pages/admin_utils.py

import os

import pandas as pd
import pinecone
from dotenv import load_dotenv
from langchain.embeddings import OpenAIEmbeddings
from langchain.embeddings.sentence_transformer import \
    SentenceTransformerEmbeddings
from langchain.llms import OpenAI
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.vectorstores import Pinecone
from pypdf import PdfReader
from sklearn.model_selection import train_test_split
from functools import lru_cache


#**********Functions to help you load documents to PINECONE***********

#Read PDF data
def read_pdf_data(pdf_file):
    pdf_page = PdfReader(pdf_file)
    text = ""
    for page in pdf_page.pages:
        text += page.extract_text()
    return text

#Split data into chunks
def split_data(text):
    text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=50)
    docs = text_splitter.split_text(text)
    docs_chunks =text_splitter.create_documents(docs)
    return docs_chunks

#Create embeddings instance
def create_embeddings_load_data():
    #embeddings = OpenAIEmbeddings()
    embeddings = SentenceTransformerEmbeddings(model_name="all-MiniLM-L6-v2")
    return embeddings


@lru_cache
def pine_cone_index(pinecone_index_name: str | None):
    load_dotenv()
    pinecone.init(
    api_key=os.getenv('PINECONE_API_KEY'),
    environment=os.getenv('PINECONE_ENV'),
    )

    index_name = pinecone_index_name or os.getenv('PINECONE_INDEX_NAME')
    if index_name is None:
        raise ValueError('PINECONE_INDEX_NAME is not set')
    return index_name


def push_to_pinecone(embeddings,docs,pinecone_index_name: str | None=None):
    index_name = pine_cone_index(pinecone_index_name)
    index = Pinecone.from_documents(docs, embeddings, index_name=index_name)
    return index

#*********Functions for dealing with Model related tasks...************

#Read dataset for model creation
def read_data(data):
    df = pd.read_csv(data,delimiter=',', header=None)  
    return df

#Create embeddings instance
def get_embeddings():
    embeddings = SentenceTransformerEmbeddings(model_name="all-MiniLM-L6-v2")
    return embeddings

#Generating embeddings for our input dataset
def create_embeddings(df,embeddings):
    df[2] = df[0].apply(lambda x: embeddings.embed_query(x))
    return df

#Splitting the data into train & test
def split_train_test__data(df_sample):
    # Split into training and testing sets
    sentences_train, sentences_test, labels_train, labels_test = train_test_split(
    list(df_sample[2]), list(df_sample[1]), test_size=0.25, random_state=0)
    print(len(sentences_train))
    return sentences_train, sentences_test, labels_train, labels_test

#Get the accuracy score on test data
def get_score(svm_classifier,sentences_test,labels_test):
    score = svm_classifier.score(sentences_test, labels_test)
    return score