LLM 企業客戶服務應用與最佳化

if hf == True:
    from grequests import download
    directory = "Chapter05"
    filename = "human_feedback.txt"
    download(directory, filename, private_token)
    
    # 檢查檔案是否存在
    efile = os.path.exists('human_feedback.txt')
    if efile:
        # 讀取並清理檔案內容
        with open('human_feedback.txt', 'r') as file:
            content = file.read().replace('\n', ' ').replace('#', '')
            text_input = content
            print(text_input)
    else:
        print("檔案未找到")
        hf = False

一個大語言模型（LLM）是一種先進的AI系統，透過在大量文字資料上的訓練，能夠生成類別似人類的文字回應。它能夠根據訓練資料中的模式和資訊理解語言，對於各種根據語言的任務非常有效，如回答問題、提供建議和促進對話。

if ranking >= 5:
    max_words = 100  # 限制資料大小
    rdata = process_query(user_input, max_words)
    print(rdata)  # 用於維護
    if rdata:
        rdata_clean = rdata.replace('\n', ' ').replace('#', '')
        rdata_sentences = rdata_clean.split('. ')
        text_input = rdata
        print(text_input)

!pip install openai==1.40.3

# 檢索API金鑰
from google.colab import drive
drive.mount('/content/drive')
f = open("drive/MyDrive/files/api_key.txt", "r")
API_KEY = f.readline().strip()
f.close()

import os
import openai
os.environ['OPENAI_API_KEY'] = API_KEY
openai.api_key = os.getenv("OPENAI_API_KEY")

import openai
from openai import OpenAI
import time
client = OpenAI()
gpt_model = "gpt-4o"
start_time = time.time()  # 在請求前開始計時

def call_gpt4_with_full_text(input_text):
    # 將所有行組合成一個字串
    text_input = '\n'.join(input_text)
    prompt = f"請總結或闡述以下內容：{text_input}"
    try:
        response = client.chat.completions.create(
            model=gpt_model,
            messages=[
                {"role": "system", "content": "你是一位專家"},
                {"role": "assistant", "content": "1. 你可以解釋相關內容"},
                {"role": "user", "content": prompt}
            ],
            temperature=0.1  # 在此新增溫度引數
        )
        return response.choices[0].message.content.strip()
    except Exception as e:
        return str(e)

import textwrap
def print_formatted_response(response):
    # 定義文字包裝的寬度
    wrapper = textwrap.TextWrapper(width=80)  # 設定為 80 欄
    wrapped_text = wrapper.fill(text=response)
    # 列印格式化的回應，包含標頭和標尾
    print("GPT-4 回應:")
    print("---------------")
    print(wrapped_text)
    print("---------------\n")
# 假設 'gpt4_response' 包含前一步驟的回應
print_formatted_response(gpt4_response)

import time
...
start_time = time.time()  # 在請求前開始計時
...
response_time = time.time() - start_time  # 測量回應時間
print(f"回應時間：{response_time:.2f} 秒")  # 列印回應時間

from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.metrics.pairwise import cosine_similarity
def calculate_cosine_similarity(text1, text2):
    vectorizer = TfidfVectorizer()
    tfidf = vectorizer.fit_transform([text1, text2])
    similarity = cosine_similarity(tfidf[0:1], tfidf[1:2])
    return similarity[0][0]
# 使用範例
similarity_score = calculate_cosine_similarity(text_input, gpt4_response)
print(f"餘弦相似度分數：{similarity_score:.3f}")

# 評分引數
counter = 20  # 回饋查詢數量
score_history = 30  # 人工回饋總分
threshold = 4  # 觸發人工專家回饋的最低評分門檻

def evaluate_response(response):
    # 評分邏輯
    while True:
        try:
            score = int(input("請輸入評分（1-5）："))
            if 1 <= score <= 5:
                return score
            else:
                print("請輸入有效的評分（1-5）。")
        except ValueError:
            print("無效輸入，請重新輸入。")

import numpy as np

def evaluate_response(response):
    print("\nGenerated Response:")
    print(response)
    print("\nPlease evaluate the response based on the following criteria:")
    print("1 - Poor, 2 - Fair, 3 - Good, 4 - Very Good, 5 - Excellent")
    score = input("Enter the relevance and coherence score (1-5): ")
    try:
        score = int(score)
        if 1 <= score <= 5:
            return score
        else:
            print("Invalid score. Please enter a number between 1 and 5.")
            return evaluate_response(response)  # 遞迴呼叫
    except ValueError:
        print("Invalid input. Please enter a numerical value.")
        return evaluate_response(response)  # 遞迴呼叫

# 呼叫函式
score = evaluate_response(gpt4_response)
print("Evaluator Score:", score)

counter += 1
score_history += score
mean_score = round(np.mean(score_history / counter), 2)
if counter > 0:
    print("Rankings:", counter)
    print("Score history:", mean_score)

threshold = 4
counter_threshold = 10
score_threshold = 4

if counter > counter_threshold and mean_score <= score_threshold:
    print("Human expert evaluation is required for the feedback loop.")

import base64
from google.colab import output
from IPython.display import display, HTML

def image_to_data_uri(file_path):
    with open(file_path, 'rb') as image_file:
        encoded_string = base64.b64encode(image_file.read()).decode()
    return f'data:image/png;base64,{encoded_string}'

thumbs_up_data_uri = image_to_data_uri('/content/thumbs_up.png')
thumbs_down_data_uri = image_to_data_uri('/content/thumbs_down.png')

def display_icons():
    # 定義包含兩個可點選圖片的HTML內容
    html_content = f"""
    <img src="{thumbs_up_data_uri}" onclick="alert('Thumbs up clicked!')">
    <img src="{thumbs_down_data_uri}" onclick="prompt('Enter feedback:')">
    """
    display(HTML(html_content))

def save_feedback(feedback):
    with open('/content/expert_feedback.txt', 'w') as f:
        f.write(feedback)
    print("Feedback saved successfully.")

# 註冊回撥函式
output.register_callback('notebook.save_feedback', save_feedback)
print("Human Expert Adaptive RAG activated")
display_icons()