main.py

import numpy as np
from sklearn.metrics.pairwise import cosine_similarity
from utils.convert_embedding import GetEmbedding
import random
import pickle
import os 
from utils.rag import RAG
from faster_whisper import WhisperModel


def process(user_query:str):
    # dump_user_question(user_query)
    user_embedding = GetEmbedding([user_query]).user_query_emb()
    with open(r"all_mix_embedding.pkl","rb") as f:
        load_embedding = pickle.load(f)

    with open(r"all_answers.pkl","rb") as f:
        ans = pickle.load(f)
    similarity_scores = cosine_similarity(user_embedding, load_embedding)
    index = np.argmax(similarity_scores)
    answer = ans[index]
    score = similarity_scores[0,index]
    print(f"Index : {index}:\tscore:{score} \tquery: {user_query}")

    if float(score) > 0.60 :
        final_output = random.choice(answer)
    else:
        final_output = RAG().pipeline(query=user_query)

    return final_output


def audio_process(audio):
    try:
        model = WhisperModel(model_size_or_path="medium.en")
        segments, info = model.transcribe(audio)
        transcription = " ".join([seg.text for seg in segments])
        result = process(user_query=transcription)
        return result
    except Exception as e:
        print("Error:", e)
        return str(e)




if __name__ == "__main__":
    res = audio_process(r"C:\Users\lalit\Documents\Sound recordings\who_is_lalit.m4a")
    print(res)
    # for _ in range(3):
    
    #     user = input("How can i help you :? \n")
    #     result = process(user)
    #     print(result)

    # with open(r"data\question_data.pkl","rb") as f:
    #     que = pickle.load(f)
    # print(que)