app.py

import gradio as gr
import httpx
import time
from datetime import datetime
from llama_index.core import VectorStoreIndex, SimpleDirectoryReader
from llama_parse import LlamaParse
from llama_index.embeddings.huggingface import HuggingFaceEmbedding
from llama_index.llms.huggingface_api import HuggingFaceInferenceAPI

# Define the LLM model and document processing
llm = HuggingFaceInferenceAPI(model_name="tiiuae/falcon-7b-instruct")
parser = LlamaParse(api_key='llx-zKtsC5UBLs8DOApOsLluXMBdQhC75ea0Vs80SmPSjsmDzuhh', result_type='markdown')
file_extractor = {'.pdf': parser}
documents = SimpleDirectoryReader('data/', file_extractor=file_extractor).load_data()

embed_model = HuggingFaceEmbedding(model_name="BAAI/bge-small-en-v1.5")
vector_index = VectorStoreIndex.from_documents(documents, embed_model=embed_model)
query_engine = vector_index.as_query_engine(llm=llm)

# Function to process query with retries
def query_with_retry(query, max_retries=3, wait_time=5):
    for attempt in range(max_retries):
        try:
            start_time = datetime.now()
            response = query_engine.query(query)
            end_time = datetime.now()
            duration = (end_time - start_time).total_seconds()
            return response
        except httpx.ReadTimeout:
            if attempt < max_retries - 1:
                time.sleep(wait_time)
            else:
                raise
        except Exception as e:
            return f"An error occurred: {e}"

# Define the Gradio interface
def gradio_interface(query):
    response = query_with_retry(query)
    return response

# Set up the Gradio UI
iface = gr.Interface(
    fn=gradio_interface,
    inputs=gr.Textbox(label="Ask a question", placeholder="Enter your query here...", lines=2),
    outputs=gr.Textbox(label="Response", interactive=False),
    title="Document-based Question Answering",
    description="Ask questions based on the documents you uploaded. This model answers queries using your uploaded PDFs.",
    live=True
)

iface.launch()