app.py

import gradio as gr
import random
import time
from sentence_transformers import SentenceTransformer
import pinecone
import os

retriever = SentenceTransformer("sentence-transformers/all-MiniLM-L12-v2")

pinecone_key = os.environ['PINECONE_SECRET']

pinecone.init(
    api_key=os.environ['PINECONE_SECRET'],
    environment="eu-west1-gcp"
)

index_name = 'quran-semantic-search'

index = pinecone.Index(index_name)


def query_pinecone(query, top_k):
  xq = retriever.encode([query]).tolist()

  # search vector database for similar vector
  xc = index.query(xq, top_k=top_k, include_metadata=True)

  return xc['matches']

def format_search_result(result):
    data = result['metadata']
    message = f"Ayah no: {data['ayah']}\nSurah no: {data['surah']}\nSentence:{data['arabic-text']}\nTranslation: {data['en-translation']}\n Tafsir:{data['en-tafsir-mokhtasar']}\n Relevant Tafsir: {data['vector-chunk']}"

    return message


with gr.Blocks() as demo:
    gr.Markdown(
    """
    # Quran Semantic Search!
    Ask questions and seek advice and see what the Holy book has to say about it.
    """)
    chatbot = gr.Chatbot()
    msg = gr.Textbox()
    clear = gr.Button("Clear")

    def user(user_message, history):
        return "", history + [[user_message, None]]

    def bot(history):
        
        query = history[-1][0]

        results = query_pinecone(query, top_k=3)

        for match in results:
            if history[-1][1] == None:
                history[-1][1] = format_search_result(match)

            else:
                history.append([None, format_search_result(match)])  
        
        return history

    msg.submit(user, [msg, chatbot], [msg, chatbot], queue=False).then(
        bot, chatbot, chatbot
    )
    clear.click(lambda: None, None, chatbot, queue=False)


demo.launch()