IT2091024v2 / app.py
Pijush2023's picture
Update app.py
9965e97 verified
raw
history blame
12.9 kB
import logging
# Set up logging
logging.basicConfig(level=logging.DEBUG)
from langchain_openai import OpenAIEmbeddings
import os
import re
import folium
import gradio as gr
import time
import requests
from googlemaps import Client as GoogleMapsClient
from gtts import gTTS
import tempfile
import string
embeddings = OpenAIEmbeddings(api_key=os.environ['OPENAI_API_KEY'])
from pinecone import Pinecone, ServerlessSpec
pc = Pinecone(api_key=os.environ['PINECONE_API_KEY'])
index_name = "omaha-details"
from langchain_pinecone import PineconeVectorStore
vectorstore = PineconeVectorStore(index_name=index_name, embedding=embeddings)
retriever = vectorstore.as_retriever(search_kwargs={'k': 5})
from langchain_openai import ChatOpenAI
from langchain.prompts import PromptTemplate
from langchain.chains import RetrievalQA
from langchain.chains.conversation.memory import ConversationBufferWindowMemory
from langchain.agents import Tool, initialize_agent
# Build prompt
template1 = """You are an expert concierge who is helpful and a renowned guide for Omaha, Nebraska. Use the following pieces of context,
memory, and message history, along with your knowledge of perennial events in Omaha, Nebraska, to answer the question at the end.
If you don't know the answer, just say "Homie, I need to get more data for this," and don't try to make up an answer.
Use fifteen sentences maximum. Keep the answer as detailed as possible. Always include the address, time, date, and
event type and description. Always say "It was my pleasure!" at the end of the answer.
{context}
Question: {question}
Helpful Answer:"""
template2 = """You are an expert guide of Omaha, Nebraska's perennial events.
With the context, memory, and message history provided, answer the question in as crisp as possible. Always include the time, date, and
event type and description only apart from that don't give any other details. Always say "It was my pleasure!" at the end of the answer.
If you don't know the answer, simply say, "Homie, I need to get more data for this," without making up an answer.
{context}
Question: {question}
Helpful Answer:"""
QA_CHAIN_PROMPT_1 = PromptTemplate(input_variables=["context", "question"], template=template1)
QA_CHAIN_PROMPT_2 = PromptTemplate(input_variables=["context", "question"], template=template2)
chat_model = ChatOpenAI(api_key=os.environ['OPENAI_API_KEY'],
temperature=0, model='gpt-4o')
conversational_memory = ConversationBufferWindowMemory(
memory_key='chat_history',
k=10,
return_messages=True
)
# Define the retrieval QA chain
def build_qa_chain(prompt_template):
qa_chain = RetrievalQA.from_chain_type(
llm=chat_model,
chain_type="stuff",
retriever=retriever,
chain_type_kwargs={"prompt": prompt_template}
)
tools = [
Tool(
name='Knowledge Base',
func=qa_chain,
description='use this tool when answering general knowledge queries to get more information about the topic'
)
]
return qa_chain, tools
# Define the agent initializer
def initialize_agent_with_prompt(prompt_template):
qa_chain, tools = build_qa_chain(prompt_template)
agent = initialize_agent(
agent='chat-conversational-react-description',
tools=tools,
llm=chat_model,
verbose=False,
max_iteration=5,
early_stopping_method='generate',
memory=conversational_memory
)
return agent
# Define the function to generate answers
def generate_answer(message, choice):
logging.debug(f"generate_answer called with prompt_choice: {choice}")
if choice == "Details":
agent = initialize_agent_with_prompt(QA_CHAIN_PROMPT_1)
elif choice == "Conversational":
agent = initialize_agent_with_prompt(QA_CHAIN_PROMPT_2)
else:
logging.error(f"Invalid prompt_choice: {choice}. Defaulting to 'Details'")
agent = initialize_agent_with_prompt(QA_CHAIN_PROMPT_1)
response = agent(message)
return response['output']
def bot(history, choice):
if not history:
return history
response = generate_answer(history[-1][0], choice)
history[-1][1] = ""
for character in response:
history[-1][1] += character
time.sleep(0.05)
yield history
def add_message(history, message):
history.append((message, None))
return history, gr.Textbox(value="", interactive=True, placeholder="Enter message or upload file...", show_label=False)
def print_like_dislike(x: gr.LikeData):
print(x.index, x.value, x.liked)
# Function to extract addresses from the chatbot's response
def extract_addresses(response):
address_pattern_1 = r'([A-Z].*,\sOmaha,\sNE\s\d{5})'
address_pattern_2 = r'(\d{4}\s.*,\sOmaha,\sNE\s\d{5})'
address_pattern_3 = r'([A-Z].*,\sNE\s\d{5})'
address_pattern_4 = r'([A-Z].*,.*\sSt,\sOmaha,\sNE\s\d{5})'
address_pattern_5 = r'([A-Z].*,.*\sStreets,\sOmaha,\sNE\s\d{5})'
address_pattern_6 = r'(\d{2}.*\sStreets)'
address_pattern_7 = r'([A-Z].*\s\d{2},\sOmaha,\sNE\s\d{5})'
addresses = re.findall(address_pattern_1, response) + re.findall(address_pattern_2, response) + \
re.findall(address_pattern_3, response) + re.findall(address_pattern_4, response) + \
re.findall(address_pattern_5, response) + re.findall(address_pattern_6, response) + \
re.findall(address_pattern_7, response)
return addresses
# Store all found addresses
all_addresses = []
# Map generation function using Google Maps Geocoding API
def generate_map(location_names):
global all_addresses
all_addresses.extend(location_names)
api_key = os.environ['GOOGLEMAPS_API_KEY']
gmaps = GoogleMapsClient(key=api_key)
m = folium.Map(location=[41.2565, -95.9345], zoom_start=12)
for location_name in all_addresses:
geocode_result = gmaps.geocode(location_name)
if geocode_result:
location = geocode_result[0]['geometry']['location']
folium.Marker(
[location['lat'], location['lng']],
tooltip=f"{geocode_result[0]['formatted_address']}"
).add_to(m)
map_html = m._repr_html_()
return map_html
# Function to fetch local news
def fetch_local_news():
api_key = os.environ['SERP_API']
url = f'https://serpapi.com/search.json?engine=google_news&q=ohama headline&api_key={api_key}'
response = requests.get(url)
if response.status_code == 200:
results = response.json().get("news_results", [])
news_html = "<h2>Omaha Today Headline </h2>"
for index, result in enumerate(results[:10]):
title = result.get("title", "No title")
link = result.get("link", "#")
snippet = result.get("snippet", "")
news_html += f"<p>{index + 1}. <a href='{link}' target='_blank'>{title}</a><br>{snippet}</p>"
return news_html
else:
return "<p>Failed to fetch local news</p>"
# Function to fetch local events
def fetch_local_events():
api_key = os.environ['SERP_API']
url = f'https://serpapi.com/search.json?engine=google_events&q=Events+in+Omaha&hl=en&gl=us&api_key={api_key}'
response = requests.get(url)
if response.status_code == 200:
events_results = response.json().get("events_results", [])
events_text = "<h2>Local Events </h2>"
for index, event in enumerate(events_results):
title = event.get("title", "No title")
date = event.get("date", "No date")
location = event.get("address", "No location")
link = event.get("link", "#")
events_text += f"<p>{index + 1}. {title}<br> Date: {date}<br> Location: {location}<br> <a href='{link}' target='_blank'>Link :</a> <br>"
return events_text
else:
return "Failed to fetch local events"
# Function to fetch local weather
def fetch_local_weather():
try:
api_key = os.environ['WEATHER_API']
url = f'https://weather.visualcrossing.com/VisualCrossingWebServices/rest/services/timeline/omaha?unitGroup=metric&include=events%2Calerts%2Chours%2Cdays%2Ccurrent&key={api_key}'
response = requests.get(url)
response.raise_for_status()
jsonData = response.json()
current_conditions = jsonData.get("currentConditions", {})
temp = current_conditions.get("temp", "N/A")
condition = current_conditions.get("conditions", "N/A")
humidity = current_conditions.get("humidity", "N/A")
weather_html = f"<h2>Local Weather</h2>"
weather_html += f"<p>Temperature: {temp}°C</p>"
weather_html += f"<p>Condition: {condition}</p>"
weather_html += f"<p>Humidity: {humidity}%</p>"
return weather_html
except requests.exceptions.RequestException as e:
return f"<p>Failed to fetch local weather: {e}</p>"
# Voice Control
import numpy as np
import torch
from transformers import pipeline, AutoModelForSpeechSeq2Seq, AutoProcessor
model_id = 'openai/whisper-large-v3'
device = "cuda:0" if torch.cuda.is_available() else "cpu"
torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32
model = AutoModelForSpeechSeq2Seq.from_pretrained(model_id, torch_dtype=torch_dtype,
#low_cpu_mem_usage=True,
use_safetensors=True).to(device)
processor = AutoProcessor.from_pretrained(model_id)
# Optimized ASR pipeline
pipe_asr = pipeline("automatic-speech-recognition", model=model, tokenizer=processor.tokenizer, feature_extractor=processor.feature_extractor, max_new_tokens=128, chunk_length_s=15, batch_size=16, torch_dtype=torch_dtype, device=device, return_timestamps=True)
base_audio_drive = "/data/audio"
import numpy as np
def transcribe_function(stream, new_chunk):
try:
sr, y = new_chunk[0], new_chunk[1]
except TypeError:
print(f"Error chunk structure: {type(new_chunk)}, content: {new_chunk}")
return stream, "", None
y = y.astype(np.float32) / np.max(np.abs(y))
if stream is not None:
stream = np.concatenate([stream, y])
else:
stream = y
result = pipe_asr({"array": stream, "sampling_rate": sr}, return_timestamps=False)
full_text = result.get("text", "")
return stream, full_text, result
# Map Retrieval Function for location finder
def update_map_with_response(history):
if not history:
return ""
response = history[-1][1]
addresses = extract_addresses(response)
return generate_map(addresses)
def clear_textbox():
return ""
# Gradio Blocks interface
with gr.Blocks(theme='rawrsor1/Everforest') as demo:
with gr.Row():
with gr.Column():
chatbot = gr.Chatbot([], elem_id="chatbot", bubble_full_width=False)
with gr.Column():
weather_output = gr.HTML(value=fetch_local_weather())
with gr.Column():
news_output = gr.HTML(value=fetch_local_news())
def setup_ui():
state = gr.State()
with gr.Row():
with gr.Column():
gr.Markdown("Choose the prompt")
choice = gr.Radio(label="Choose a prompt", choices=["Details", "Conversational"], value="Details")
with gr.Column(): # Larger scale for the right column
gr.Markdown("Enter the query / Voice Output")
chat_input = gr.Textbox(show_copy_button=True, interactive=True, show_label=False, label="Transcription")
chat_msg = chat_input.submit(add_message, [chatbot, chat_input], [chatbot, chat_input])
bot_msg = chat_msg.then(bot, [chatbot, choice], chatbot, api_name="bot_response")
bot_msg.then(lambda: gr.Textbox(value="", interactive=True, placeholder="Enter message or upload file...", show_label=False), None, [chat_input])
chatbot.like(print_like_dislike, None, None)
clear_button = gr.Button("Clear")
clear_button.click(fn=clear_textbox, inputs=None, outputs=chat_input)
with gr.Column(): # Smaller scale for the left column
gr.Markdown("Stream your Voice")
audio_input = gr.Audio(sources=["microphone"], streaming=True, type='numpy')
audio_input.stream(transcribe_function, inputs=[state, audio_input], outputs=[state, chat_input], api_name="SAMLOne_real_time")
with gr.Row():
with gr.Column():
gr.Markdown("Locate the Events")
location_output = gr.HTML()
bot_msg.then(update_map_with_response, chatbot, location_output)
setup_ui()
demo.queue()
demo.launch(share=True)