utils/gpt_processor.py

import re
import os
import logging
from typing import List

from opencc import OpenCC

import openai
import tiktoken

OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")


class GPTAgent:
    def __init__(self, model):
        openai.api_key = OPENAI_API_KEY
        self.model = model
        self.temperature = 0.8
        self.frequency_penalty = 0
        self.presence_penalty = 0.6
        self.max_tokens = 2048
        self.split_max_tokens = 13000

    def request(self, messages):
        response = self.agent.complete(messages=messages)
        return response.choices[0].message["content"]

    def split_into_many(self, text) -> List[str]:
        tokenizer = tiktoken.get_encoding("cl100k_base")
        # Split the text into sentences
        sentences = text.split("。")

        # Get the number of tokens for each sentence
        n_tokens = [len(tokenizer.encode(" " + sentence)) for sentence in sentences]

        chunks = []
        tokens_so_far = 0
        chunk = []

        # Loop through the sentences and tokens joined together in a tuple
        for sentence, token in zip(sentences, n_tokens):
            # If the number of tokens so far plus the number of tokens in the current sentence is greater
            # than the max number of tokens, then add the chunk to the list of chunks and reset
            # the chunk and tokens so far
            if tokens_so_far + token > self.split_max_tokens:
                chunks.append("。".join(chunk) + "。")
                chunk = []
                tokens_so_far = 0

            # If the number of tokens in the current sentence is greater than the max number of
            # tokens, go to the next sentence
            if token > self.split_max_tokens:
                continue

            # Otherwise, add the sentence to the chunk and add the number of tokens to the total
            chunk.append(sentence)
            tokens_so_far += token + 1

        # if the length of the text is less than the max number of tokens, then return the text
        return [text] if len(chunks) == 0 else chunks

    def preprocess(self, text):
        text = text.replace("\n", " ").replace("\r", "")
        return text

    def parse_result(self, result):
        parsed_result = []
        chinese_converter = OpenCC("s2tw")
        for i in range(len(result)):
            result[i] = result[i].split(",")
            if len(result[i]) == 1:
                result[i] = result[i][0].split("、")
            if len(result[i]) == 1:
                result[i] = result[i][0].split("，")
            for word in result[i]:
                try:
                    parsed_result.append(
                        chinese_converter.convert(word).strip().replace("。", "")
                    )
                except Exception as e:
                    logging.error(e)
                    logging.error("Failed to parse result")
        return parsed_result


class Translator(GPTAgent):
    def __init__(self):
        super().__init__("gpt-3.5-turbo")

    def translate_to_chinese(self, text):
        system_prompt = """
            I want you to act as an Chinese translator, spelling corrector and improver. 
            I will speak to you in English, translate it and answer in the corrected and improved version of my text, in Traditional Chinese.  
            Keep the meaning same, but make them more literary. I want you to only reply the correction, the improvements and nothing else, do not write explanations and DO NOT use any Simplified Chinese. 
        """
        system_prompt_zh_tw = """
            我希望你擔任中文翻譯、拼寫糾正及改進的角色。
            我將用英文與你交流，請將其翻譯並用繁體中文回答，同時對我的文本進行糾正和改進。
            保持原意不變，但使其更具文學性。我希望你僅回覆更正、改進的部分，不要寫解釋，也不要使用任何简体中文。
        """
        messages = [
            {"role": "system", "content": f"{system_prompt_zh_tw}"},
            {"role": "user", "content": text},
        ]
        try:
            response = openai.ChatCompletion.create(
                model=self.model,
                messages=messages,
                temperature=self.temperature,
                frequency_penalty=self.frequency_penalty,
                presence_penalty=self.presence_penalty,
            )
        except Exception as e:
            logging.error(e)
            logging.error("Failed to translate to Chinese")

        # translate from simplified chinese to traditional chinese
        chinese_converter = OpenCC("s2tw")
        return chinese_converter.convert(
            response["choices"][0]["message"]["content"].strip()
        )


class EmbeddingGenerator(GPTAgent):
    def __init__(self):
        super().__init__("text-davinci-002")

    def get_embedding(self, text):
        return openai.Embedding.create(input=text, engine="text-embedding-ada-002")[
            "data"
        ][0]["embedding"]


class KeywordsGenerator(GPTAgent):
    def __init__(self):
        super().__init__("gpt-3.5-turbo")

    def extract_keywords(self, text):
        system_prompt = """
            請你為以下內容抓出 5 個關鍵字用以搜尋這篇文章，並用「,」來分隔
        """
        text_chunks = self.split_into_many(text)
        keywords = []
        for i in range(len(text_chunks)):
            text = text_chunks[i]
            messages = [
                {"role": "system", "content": f"{system_prompt}"},
                {"role": "user", "content": f"{self.preprocess(text)}"},
            ]
            try:
                response = openai.ChatCompletion.create(
                    model=self.model,
                    messages=messages,
                    temperature=0,
                    max_tokens=self.max_tokens,
                    frequency_penalty=self.frequency_penalty,
                    presence_penalty=self.presence_penalty,
                )
                keywords.append(response["choices"][0]["message"]["content"].strip())
            except Exception as e:
                logging.error(e)
                logging.error("Failed to extract keywords")
        return self.parse_result(keywords)


class TopicsGenerator(GPTAgent):
    def __init__(self):
        super().__init__("gpt-3.5-turbo")

    def extract_topics(self, text):
        system_prompt = """
            請你為以下內容給予 3 個高度抽象的主題分類這篇文章，並用「,」來分隔
        """
        text_chunks = self.split_into_many(text)
        topics = []
        for i in range(len(text_chunks)):
            text = text_chunks[i]
            messages = [
                {"role": "system", "content": f"{system_prompt}"},
                {"role": "user", "content": f"{self.preprocess(text)}"},
            ]
            try:
                response = openai.ChatCompletion.create(
                    model=self.model,
                    messages=messages,
                    temperature=0,
                    max_tokens=self.max_tokens,
                    frequency_penalty=self.frequency_penalty,
                    presence_penalty=self.presence_penalty,
                )
                topics.append(response["choices"][0]["message"]["content"].strip())
            except Exception as e:
                logging.error(e)
                logging.error("Failed to extract topics")
        return self.parse_result(topics)


class Summarizer(GPTAgent):
    def __init__(self):
        super().__init__("gpt-3.5-turbo-16k")

    def summarize(self, text):
        system_prompt = """
            請幫我總結以下的文章。  
        """
        messages = [
            {"role": "system", "content": f"{system_prompt}"},
            {"role": "user", "content": text},
        ]
        try:
            response = openai.ChatCompletion.create(
                model=self.model,
                messages=messages,
                temperature=self.temperature,
                max_tokens=self.max_tokens,
                frequency_penalty=self.frequency_penalty,
                presence_penalty=self.presence_penalty,
            )
        except Exception as e:
            logging.error(e)
            logging.error("Failed to summarize")
        chinese_converter = OpenCC("s2tw")
        print(f'the summary is {response["choices"][0]["message"]["content"].strip()}')
        response = chinese_converter.convert(
            response["choices"][0]["message"]["content"]
        )

        return re.sub(r"\n+", "<br>", response)


class QuestionAnswerer(GPTAgent):
    def __init__(self):
        super().__init__("gpt-3.5-turbo-16k")

    def answer_chunk_question(self, text, question):
        system_prompt = """
            你是一個知識檢索系統，我會給你一份文件，請幫我依照文件內容回答問題，並用繁體中文回答。以下是文件內容
        """
        text_chunks = self.split_into_many(text)
        answer_chunks = []
        for i in range(len(text_chunks)):
            text = text_chunks[i]
            messages = [
                {"role": "system", "content": f"{system_prompt} + '\n' '{text}'"},
                {"role": "user", "content": f"{question}"},
            ]
            try:
                response = openai.ChatCompletion.create(
                    model=self.model,
                    messages=messages,
                    temperature=self.temperature,
                    max_tokens=1024,
                    frequency_penalty=self.frequency_penalty,
                    presence_penalty=self.presence_penalty,
                )
            except Exception as e:
                logging.error(e)
                logging.error("Failed to answer question")
            chinese_converter = OpenCC("s2tw")
            answer_chunks.append(
                chinese_converter.convert(
                    response["choices"][0]["message"]["content"].strip()
                )
            )

        return "。".join(answer_chunks)

    def answer_question(self, context, context_page_num, context_file_name, history):
        system_prompt = """
            你是一個知識檢索系統，我會給你一份文件，請幫我依照文件內容回答問題，並用繁體中文回答。以下是文件內容
        """

        history = self.__construct_message_history(history)
        messages = [
            {"role": "system", "content": f"{system_prompt} + '\n' '''{context}'''"},
        ] + history
        try:
            response = openai.ChatCompletion.create(
                model=self.model,
                messages=messages,
                temperature=self.temperature,
                max_tokens=2048,
                frequency_penalty=self.frequency_penalty,
                presence_penalty=self.presence_penalty,
            )
            chinese_converter = OpenCC("s2tw")
            page_num_message = f"以下內容來自 {context_file_name}，第 {context_page_num} 頁\n\n"
            bot_answer = response["choices"][0]["message"]["content"]
            whole_answer = page_num_message + bot_answer

            return chinese_converter.convert(whole_answer)
        except Exception as e:
            logging.error(e)
            logging.error("Failed to answer question")

    def __construct_message_history(self, history):
        print(f"history is {history}")
        max_history_length = 10
        if len(history) > max_history_length:
            history = history[-max_history_length:]

        messages = []
        for i in range(len(history)):
            messages.append({"role": "user", "content": history[i][0]})
            if history[i][1] is not None:
                messages.append({"role": "assistant", "content": history[i][1]})

        return messages