app_old.py

import requests
import os
os.system('pip install openpyxl')
os.system('pip install sentence-transformers')

def chatgpt3_question(context, question):
    api_key = "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"
    url = "https://api.openai.com/v1/chat/completions"

    prompt = f"""
    based on this context: {context}
    answer this use question: {question}
    """

    headers = {
        "Content-Type": "application/json",
        "Authorization": f"Bearer {api_key}"
    }

    data = {
        "model": "gpt-3.5-turbo",
        "messages": [{"role": "user", "content": prompt}]
    }

    response = requests.post(url, headers=headers, json=data)
    generated_text = response.json()['choices'][0]['message']['content']

    return generated_text

import os
import requests
import pandas as pd

def text2vec(query):
    headers = {
        'Content-Type': 'application/json',
        'Authorization': 'Bearer ' + "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",
    }

    json_data = {
        'input': query,
        'model': 'text-embedding-ada-002',
    }

    response = requests.post('https://api.openai.com/v1/embeddings', headers=headers, json=json_data)
    query = response.json()['data'][0]['embedding'] #len=1536 #pricing=0.0004
    return query

import pandas as pd
from sentence_transformers import SentenceTransformer, util

df = pd.read_parquet('df.parquet')
df_qa = pd.read_parquet('df_qa.parquet')

df_qa_ = df_qa.copy()
df_ = df.copy()

def qa(df_, df_qa_, min_qa_score, min_context_score, verbose, query):
    query_vec = text2vec(query)

    #first check if there is already a question in df_qa
    df_qa_['score'] = df_qa_['text_vector_'].apply(lambda x : float(util.cos_sim(x, query_vec)))
    df_qa_ = df_qa_.sort_values('score', ascending=False)
    df_qa_ = df_qa_[df_qa_['score']>=min_qa_score]
    #if we find at least one possible preset answer
    if len(df_qa_) > 0:
        if verbose : display(df_qa_)
        answer = df_qa_[0:1]['answer'].values.tolist()[0]
        return answer
    
    #then check if we can use the context to answer a question
    df_['score'] = df_['text_vector_'].apply(lambda x : float(util.cos_sim(x, query_vec)))
    df_ = df_.sort_values('score', ascending=False)
    df_ = df_[df_['score']>=min_context_score]
    #if we find at least one possible preset answer
    if len(df_) > 0:
        if verbose : display(df_)
        #in case we might decide to merge multiple context
        context = ' '.join(df_['description'][0:1].values.tolist())
        answer = chatgpt3_question(context, query)
        return answer
    else:
        return 'impossible to give an answer'

# print(
#     qa(
#         df_, 
#         df_qa_, 
#         min_qa_score=0.92, 
#         min_context_score=.75, 
#         verbose=False, 
#         query='What is a recommender system?'
#     )
# )

import subprocess
import random
import gradio as gr
import requests

history = None
history_prompt = None

def predict(input, history):
    #WE CAN PLAY WITH user_input AND bot_answer, as well as history
    user_input = input

    global history_prompt
    global block_predict

    bot_answer = qa(
        df_, 
        df_qa_, 
        min_qa_score=0.92, 
        min_context_score=.75, 
        verbose=False, 
        query=input
    )

    response = list()
    response = [(input, bot_answer)]
    
    history.append(response[0])
    response = history

    # print('#history', history)
    # print('#response', response)

    return response, history

demo = gr.Blocks()
with demo:
    gr.Markdown(
    """
        Chatbot
    """
    )
    state = gr.Variable(value=[]) #beginning
    chatbot = gr.Chatbot() #color_map=("#00ff7f", "#00d5ff")
    text = gr.Textbox(
        label="Question",
        value="What is a recommendation system?",
        placeholder="",
        max_lines=1,
    )
    text.submit(predict, [text, state], [chatbot, state])
    text.submit(lambda x: "", text, text)
    # btn = gr.Button(value="submit")
    # btn.click(chatbot_foo, None, [chatbot, state])

demo.launch(share=False)