recommend.py

import cohere
import os
import pinecone
import pandas as pd 
import numpy as np
from dotenv import dotenv_values

env_name = ".env"
config = dotenv_values(env_name)

co = cohere.Client(os.environ['cohere_api'])

pinecone.init(
    api_key= config['pinecone_api'],
    environment= config['pinecone_env']
)

index = pinecone.Index(config['pinecone_index'])



def embed_text(text):
    embeddings = co.embed(
        texts=[text],
        model='embed-english-v3.0',
        input_type='search_query'
    )

    return embeddings



def vector_search(desc):

    results = []

    embeddings = co.embed(
        texts=[desc],
        model='embed-english-v3.0',
        input_type='search_query'
    )

    res = index.query([embeddings.embeddings[0]], top_k=2, include_metadata=True)

    for match in res['matches']:
        results.append(match['metadata']['text'])

    return results

def get_anime(res):
    data = pd.read_csv('mini_data.csv')

    df = pd.DataFrame()

    for desc in res:
        anime = data[data['description'] == desc]

        df = pd.concat([df, anime])

    return df