Upload 2 files
Browse files- app (3).py +293 -0
- requirements (4).txt +9 -0
app (3).py
ADDED
@@ -0,0 +1,293 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# app.py
|
2 |
+
|
3 |
+
import os
|
4 |
+
import openai
|
5 |
+
import gradio as gr
|
6 |
+
from langchain.chains import ConversationalRetrievalChain
|
7 |
+
from langchain.text_splitter import CharacterTextSplitter
|
8 |
+
from langchain_community.document_loaders import PyMuPDFLoader, PyPDFLoader
|
9 |
+
from langchain.vectorstores import Chroma
|
10 |
+
from langchain_community.embeddings import OpenAIEmbeddings
|
11 |
+
from langchain_community.chat_models import ChatOpenAI
|
12 |
+
import shutil # 用於文件複製
|
13 |
+
import logging
|
14 |
+
|
15 |
+
# 設置日誌配置
|
16 |
+
logging.basicConfig(level=logging.INFO)
|
17 |
+
logger = logging.getLogger(__name__)
|
18 |
+
|
19 |
+
# 獲取 OpenAI API 密鑰(初始不使用固定密鑰)
|
20 |
+
api_key_env = os.getenv("OPENAI_API_KEY")
|
21 |
+
if api_key_env:
|
22 |
+
openai.api_key = api_key_env
|
23 |
+
logger.info("OpenAI API 密鑰已設置。")
|
24 |
+
else:
|
25 |
+
logger.info("未設置固定的 OpenAI API 密鑰。將使用使用者提供的密鑰。")
|
26 |
+
|
27 |
+
# 確保向量資料庫目錄存在且有寫入權限
|
28 |
+
VECTORDB_DIR = os.path.abspath("./data")
|
29 |
+
os.makedirs(VECTORDB_DIR, exist_ok=True)
|
30 |
+
os.chmod(VECTORDB_DIR, 0o755) # 設置適當的權限
|
31 |
+
logger.info(f"VECTORDB_DIR set to: {VECTORDB_DIR}")
|
32 |
+
|
33 |
+
# 定義測試 PDF 加載器的函數
|
34 |
+
def test_pdf_loader(file_path, loader_type='PyMuPDFLoader'):
|
35 |
+
logger.info(f"Testing PDF loader ({loader_type}) with file: {file_path}")
|
36 |
+
try:
|
37 |
+
if loader_type == 'PyMuPDFLoader':
|
38 |
+
loader = PyMuPDFLoader(file_path)
|
39 |
+
elif loader_type == 'PyPDFLoader':
|
40 |
+
loader = PyPDFLoader(file_path)
|
41 |
+
else:
|
42 |
+
logger.error(f"Unknown loader type: {loader_type}")
|
43 |
+
return
|
44 |
+
loaded_docs = loader.load()
|
45 |
+
if loaded_docs:
|
46 |
+
logger.info(f"Successfully loaded {file_path} with {len(loaded_docs)} documents.")
|
47 |
+
logger.info(f"Document content (first 500 chars): {loaded_docs[0].page_content[:500]}")
|
48 |
+
else:
|
49 |
+
logger.error(f"No documents loaded from {file_path}.")
|
50 |
+
except Exception as e:
|
51 |
+
logger.error(f"Error loading {file_path} with {loader_type}: {e}")
|
52 |
+
|
53 |
+
# 定義載入和處理 PDF 文件的函數
|
54 |
+
def load_and_process_documents(file_paths, loader_type='PyMuPDFLoader', api_key=None):
|
55 |
+
if not api_key:
|
56 |
+
raise ValueError("未提供 OpenAI API 密鑰。")
|
57 |
+
documents = []
|
58 |
+
logger.info("開始載入上傳的 PDF 文件。")
|
59 |
+
|
60 |
+
for file_path in file_paths:
|
61 |
+
logger.info(f"載入 PDF 文件: {file_path}")
|
62 |
+
if not os.path.exists(file_path):
|
63 |
+
logger.error(f"文件不存在: {file_path}")
|
64 |
+
continue
|
65 |
+
try:
|
66 |
+
if loader_type == 'PyMuPDFLoader':
|
67 |
+
loader = PyMuPDFLoader(file_path)
|
68 |
+
elif loader_type == 'PyPDFLoader':
|
69 |
+
loader = PyPDFLoader(file_path)
|
70 |
+
else:
|
71 |
+
logger.error(f"Unknown loader type: {loader_type}")
|
72 |
+
continue
|
73 |
+
loaded_docs = loader.load()
|
74 |
+
if loaded_docs:
|
75 |
+
logger.info(f"載入 {file_path} 成功,包含 {len(loaded_docs)} 個文檔。")
|
76 |
+
# 打印第一個文檔的部分內容以確認
|
77 |
+
logger.info(f"第一個文檔內容: {loaded_docs[0].page_content[:500]}")
|
78 |
+
documents.extend(loaded_docs)
|
79 |
+
else:
|
80 |
+
logger.error(f"載入 {file_path} 但未找到任何文檔。")
|
81 |
+
except Exception as e:
|
82 |
+
logger.error(f"載入 {file_path} 時出現錯誤: {e}")
|
83 |
+
|
84 |
+
if not documents:
|
85 |
+
raise ValueError("沒有找到任何 PDF 文件或 PDF 文件無法載入。")
|
86 |
+
else:
|
87 |
+
logger.info(f"總共載入了 {len(documents)} 個文檔。")
|
88 |
+
|
89 |
+
# 分割長文本
|
90 |
+
text_splitter = CharacterTextSplitter(chunk_size=1000, chunk_overlap=50)
|
91 |
+
documents = text_splitter.split_documents(documents)
|
92 |
+
logger.info(f"分割後的文檔數量: {len(documents)}")
|
93 |
+
|
94 |
+
if not documents:
|
95 |
+
raise ValueError("分割後的文檔列表為空。請檢查 PDF 文件內容。")
|
96 |
+
|
97 |
+
# 初始化向量資料庫
|
98 |
+
try:
|
99 |
+
embeddings = OpenAIEmbeddings(openai_api_key=api_key) # 使用使用者的 API 密鑰
|
100 |
+
logger.info("初始化 OpenAIEmbeddings 成功。")
|
101 |
+
except Exception as e:
|
102 |
+
raise ValueError(f"初始化 OpenAIEmbeddings 時出現錯誤: {e}")
|
103 |
+
|
104 |
+
try:
|
105 |
+
vectordb = Chroma.from_documents(
|
106 |
+
documents,
|
107 |
+
embedding=embeddings,
|
108 |
+
persist_directory=VECTORDB_DIR
|
109 |
+
)
|
110 |
+
logger.info("初始化 Chroma 向量資料庫成功。")
|
111 |
+
except Exception as e:
|
112 |
+
raise ValueError(f"初始化 Chroma 向量資料庫時出現錯誤: {e}")
|
113 |
+
|
114 |
+
return vectordb
|
115 |
+
|
116 |
+
# 定義聊天處理函數
|
117 |
+
def handle_query(user_message, chat_history, vectordb, api_key):
|
118 |
+
try:
|
119 |
+
if not user_message:
|
120 |
+
return chat_history
|
121 |
+
|
122 |
+
# 添加角色指令前綴
|
123 |
+
preface = """
|
124 |
+
指令: 以繁體中文回答問題,200字以內。你是一位專業心理學家與調酒師,專精於 MBTI 人格與經典調酒主題。
|
125 |
+
非相關問題,請回應:「目前僅支援 MBTI 分析與經典調酒主題。」。
|
126 |
+
"""
|
127 |
+
query = f"{preface} 查詢內容:{user_message}"
|
128 |
+
|
129 |
+
# 初始化 ConversationalRetrievalChain,並傳遞 openai_api_key
|
130 |
+
pdf_qa = ConversationalRetrievalChain.from_llm(
|
131 |
+
ChatOpenAI(temperature=0.7, model="gpt-4", openai_api_key=api_key),
|
132 |
+
retriever=vectordb.as_retriever(search_kwargs={'k': 6}),
|
133 |
+
return_source_documents=True
|
134 |
+
)
|
135 |
+
|
136 |
+
# 呼叫模型並處理查詢
|
137 |
+
result = pdf_qa.invoke({"question": query, "chat_history": chat_history})
|
138 |
+
|
139 |
+
# 檢查結果並更新聊天歷史
|
140 |
+
if "answer" in result:
|
141 |
+
chat_history = chat_history + [(user_message, result["answer"])]
|
142 |
+
else:
|
143 |
+
chat_history = chat_history + [(user_message, "抱歉,未能獲得有效回應。")]
|
144 |
+
return chat_history
|
145 |
+
|
146 |
+
except Exception as e:
|
147 |
+
logger.error(f"Error in handle_query: {e}")
|
148 |
+
return chat_history + [("系統", f"出現錯誤: {str(e)}")]
|
149 |
+
|
150 |
+
# 定義保存 API 密鑰的函數
|
151 |
+
def save_api_key(api_key, state):
|
152 |
+
if not api_key.startswith("sk-"):
|
153 |
+
return "請輸入有效的 OpenAI API 密鑰。", state
|
154 |
+
state['api_key'] = api_key
|
155 |
+
logger.info("使用者已保存自己的 OpenAI API 密鑰。")
|
156 |
+
return "API 密鑰已成功保存。您現在可以上傳 PDF 文件並開始提問。", state
|
157 |
+
|
158 |
+
# 定義 Gradio 的處理函數
|
159 |
+
def process_files(files, state):
|
160 |
+
logger.info("process_files called")
|
161 |
+
if files:
|
162 |
+
try:
|
163 |
+
# 檢查是否已保存 API 密鑰
|
164 |
+
api_key = state.get('api_key', None)
|
165 |
+
if not api_key:
|
166 |
+
logger.error("使用者未提供 OpenAI API 密鑰。")
|
167 |
+
return "請先在「設定 API 密鑰」標籤中輸入並保存您的 OpenAI API 密鑰。", state
|
168 |
+
|
169 |
+
logger.info(f"Received {len(files)} files")
|
170 |
+
saved_file_paths = []
|
171 |
+
for idx, file_data in enumerate(files):
|
172 |
+
# 為每個文件分配唯一的文件名
|
173 |
+
filename = f"uploaded_{idx}.pdf"
|
174 |
+
save_path = os.path.join(VECTORDB_DIR, filename)
|
175 |
+
with open(save_path, "wb") as f:
|
176 |
+
f.write(file_data)
|
177 |
+
# 確認文件是否存在並檢查文件大小
|
178 |
+
if os.path.exists(save_path):
|
179 |
+
file_size = os.path.getsize(save_path)
|
180 |
+
if file_size > 0:
|
181 |
+
logger.info(f"File successfully saved to: {save_path} (Size: {file_size} bytes)")
|
182 |
+
else:
|
183 |
+
logger.error(f"File saved to {save_path} is empty.")
|
184 |
+
raise ValueError(f"上傳的文件 {filename} 為空。")
|
185 |
+
else:
|
186 |
+
logger.error(f"Failed to save file to: {save_path}")
|
187 |
+
raise FileNotFoundError(f"無法保存文件到 {save_path}")
|
188 |
+
saved_file_paths.append(save_path)
|
189 |
+
# 測試 PDF 加載器,先用 PyMuPDFLoader,再用 PyPDFLoader
|
190 |
+
try:
|
191 |
+
test_pdf_loader(save_path, loader_type='PyMuPDFLoader')
|
192 |
+
except Exception as e:
|
193 |
+
logger.error(f"PyMuPDFLoader failed: {e}")
|
194 |
+
logger.info("Attempting to load with PyPDFLoader...")
|
195 |
+
test_pdf_loader(save_path, loader_type='PyPDFLoader')
|
196 |
+
# 列出 VECTORDB_DIR 中的所有文件
|
197 |
+
saved_files = os.listdir(VECTORDB_DIR)
|
198 |
+
logger.info(f"Files in VECTORDB_DIR ({VECTORDB_DIR}): {saved_files}")
|
199 |
+
# 列出文件大小
|
200 |
+
file_sizes = {file: os.path.getsize(os.path.join(VECTORDB_DIR, file)) for file in saved_files}
|
201 |
+
logger.info(f"File sizes in VECTORDB_DIR: {file_sizes}")
|
202 |
+
vectordb = load_and_process_documents(saved_file_paths, loader_type='PyMuPDFLoader', api_key=api_key)
|
203 |
+
state['vectordb'] = vectordb
|
204 |
+
return "PDF 文件已成功上傳並處理。您現在可以開始提問。", state
|
205 |
+
except Exception as e:
|
206 |
+
logger.error(f"Error in process_files: {e}")
|
207 |
+
return f"處理文件時出現錯誤: {e}", state
|
208 |
+
else:
|
209 |
+
return "請上傳至少一個 PDF 文件。", state
|
210 |
+
|
211 |
+
def chat_interface(user_message, chat_history, state):
|
212 |
+
vectordb = state.get('vectordb', None)
|
213 |
+
api_key = state.get('api_key', None)
|
214 |
+
if not vectordb:
|
215 |
+
return chat_history, state, "請先上傳 PDF 文件以進行處理。"
|
216 |
+
if not api_key:
|
217 |
+
return chat_history, state, "請先在「設定 API 密鑰」標籤中輸入並保存您的 OpenAI API 密鑰。"
|
218 |
+
|
219 |
+
# 處理查詢
|
220 |
+
updated_history = handle_query(user_message, chat_history, vectordb, api_key)
|
221 |
+
return updated_history, state, ""
|
222 |
+
|
223 |
+
# 設計 Gradio 介面
|
224 |
+
with gr.Blocks() as demo:
|
225 |
+
gr.Markdown("<h1 style='text-align: center;'>MBTI 與經典調酒 AI 助理</h1>")
|
226 |
+
|
227 |
+
# 定義共享的 state
|
228 |
+
state = gr.State({"vectordb": None, "api_key": None})
|
229 |
+
|
230 |
+
with gr.Tab("設定 API 密���"):
|
231 |
+
with gr.Row():
|
232 |
+
with gr.Column(scale=1):
|
233 |
+
api_key_input = gr.Textbox(
|
234 |
+
label="輸入您的 OpenAI API 密鑰",
|
235 |
+
placeholder="sk-...",
|
236 |
+
type="password",
|
237 |
+
interactive=True
|
238 |
+
)
|
239 |
+
save_api_key_btn = gr.Button("保存 API 密鑰")
|
240 |
+
api_key_status = gr.Textbox(label="狀態", interactive=False)
|
241 |
+
|
242 |
+
with gr.Tab("上傳 PDF 文件"):
|
243 |
+
with gr.Row():
|
244 |
+
with gr.Column(scale=1):
|
245 |
+
upload = gr.File(
|
246 |
+
file_count="multiple",
|
247 |
+
file_types=[".pdf"],
|
248 |
+
label="上傳 PDF 文件",
|
249 |
+
interactive=True,
|
250 |
+
type="binary" # 修改為 'binary'
|
251 |
+
)
|
252 |
+
upload_btn = gr.Button("上傳並處理")
|
253 |
+
upload_status = gr.Textbox(label="上傳狀態", interactive=False)
|
254 |
+
|
255 |
+
with gr.Tab("聊天機器人"):
|
256 |
+
chatbot = gr.Chatbot()
|
257 |
+
|
258 |
+
with gr.Row():
|
259 |
+
with gr.Column(scale=0.85):
|
260 |
+
txt = gr.Textbox(show_label=False, placeholder="請輸入您的問題...")
|
261 |
+
with gr.Column(scale=0.15, min_width=0):
|
262 |
+
submit_btn = gr.Button("提問")
|
263 |
+
|
264 |
+
# 綁定提問按鈕
|
265 |
+
submit_btn.click(
|
266 |
+
chat_interface,
|
267 |
+
inputs=[txt, chatbot, state],
|
268 |
+
outputs=[chatbot, state, txt]
|
269 |
+
)
|
270 |
+
|
271 |
+
# 綁定輸入框的提交事件
|
272 |
+
txt.submit(
|
273 |
+
chat_interface,
|
274 |
+
inputs=[txt, chatbot, state],
|
275 |
+
outputs=[chatbot, state, txt]
|
276 |
+
)
|
277 |
+
|
278 |
+
# 綁定保存 API 密鑰按鈕
|
279 |
+
save_api_key_btn.click(
|
280 |
+
save_api_key,
|
281 |
+
inputs=[api_key_input, state],
|
282 |
+
outputs=[api_key_status, state]
|
283 |
+
)
|
284 |
+
|
285 |
+
# 綁定上傳按鈕
|
286 |
+
upload_btn.click(
|
287 |
+
process_files,
|
288 |
+
inputs=[upload, state],
|
289 |
+
outputs=[upload_status, state]
|
290 |
+
)
|
291 |
+
|
292 |
+
# 啟動 Gradio 應用
|
293 |
+
demo.launch()
|
requirements (4).txt
ADDED
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
gradio
|
2 |
+
langchain
|
3 |
+
langchain-community
|
4 |
+
chromadb
|
5 |
+
pymupdf
|
6 |
+
python-dotenv
|
7 |
+
tiktoken
|
8 |
+
openai==0.27.8
|
9 |
+
PyPDF2
|