app.py

import gradio as gr
from huggingface_hub import InferenceClient
import os
import openai
import pandas as pd
import faiss
import pickle
from sentence_transformers import SentenceTransformer


embedding_model = SentenceTransformer('nomic-ai/nomic-embed-text-v1.5', trust_remote_code=True)
openai.api_key = os.getenv("OPENAI_API_KEY")
db_index = faiss.read_index("db_index.faiss")
df = pd.read_csv('cleaned_data.csv')
with open('metadata_info.pkl', 'rb') as file:
    metadata_info = pickle.load(file)


def search(query):
    cleaned_query = query
    query_embedding = embedding_model.encode(cleaned_query).reshape(1, -1).astype('float32')
    D, I = db_index.search(query_embedding, k=10)
    results = []
    for idx in I[0]:
        if idx < 3327:
            doc_index = idx
            results.append({
                'type': 'metadata',
                'title': df.iloc[doc_index]['title'],
                'author': df.iloc[doc_index]['author'],
                'publish_date': df.iloc[doc_index]['publish_date'],
                'full_text': df.iloc[doc_index]['full_text'],
                'source': df.iloc[doc_index]['url']
            })
        else:
            chunk_index = idx - 3327
            metadata = metadata_info[chunk_index]
            doc_index = metadata['index']
            chunk_text = metadata['chunk']
            results.append({
                'type': 'content',
                'title': df.iloc[doc_index]['title'],
                'author': df.iloc[doc_index]['author'],
                'publish_date': df.iloc[doc_index]['publish_date'],
                'content': chunk_text,
                'source': df.iloc[doc_index]['url']
            })

    return results


def generate_answer(query):
    prompt = f"""
    Based on the following query from a user, please generate a detailed answer based on the context
    focusing on which is the best based on the query. You should responsd as you are a news and politician expert agent and are conversing with the
    user in a nice cordial way. If the query question is not in the context say I don't know, and always provide the url as the source of the information.
    Remove the special characters and (/n ) , make the output clean and concise.

    ###########
    query:
    "{query}"

    ########

    context:"
    "{search(query)}"
    #####

    Return in Markdown format with each hotel highlighted.
    """

    messages = [
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": prompt}
    ]
    response = openai.ChatCompletion.create(
        model="gpt-4o-mini",
        max_tokens=1500,
        n=1,
        stop=None,
        temperature=0.2, #higher temperature means more creative or more hallucination
        messages = messages


    )

    # Extract the generated response from the API response
    generated_text = response.choices[0].message['content'].strip()

    return generated_text

"""
For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
"""
client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")


def respond(
    message,
    history: list[tuple[str, str]],
    system_message,
    max_tokens,
    temperature,
    top_p,
):
    response = generate_answer(message)
    yield response

"""
For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
"""
demo = gr.ChatInterface(
    respond,
    additional_inputs=[
        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
        gr.Slider(
            minimum=0.1,
            maximum=1.0,
            value=0.95,
            step=0.05,
            label="Top-p (nucleus sampling)",
        ),
    ],
)


if __name__ == "__main__":
    demo.launch()