app.py

import os
import textwrap
from dotenv import load_dotenv
import gradio as gr
from haystack import Pipeline
from haystack.utils import Secret
from haystack.components.builders import PromptBuilder
from haystack.components.generators import OpenAIGenerator
import pandas as pd

load_dotenv()

MODEL = "microsoft/Phi-3-mini-4k-instruct"

# Load the CSV file
df = pd.read_csv("dataset.csv")

# Set up components
prompt_template = """
Based on the Indian Union Budget data for FY 21-22 to 23-24:

{{budget_data}}

Answer the given question: {{query}}
Answer:
"""
prompt_builder = PromptBuilder(template=prompt_template)
llm = OpenAIGenerator(
    api_key=Secret.from_env_var("MONSTER_API_KEY"),
    api_base_url="https://llm.monsterapi.ai/v1/", 
    model=MODEL,
    generation_kwargs={"max_tokens": 512}
)
pipeline = Pipeline()
pipeline.add_component("prompt", prompt_builder)
pipeline.add_component("llm", llm)

pipeline.connect("prompt.prompt", "llm.prompt")

# Function to handle the query
# def answer_query(query):
#     # Convert DataFrame to string representation
#     budget_data = df.to_string()
#     result = pipeline.run({"prompt": {"budget_data": budget_data, "query": query}})
#     return result["llm"]["replies"][0]


def answer_query(query):
    try:
        # Select a subset of the data (adjust as needed)
        sample_data = df.sample(n=10).to_string()
        
        # Truncate the data if it's too long
        budget_data = textwrap.shorten(sample_data, width=1000, placeholder="...")
        
        result = pipeline.run({"prompt": {"budget_data": budget_data, "query": query}})
        return result["llm"]["replies"][0]
    except Exception as e:
        return f"An error occurred: {str(e)}"

# Gradio interface
def chat_interface(query):
    return answer_query(query)

with gr.Blocks() as demo:
    gr.Markdown("# Indian 2024 Budget Chatbot")
    query_input = gr.Textbox(label="Enter Your Question")
    submit_button = gr.Button("Get Answer")
    output_text = gr.Textbox(label="Answer", interactive=False)
    
    submit_button.click(fn=chat_interface, inputs=[query_input], outputs=output_text)

# Run the app locally
if __name__ == "__main__":
    demo.launch()