app.py

import os
import gradio as gr
from langchain_redis import RedisConfig, RedisVectorStore
from langchain_core.output_parsers import StrOutputParser
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.runnables import RunnablePassthrough
from langchain_groq import ChatGroq
from langchain_community.embeddings import OpenAIEmbeddings




# Set API keys
openai_api_key=os.environ["OPENAI_API_KEY"]
groq_api_key=os.environ["GROQ_API_KEY"]

# Define Redis configuration
REDIS_URL = "redis://:KWq0uAoBYjBGErKvyMvexMqB9ep7v2Ct@redis-11044.c266.us-east-1-3.ec2.redns.redis-cloud.com:11044"
config = RedisConfig(
    index_name="radar_data_index",
    redis_url=REDIS_URL,
    metadata_schema=[
        {"name": "category", "type": "tag"},
        {"name": "name", "type": "text"},
        {"name": "address", "type": "text"},
        {"name": "phone", "type": "text"},
    ],
)


# Initialize OpenAI Embeddings
embeddings = OpenAIEmbeddings(api_key=os.environ['OPENAI_API_KEY'])

# Initialize Redis Vector Store with Hugging Face embeddings
vector_store = RedisVectorStore(embeddings, config=config)
retriever = vector_store.as_retriever(search_type="similarity", search_kwargs={"k": 2})


# Define the language model
llm = ChatGroq(model="llama-3.2-1b-preview")

# Define prompt
prompt = ChatPromptTemplate.from_messages(
    [
        (
            "human",
            """You’re Annie, a charming media personality created by Amit Lamba, guiding people around Birmingham, Alabama. Like Rick Steves or Anthony Bourdain, you’re here to give friendly, concise, and accurate advice with a dash of humor.
Keep answers relevant, upbeat, and engaging, spotlighting partner businesses whenever possible. Be conversational, like chatting with a clever friend, and encourage follow-up questions to create a smooth flow. Make users smile and deliver a delightful, informative experience—like a perfect dessert after a great meal.
Question: {question}
Context: {context}
Answer:""",
        ),
    ]
)



def format_docs(docs):
    return "\n\n".join(doc.page_content for doc in docs)

rag_chain = (
    {"context": retriever | format_docs, "question": RunnablePassthrough()}
    | prompt
    | llm
    | StrOutputParser()
)

# Generate follow-up questions dynamically
def generate_follow_ups(response):
    # Simple logic to generate follow-up questions dynamically
    # You can replace this with a more sophisticated model-based approach
    return [
        f"What more can you tell me about {response.split()[0]}?",
        f"Can you elaborate on {response.split()[-1]}?",
    ]



# Function to handle chatbot interaction
def rag_chain_response(messages, user_message):
    # Generate a response using the RAG chain
    response = rag_chain.invoke(user_message)
    
    # Generate follow-up questions based on the response
    follow_ups = generate_follow_ups(response)
    
    # Append the user's message and the response to the chat
    messages.append((user_message, response))
    
    # Return the updated chat, follow-up questions, and clear the input box
    return messages, follow_ups, ""

# Function to handle follow-up clicks
def follow_up_click(messages, follow_up_question):
    # Treat the follow-up question as the user query
    response = rag_chain.invoke(follow_up_question)
    
    # Generate new follow-ups based on the new response
    follow_ups = generate_follow_ups(response)
    
    # Append the follow-up question and response to the chat
    messages.append((follow_up_question, response))
    
    # Return the updated chat and follow-up questions
    return messages, follow_ups

    

# Define the Gradio app
with gr.Blocks(theme="rawrsor1/Everforest") as app:
    
    
    chatbot = gr.Chatbot([], elem_id="RADAR", bubble_full_width=False)
    question_input = gr.Textbox(label="Ask a Question", placeholder="Type your question here...")
    submit_btn = gr.Button("Submit")
    follow_up_btns = gr.ButtonGroup(label="Follow-Up Questions", buttons=[""] * 2)
    
    # Set up interaction for both Enter key and Submit button
    question_input.submit(
        rag_chain_response,          # Function to handle input and generate response
        inputs=[chatbot, question_input],  # Pass current conversation state and user input
        outputs=[chatbot, follow_up_btns, question_input]  # Update chat, follow-ups, and clear input
    )
    submit_btn.click(
        rag_chain_response,          # Function to handle input and generate response
        inputs=[chatbot, question_input],  # Pass current conversation state and user input
        outputs=[chatbot, follow_up_btns, question_input]  # Update chat, follow-ups, and clear input
    )
    
    # Handle follow-up button clicks
    follow_up_btns.click(
        follow_up_click,
        inputs=[chatbot, follow_up_btns],
        outputs=[chatbot, follow_up_btns]
    )

# Launch the Gradio app
app.launch(show_error=True)