from datetime import datetime, timedelta
# Streamlit & UI components
import streamlit as st
import streamlit.components.v1 as components
# Firebase (Realtime + Pyrebase4)
import pyrebase
import firebase_admin
from firebase_admin import credentials, initialize_app, db as firebase_db
# Langchain core
from langchain.prompts import PromptTemplate, ChatPromptTemplate
from langchain.memory import ConversationBufferWindowMemory
from langchain.chains import ConversationalRetrievalChain
from langchain_community.vectorstores import FAISS
from langchain_community.embeddings import HuggingFaceEmbeddings
from langchain_core.output_parsers import StrOutputParser
from langchain_core.runnables import RunnablePassthrough
from langchain_together import Together
# NLP & translation
from transformers import pipeline
from deep_translator import GoogleTranslator
from langdetect import detect
import textwrap
# Audio processing
import soundfile as sf
import numpy as np
import wave
import sounddevice as sd
# Utility
import requests
from dotenv import load_dotenv
from dateutil.parser import parse
# Local modules
from footer import footer
# ----------------- Streamlit Config -------------------
st.set_page_config(page_title="ApnaLawyer", layout="centered")
# ----------------- Load Environment Variables -------------------
load_dotenv()
SPEECHMATICS_API_KEY = os.getenv('SPEECHMATICS_API_KEY')
TOGETHER_API_KEY = os.getenv('TOGETHER_API_KEY')
if not TOGETHER_API_KEY:
st.error("Please set TOGETHER_API_KEY environment variable")
st.stop()
# ----------------- LLM Init -------------------
llm = Together(
model="mistralai/Mixtral-8x7B-Instruct-v0.1",
temperature=0.7,
max_tokens=1024,
together_api_key=TOGETHER_API_KEY
)
# ----------------- Firebase Init (Pyrebase) -------------------
firebase_config = {
"apiKey": os.environ["FIREBASE_API_KEY"],
"authDomain": os.environ["FIREBASE_AUTH_DOMAIN"],
"databaseURL": os.environ["FIREBASE_DATABASE_URL"],
"projectId": os.environ["FIREBASE_PROJECT_ID"],
"storageBucket": os.environ["FIREBASE_STORAGE_BUCKET"],
"messagingSenderId": os.environ["FIREBASE_MESSAGING_SENDER_ID"],
"appId": os.environ["FIREBASE_APP_ID"]
}
firebase = pyrebase.initialize_app(firebase_config)
auth = firebase.auth()
# ----------------- Firebase Admin SDK Init -------------------
def initialize_firebase():
try:
if not firebase_admin._apps:
creds_dict = json.loads(os.environ["FIREBASE_CREDS_JSON"])
cred = credentials.Certificate(creds_dict)
firebase_app = initialize_app(cred, {
'databaseURL': os.environ["FIREBASE_DB_URL"]
})
except Exception as e:
st.error(f"Firebase initialization error: {str(e)}")
initialize_firebase()
# ----------------- UI -------------------
col1, col2, col3 = st.columns([1, 30, 1])
with col2:
st.image("images/banner.jpg", use_container_width=True)
st.markdown("""
<style>
#MainMenu {visibility: hidden;}
footer {visibility: hidden;}
</style>
""", unsafe_allow_html=True)
# ----------------- Translation Logic -------------------
supported_languages = {
"hindi": "hi",
"english": "en",
"hinglish": "hi"
}
devanagari_regex = re.compile(r'[\u0900-\u097F]+')
def detect_target_language(prompt):
prompt_lower = prompt.lower().strip()
if re.search(r'[\u0C80-\u0CFF]', prompt): # Kannada block
return "hindi"
for lang_name, lang_code in supported_languages.items():
if f"in {lang_name}" in prompt_lower:
return lang_name
if devanagari_regex.search(prompt):
return "hindi"
if re.search(r'\bdhara\b|\bkanoon\b|\bnyay\b', prompt_lower) and detect(prompt) == 'en':
return "hinglish"
try:
detected = detect(prompt)
for name, code in supported_languages.items():
if code == detected:
return name
except:
pass
return "english"
def translate_text(text, target_language):
try:
return GoogleTranslator(source='auto', target=target_language).translate(text)
except Exception as e:
return f"⚠ Translation failed: {str(e)}"
# ----------------- Login/Signup Interface -------------------
import json
import requests
from streamlit.components.v1 import html
# Custom CSS for the enhanced UI
def inject_custom_css():
st.markdown("""
<style>
/* Main container styles */
.stApp {
background: radial-gradient(circle at top left, #0f0f0f, #050505);
color: white;
}
/* Sidebar styles */
[data-testid="stSidebar"] {
background: rgba(255, 255, 255, 0.05) !important;
border: 1px solid rgba(255, 255, 255, 0.1) !important;
backdrop-filter: blur(12px) !important;
box-shadow: 0 0 30px rgba(0, 255, 255, 0.2) !important;
border-radius: 20px !important;
padding: 30px !important;
margin: 20px !important;
animation: fadeIn 1s ease !important;
}
@keyframes fadeIn {
from { opacity: 0; transform: translateY(20px); }
to { opacity: 1; transform: translateY(0); }
}
/* Input field styles */
.stTextInput>div>div>input, .stPassword>div>div>input {
background: rgba(255, 255, 255, 0.1) !important;
color: white !important;
border: 1px solid transparent !important;
border-radius: 10px !important;
padding: 12px 20px !important;
transition: all 0.3s ease !important;
}
.stTextInput>div>div>input:focus, .stPassword>div>div>input:focus {
border-color: #00ffff !important;
background: rgba(255, 255, 255, 0.15) !important;
box-shadow: 0 0 10px rgba(0, 255, 255, 0.2) !important;
outline: none !important;
}
/* Button styles */
.stButton>button {
width: 100% !important;
padding: 12px 20px !important;
border-radius: 10px !important;
background: linear-gradient(45deg, #00ffff, #007fff) !important;
color: black !important;
font-weight: 600 !important;
border: none !important;
transition: all 0.3s ease !important;
box-shadow: 0 4px 15px rgba(0, 255, 255, 0.3) !important;
}
.stButton>button:hover {
transform: translateY(-2px) !important;
box-shadow: 0 6px 20px rgba(0, 255, 255, 0.4) !important;
}
/* Radio button styles */
.stRadio>div {
flex-direction: column !important;
gap: 10px !important;
}
.stRadio>div>label {
color: white !important;
font-weight: 500 !important;
padding: 8px 12px !important;
border-radius: 8px !important;
background: rgba(255, 255, 255, 0.05) !important;
transition: all 0.3s ease !important;
}
.stRadio>div>label:hover {
background: rgba(255, 255, 255, 0.1) !important;
}
.stRadio>div>label[data-baseweb="radio"]>div:first-child {
border-color: #00ffff !important;
}
/* Link styles */
a {
color: #00ffff !important;
text-decoration: none !important;
transition: all 0.3s ease !important;
}
a:hover {
text-shadow: 0 0 10px rgba(0, 255, 255, 0.5) !important;
}
/* Error message styles */
.stAlert {
border-radius: 10px !important;
background: rgba(255, 0, 0, 0.1) !important;
border: 1px solid rgba(255, 0, 0, 0.2) !important;
}
/* Success message styles */
.stSuccess {
border-radius: 10px !important;
background: rgba(0, 255, 0, 0.1) !important;
border: 1px solid rgba(0, 255, 0, 0.2) !important;
}
/* Audio recording animation */
@keyframes pulse {
0% { transform: scale(1); }
50% { transform: scale(1.1); }
100% { transform: scale(1); }
}
.recording-active {
animation: pulse 1.5s infinite;
color: #ff4b4b !important;
}
</style>
""", unsafe_allow_html=True)
def store_user_data(user_id, name, email):
try:
ref = firebase_db.reference(f'users/{user_id}')
ref.set({
'name': name,
'email': email,
'created_at': time.time()
})
return True
except Exception as e:
st.error(f"Error storing user data: {str(e)}")
return False
# Add these Firebase functions right after your existing Firebase functions
def save_chat_to_history(user_id, chat_title, messages):
try:
ref = firebase_db.reference(f'users/{user_id}/chats')
new_chat_ref = ref.push()
new_chat_ref.set({
'title': chat_title,
'messages': messages,
'timestamp': time.time(),
'last_updated': time.time()
})
return new_chat_ref.key
except Exception as e:
st.error(f"Error saving chat history: {str(e)}")
return None
def update_chat_history(user_id, chat_id, messages):
try:
ref = firebase_db.reference(f'users/{user_id}/chats/{chat_id}')
ref.update({
'messages': messages,
'last_updated': time.time()
})
return True
except Exception as e:
st.error(f"Error updating chat history: {str(e)}")
return False
def get_chat_history(user_id):
try:
ref = firebase_db.reference(f'users/{user_id}/chats')
chats = ref.get()
if chats:
return sorted(
[(chat_id, chat_data) for chat_id, chat_data in chats.items()],
key=lambda x: x[1]['last_updated'],
reverse=True
)
return []
except Exception as e:
st.error(f"Error fetching chat history: {str(e)}")
return []
def delete_chat_history(user_id, chat_id):
try:
ref = firebase_db.reference(f'users/{user_id}/chats/{chat_id}')
ref.delete()
return True
except Exception as e:
st.error(f"Error deleting chat history: {str(e)}")
return False
def generate_chat_title(messages):
"""Generate a title for the chat based on messages, with fallbacks."""
try:
for message in messages:
if message.get('role') == 'user' and message.get('content'):
user_message = message['content']
return user_message[:30] + "..." if len(user_message) > 30 else user_message
except (KeyError, TypeError):
pass
return "New Chat" # Default fallback title
def get_user_name(user_id):
try:
ref = firebase_db.reference(f'users/{user_id}')
user_data = ref.get()
return user_data.get('name', 'User') if user_data else 'User'
except Exception as e:
st.error(f"Error fetching user data: {str(e)}")
return 'User'
# Use the pyrebase auth object for login/signup
def login_signup_ui():
inject_custom_css()
st.sidebar.markdown("""
<div style="text-align: center; margin-bottom: 30px;">
<h1 style="color: #00ffff; font-size: 28px; margin-bottom: 5px;">πŸ” ApnaLawyer</h1>
<p style="color: rgba(255,255,255,0.7); font-size: 14px;">Secure Legal Assistance Portal</p>
</div>
""", unsafe_allow_html=True)
choice = st.sidebar.radio("Select Option", ["Login", "Signup", "Forgot Password"], label_visibility="collapsed")
if choice == "Signup":
st.sidebar.markdown("### Create New Account")
name = st.sidebar.text_input("Full Name", key="signup_name")
email = st.sidebar.text_input("Email Address", key="signup_email")
password = st.sidebar.text_input("Password", type="password", key="signup_password")
confirm_password = st.sidebar.text_input("Confirm Password", type="password", key="signup_confirm")
if st.sidebar.button("Create Account", key="signup_button"):
if not name:
st.sidebar.error("Please enter your full name!")
elif not email:
st.sidebar.error("Email address is required!")
elif not password or not confirm_password:
st.sidebar.error("Password fields cannot be empty!")
elif password != confirm_password:
st.sidebar.error("Passwords do not match!")
else:
try:
user = auth.create_user_with_email_and_password(email, password)
if store_user_data(user['localId'], name, email):
auth.send_email_verification(user['idToken'])
st.sidebar.success("βœ… Account created! Please verify your email before logging in.")
except Exception as e:
error_str = str(e)
if "EMAIL_EXISTS" in error_str:
st.sidebar.warning("⚠ Email already exists. Please try a different email address.")
elif "WEAK_PASSWORD" in error_str:
st.sidebar.warning("⚠ Password should be at least 6 characters long.")
else:
st.sidebar.error(f"Error: {error_str}")
elif choice == "Login":
st.sidebar.markdown("### Welcome Back")
email = st.sidebar.text_input("Email Address", key="login_email")
password = st.sidebar.text_input("Password", type="password", key="login_password")
if st.sidebar.button("Login", key="login_button"):
if not email:
st.sidebar.error("βœ‹ Please enter your email address")
elif not password:
st.sidebar.error("βœ‹ Please enter your password")
else:
try:
user = auth.sign_in_with_email_and_password(email, password)
user_info = auth.get_account_info(user['idToken'])
email_verified = user_info['users'][0]['emailVerified']
if email_verified:
user_name = get_user_name(user['localId'])
st.session_state.logged_in = True
st.session_state.user_email = email
st.session_state.user_token = user['idToken']
st.session_state.user_name = user_name
st.sidebar.success(f"πŸŽ‰ Welcome back, {user_name}!")
st.rerun()
else:
st.sidebar.warning("πŸ“§ Email not verified. Please check your inbox.")
if st.sidebar.button("πŸ” Resend Verification Email", key="resend_verification"):
auth.send_email_verification(user['idToken'])
st.sidebar.info("πŸ“¬ Verification email sent again!")
except Exception as e:
error_str = str(e)
if "EMAIL_NOT_FOUND" in error_str or "no user record" in error_str.lower():
st.sidebar.error("πŸ“­ Account not found")
st.sidebar.warning("Don't have an account? Please sign up.")
elif "INVALID_PASSWORD" in error_str or "INVALID_LOGIN_CREDENTIALS" in error_str:
st.sidebar.error("πŸ” Incorrect password")
else:
st.sidebar.error("⚠ Login error. Please try again.")
elif choice == "Forgot Password":
st.sidebar.markdown("### Reset Your Password")
email = st.sidebar.text_input("Enter your email address", key="reset_email")
if st.sidebar.button("Send Reset Link", key="reset_button"):
if not email:
st.sidebar.error("Please enter your email address!")
else:
try:
auth.send_password_reset_email(email)
st.sidebar.success("πŸ“¬ Password reset link sent to your email.")
except Exception as e:
error_str = str(e)
if "EMAIL_NOT_FOUND" in error_str:
st.sidebar.error("❌ No account found with this email address")
st.sidebar.warning("⚠ Don't have an account? Please sign up.")
else:
st.sidebar.error(f"Error: {error_str}")
# In your login_signup_ui() function, replace the Google login section with:
st.sidebar.markdown("---")
st.sidebar.markdown("### Or continue with")
if st.sidebar.button("Continue with Google", key="google_login"):
try:
# Generate a unique state token
state_token = hashlib.sha256(str(time.time()).encode()).hexdigest()
# Get Firebase config values
client_id = "546645596018-nvtkegm7mi8e83upfv771tv6t58c7snn.apps.googleusercontent.com" # Use actual OAuth client ID
firebase_auth_domain = firebaseConfig["authDomain"]
redirect_uri = f"https://{firebase_auth_domain}/auth/handler"
# Build the Google OAuth URL
auth_url = (
f"https://accounts.google.com/o/oauth2/v2/auth?"
f"response_type=code&"
f"client_id={client_id}&"
f"redirect_uri={urllib.parse.quote(redirect_uri)}&"
f"scope=email%20profile%20openid&"
f"state={state_token}"
)
# Store state token in session
st.session_state.oauth_state = state_token
# Open OAuth flow in new tab
components.html(
f"""
<script>
window.open('{auth_url}', '_blank').focus();
</script>
""",
height=0
)
st.sidebar.info("Google login window should open. Please allow popups if it doesn't appear.")
except Exception as e:
st.sidebar.error(f"Failed to start Google login: {str(e)}")
def check_google_callback():
try:
if 'code' in st.query_params and 'state' in st.query_params:
# Verify state token matches
if st.query_params['state'] != st.session_state.get('oauth_state'):
st.error("Security verification failed. Please try logging in again.")
return
auth_code = st.query_params['code']
# Initialize the Google Auth Provider
provider = firebase.auth.GoogleAuthProvider()
# Sign in with the auth code
credential = provider.credential(
None, # No ID token needed for code flow
auth_code
)
# Sign in with credential
user = auth.sign_in_with_credential(credential)
# Store user in session
st.session_state.logged_in = True
st.session_state.user_email = user['email']
st.session_state.user_name = user.get('displayName', 'Google User')
st.session_state.user_token = user['idToken']
# Store user data if new
if not user_exists(user['localId']):
store_user_data(
user['localId'],
st.session_state.user_name,
user['email']
)
# Clear the OAuth code from URL
st.query_params.clear()
st.rerun()
except Exception as e:
st.error(f"Google login failed: {str(e)}")
# Initialize speech-to-text model (cached to avoid reloading)
# @st.cache_resource
# def load_speech_to_text_model():
# return pipeline("automatic-speech-recognition", model="openai/whisper-base")
# Function to translate text
def transcribe_audio(audio_bytes, auto_detect=False):
"""Transcribe audio with auto language detection (English/Hindi)"""
if not SPEECHMATICS_API_KEY:
st.error("API key not configured!")
return None
try:
API_BASE_URL = "https://asr.api.speechmatics.com/v2"
MAX_FILE_SIZE = 5 * 1024 * 1024 # 5MB free tier limit
TIMEOUT = 30 # seconds
# 1. Validate audio size
if len(audio_bytes) > MAX_FILE_SIZE:
st.error(f"Audio exceeds {MAX_FILE_SIZE/1024/1024}MB free tier limit")
return None
# 2. Create temp WAV file
with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as tmpfile:
tmpfile.write(audio_bytes)
tmp_path = tmpfile.name
# 3. Configure language settings
language_config = {
"language": "auto" if auto_detect else st.session_state.get('language', 'en')
}
# For auto-detect, specify possible languages (improves accuracy)
if auto_detect:
language_config["language_options"] = ["en", "hi"] # English/Hindi only
job_config = {
"type": "transcription",
"transcription_config": {
**language_config,
"operating_point": "standard",
"enable_entities": False
}
}
headers = {"Authorization": f"Bearer {SPEECHMATICS_API_KEY}"}
# 4. Create job
with open(tmp_path, 'rb') as audio_file:
response = requests.post(
f"{API_BASE_URL}/jobs",
headers=headers,
files={
'config': (None, json.dumps(job_config)),
'data_file': ('audio.wav', audio_file)
},
timeout=TIMEOUT
)
# 5. Handle response
if response.status_code != 201:
error_msg = response.json().get('error', {}).get('message', response.text)
st.error(f"Job creation failed: {error_msg}")
return None
job_id = response.json()['id']
st.session_state.current_job_id = job_id
# 6. Poll for completion
start_time = time.time()
detected_language = None
while True:
if time.time() - start_time > TIMEOUT:
raise Exception("Timeout waiting for transcription")
status_response = requests.get(
f"{API_BASE_URL}/jobs/{job_id}",
headers=headers,
timeout=TIMEOUT
)
status_data = status_response.json()
# Capture detected language if auto mode
if auto_detect and not detected_language:
detected_language = status_data['job'].get('detected_language')
if detected_language:
st.info(f"πŸ” Detected language: {detected_language.upper()}")
if status_data['job']['status'] == 'done':
break
elif status_data['job']['status'] == 'failed':
raise Exception(f"Transcription failed: {status_data.get('error')}")
time.sleep(2)
# 7. Get transcript
transcript_response = requests.get(
f"{API_BASE_URL}/jobs/{job_id}/transcript",
headers=headers,
params={'format': 'txt'},
timeout=TIMEOUT
)
if transcript_response.status_code != 200:
st.error(f"Failed to fetch transcript: {transcript_response.text}")
return None
return transcript_response.text
except Exception as e:
st.error(f"Transcription error: {str(e)}")
return None
finally:
if 'tmp_path' in locals() and os.path.exists(tmp_path):
try:
os.unlink(tmp_path)
except:
pass
def check_rate_limit():
"""Simple rate limiting for audio transcription"""
if 'last_transcription_time' not in st.session_state:
st.session_state.last_transcription_time = time.time()
return True
current_time = time.time()
time_since_last = current_time - st.session_state.last_transcription_time
if time_since_last < 10: # 10 second cooldown between transcriptions
return False
st.session_state.last_transcription_time = current_time
return True
def create_new_chat():
"""Properly reset the chat state and start a new chat session."""
# Save current chat if it has messages
if len(st.session_state.get('messages', [])) > 1: # More than just welcome message
save_current_chat()
# Reset conversation state
st.session_state.messages = [{
"role": "assistant",
"content": f"πŸŽ‰βœ¨ Welcome {st.session_state.user_name}! βœ¨πŸŽ‰\n\nI'm ApnaLawyer, your AI legal assistant. "
"I can help explain Indian laws in simple terms. What would you like to know?"
}]
st.session_state.current_chat_id = None
st.session_state.memory = ConversationBufferWindowMemory(k=2, memory_key="chat_history", return_messages=True)
# Clear any audio processing flags
if 'audio_processed' in st.session_state:
del st.session_state.audio_processed
st.rerun()
def save_current_chat():
"""Save the current chat to history before starting a new one."""
if len(st.session_state.get('messages', [])) > 1: # More than just welcome message
chat_title = generate_chat_title(st.session_state.messages)
if hasattr(st.session_state, 'current_chat_id') and st.session_state.current_chat_id:
update_chat_history(st.session_state.user_id, st.session_state.current_chat_id, st.session_state.messages)
else:
chat_id = save_chat_to_history(st.session_state.user_id, chat_title, st.session_state.messages)
st.session_state.current_chat_id = chat_id
st.session_state.chat_history = get_chat_history(st.session_state.user_id)
def load_chat(chat_id):
"""Load a specific chat from history."""
# Save current chat if it has messages
if len(st.session_state.get('messages', [])) > 1: # More than just welcome message
save_current_chat()
# Load the selected chat
chat_data = next((chat for chat in st.session_state.chat_history if chat[0] == chat_id), None)
if chat_data:
st.session_state.messages = chat_data[1]['messages']
st.session_state.current_chat_id = chat_id
# Reinitialize memory with loaded messages
st.session_state.memory = ConversationBufferWindowMemory(k=2, memory_key="chat_history", return_messages=True)
for msg in chat_data[1]['messages'][:-2]: # Skip last 2 messages to maintain window size
if msg['role'] == 'user':
st.session_state.memory.save_context({"question": msg['content']}, {"answer": ""})
st.rerun()
def delete_chat(chat_id):
"""Delete a specific chat from the user's chat history."""
try:
user_id = st.session_state.user_id
if delete_chat_history(user_id, chat_id):
# Show toast notification instead of message in chat history
st.toast("Chat deleted successfully!", icon="βœ…")
# Refresh chat history after deletion
st.session_state.chat_history = get_chat_history(user_id)
st.rerun() # Force UI refresh
else:
st.toast("Failed to delete chat.", icon="❌")
except Exception as e:
st.toast(f"Error deleting chat: {str(e)}", icon="❌")
def chatbot_ui():
# Initialize language state
target_lang_code = "en" # Default language is English
if "language" not in st.session_state:
st.session_state.language = target_lang_code # Default language is English
# Initialize with personalized welcome message if first time
if "messages" not in st.session_state:
st.session_state.messages = [{
"role": "assistant",
"content": f"πŸŽ‰βœ¨ Welcome {st.session_state.user_name}! βœ¨πŸŽ‰\n\nI'm ApnaLawyer, your AI legal assistant. "
"I can help explain Indian laws in simple terms. What would you like to know?"
}]
if "memory" not in st.session_state:
st.session_state.memory = ConversationBufferWindowMemory(k=5, memory_key="chat_history", return_messages=True)
# Get user ID
if "user_id" not in st.session_state:
try:
user_info = auth.get_account_info(st.session_state.user_token)
st.session_state.user_id = user_info['users'][0]['localId']
st.session_state.chat_history = get_chat_history(st.session_state.user_id)
except Exception as e:
st.error(f"Error getting user info: {str(e)}")
@st.cache_resource
def load_embeddings():
return HuggingFaceEmbeddings(model_name="law-ai/InLegalBERT")
embeddings = load_embeddings()
db = FAISS.load_local("ipc_embed_db", embeddings, allow_dangerous_deserialization=True)
db_retriever = db.as_retriever(search_type="similarity", search_kwargs={"k": 3})
# Define the prompt template
prompt_template = PromptTemplate(
input_variables=["context", "question", "chat_history"],
template="""
<s>[INST]
You are ApnaLawyer, a trusted and knowledgeable AI assistant for Indian citizens. You provide legally accurate help related to Indian laws β€” including the Indian Penal Code (IPC), CrPC, POCSO Act, Domestic Violence Act, and others.
### Your responsibilities:
- Use clear, simple, respectful language
- Accurately cite laws (IPC sections, CrPC, Acts) **only when asked for legal explanation**
- If the user asks you to "write", "draft", "create", or "format" a legal document or application, you must write a **formal legal draft**
- Do not mix legal explanations with the draft unless asked β€” keep your response focused
- You may write drafts such as:
- FIR applications
- Police complaints
- Legal notices
- Affidavits
- Consent forms
- When drafting, use correct legal formatting, salutation, subject lines, and placeholders (name, address, date)
### CONTEXT:
{context}
### CHAT HISTORY:
{chat_history}
### QUESTION:
{question}
---
Based on the user's intent, choose **one of the following** response types:
---
πŸ“˜ If the user is asking about the law, respond with:
βœ… **Answer**:
[Summary of the situation and legal explanation]
πŸ“˜ **Relevant Law(s)**:
[List exact IPC sections, Acts, clause-wise punishment, and applicable exceptions]
🧾 **Other Related Laws**:
[Include CrPC, POCSO, DV Act, or procedural laws if relevant]
πŸ“ **Suggested Action**:
[Practical next steps β€” where to file, what to prepare]
🧾 **Summary**:
[Short recap in plain language]
---
✍️ If the user wants you to write or draft something, respond ONLY with:
πŸ“„ **Legal Draft/Application**:
[Write the complete legal document in clean, formal format using Indian legal norms. Use placeholders where needed.]
</s>[/INST]
"""
)
# Initialize the ConversationalRetrievalChain
qa = ConversationalRetrievalChain.from_llm(
llm=llm,
memory=st.session_state.memory,
retriever=db_retriever,
combine_docs_chain_kwargs={'prompt': prompt_template}
)
# Display all previous messages
for message in st.session_state.messages:
with st.chat_message(message["role"]):
st.markdown(message["content"])
# Text input box - always shown at the bottom
text_input = st.chat_input("Ask your legal question or record audio...")
# Handle text input
if text_input and 'audio_processed' not in st.session_state:
# Display user message immediately
with st.chat_message("user"):
st.markdown(text_input)
# Add to message history
st.session_state.messages.append({"role": "user", "content": text_input})
# Generate and display response
with st.chat_message("assistant"):
with st.spinner("Thinking..."):
# Auto-detect language from user prompt
# Auto-detect language from prompt
target_lang_name = detect_target_language(text_input)
if target_lang_name == "unsupported":
st.warning("⚠ Currently only Hindi, English, and Hinglish are supported.")
return
target_lang_code = supported_languages.get(target_lang_name, "en")
# Translate if necessary
mod_input = f"""Please answer the following question in {target_lang_name} language, using clear legal terms:
{text_input}"""
result = qa.invoke(input=mod_input)
answer = result["answer"]
message_placeholder = st.empty()
full_response = "⚠ Gentle reminder: We generally ensure precise information, but do double-check. \n\n\n"
for chunk in answer:
full_response += chunk
time.sleep(0.006)
message_placeholder.markdown(full_response + " |", unsafe_allow_html=True)
st.session_state.messages.append({"role": "assistant", "content": full_response})
# Update chat history
update_chat_history_function()
# Audio input - shown below the text input
audio_file = st.audio_input("", key="audio_recorder")
# Handle audio input (only if no text input was processed in this cycle)
if audio_file and 'audio_processed' not in st.session_state and not text_input:
audio_bytes = audio_file.getvalue()
# Set flag to prevent duplicate processing
st.session_state.audio_processed = True
with st.spinner("Transcribing..."):
try:
transcribed_text = transcribe_audio(audio_bytes)
# Display transcribed text immediately
with st.chat_message("user"):
st.markdown(transcribed_text)
st.session_state.messages.append({"role": "user", "content": transcribed_text})
# Generate and display response
with st.chat_message("assistant"):
with st.spinner("Thinking..."):
# Detect desired language from user input
# Detect desired language from user input
target_lang_name = detect_target_language(transcribed_text)
if target_lang_name == "unsupported":
st.warning("⚠ Currently only Hindi, English, and Hinglish are supported.")
return
target_lang_code = supported_languages.get(target_lang_name, "en")
# Handle Hinglish separately (keep original)
mod_input = f"""Please answer the following question in {target_lang_name} language, using clear legal terms:
{transcribed_text}"""
result = qa.invoke(input=mod_input)
answer = result["answer"]
# else: leave in English or handle more languages later
message_placeholder = st.empty()
full_response = "⚠ Gentle reminder: We generally ensure precise information, but do double-check. \n\n\n"
for chunk in answer:
full_response += chunk
time.sleep(0.006)
message_placeholder.markdown(full_response + " |", unsafe_allow_html=True)
st.session_state.messages.append({"role": "assistant", "content": full_response})
# Update chat history
update_chat_history_function()
except Exception as e:
st.error(f"Error: {str(e)}")
finally:
# Clear the flag after processing
if 'audio_processed' in st.session_state:
del st.session_state.audio_processed
# Sidebar UI (unchanged from your original)
with st.sidebar:
st.markdown(f"""
<div style="margin-bottom: 20px;">
<h3 style="color: #00ffff; margin-bottom: 5px;">πŸ‘€ {st.session_state.user_name}</h3>
<p style="color: rgba(255,255,255,0.7); font-size: 12px; margin-top: 0;">{st.session_state.user_email}</p>
</div>
""", unsafe_allow_html=True)
# Button for creating a new chat
if st.button("βž• New Chat", key="new_chat_button", use_container_width=True):
create_new_chat()
# Chat history section
if hasattr(st.session_state, 'chat_history') and st.session_state.chat_history:
for chat_id, chat_data in st.session_state.chat_history:
# Safely get title with default
chat_title = chat_data.get('title', 'Untitled Chat')
timestamp = time.strftime('%d %b %Y, %I:%M %p',
time.localtime(chat_data.get('last_updated', time.time())))
col1, col2 = st.columns([0.8, 0.2])
with col1:
if st.button(
f"{chat_title}",
key=f"chat_{chat_id}", # Unique key for each chat button
help=f"Last updated: {timestamp}",
use_container_width=True
):
load_chat(chat_id)
with col2:
if st.button("πŸ—‘", key=f"delete_{chat_id}"): # Unique key for each delete button
delete_chat(chat_id)
else:
st.markdown("<p style='color: rgba(255,255,255,0.5);'>No chat history yet</p>", unsafe_allow_html=True)
# Logout button
if st.button("πŸšͺ Logout", key="logout_button", use_container_width=True):
st.session_state.logged_in = False
st.session_state.user_email = None
st.session_state.user_name = None
st.session_state.user_id = None
st.rerun()
def update_chat_history_function():
"""Helper function to update chat history"""
if len(st.session_state.messages) > 2:
chat_title = generate_chat_title(st.session_state.messages)
if hasattr(st.session_state, 'current_chat_id') and st.session_state.current_chat_id:
update_chat_history(st.session_state.user_id, st.session_state.current_chat_id, st.session_state.messages)
else:
chat_id = save_chat_to_history(st.session_state.user_id, chat_title, st.session_state.messages)
st.session_state.current_chat_id = chat_id
st.session_state.chat_history = get_chat_history(st.session_state.user_id)
# ----------------- Main App -------------------
if "logged_in" not in st.session_state:
st.session_state.logged_in = False
if not st.session_state.logged_in:
login_signup_ui()
else:
# Main chat interface
chatbot_ui()
# Footer
footer()