embedding.py

import openai
import numpy as np  # standard math module for python
from pprint import pprint


def open_file(filepath):
    with open(filepath, 'r', encoding='utf-8') as infile:
        return infile.read()


def gpt3_embedding(content, engine='text-similarity-babbage-001'):
    content = content.encode(encoding='ASCII',errors='ignore').decode()
    response = openai.Embedding.create(input=content,engine=engine)
    vector = response['data'][0]['embedding']  # this is a normal list
    return vector


def similarity(v1, v2):  # return dot product of two vectors
    return np.dot(v1, v2)


openai.api_key = open_file('openaiapikey.txt')


def match_class(vector, classes):
    results = list()
    for c in classes:
        score = similarity(vector, c['vector'])
        info = {'category': c['category'], 'score': score}
        results.append(info)
    return results


if __name__ == '__main__':
    categories = ['plant', 'reptile', 'mammal', 'fish', 'bird', 'pet', 'wild animal']
    classes = list()
    for c in categories:
        vector = gpt3_embedding(c)
        info = {'category': c, 'vector': vector}
        classes.append(info)
    #print(classes)
    #exit(0)
    while True:
        a = input('Enter a lifeform here: ')
        vector = gpt3_embedding(a)
        #print(a, vector)
        result = match_class(vector, classes)
        pprint(result)