更新專案

app.py

@@ -1,33 +1,39 @@
 # app.py
 # -------------------------------
-# 套件
+# 1. 套件載入
 # -------------------------------
 import os, glob, requests
 from langchain.docstore.document import Document
 from langchain.text_splitter import RecursiveCharacterTextSplitter
 from langchain.vectorstores import FAISS
 from langchain.chains import RetrievalQA
-from langchain.llms import HuggingFaceHub
+from langchain_huggingface import HuggingFaceHub  # <-- 正確的 Import
+from langchain.embeddings import HuggingFaceEmbeddings
 from docx import Document as DocxDocument
 import gradio as gr
 
 # -------------------------------
-# 資料路徑
+# 2. 環境變數與資料路徑
 # -------------------------------
 TXT_FOLDER = "./out_texts"
 DB_PATH = "./faiss_db"
 os.makedirs(DB_PATH, exist_ok=True)
 
+HF_TOKEN = os.environ.get("HUGGINGFACEHUB_API_TOKEN")
+if not HF_TOKEN:
+    raise ValueError("請在 Hugging Face Space 的 Settings → Repository secrets 設定 HUGGINGFACEHUB_API_TOKEN")
+
 # -------------------------------
-# 向量資料庫
+# 3. 建立或載入向量資料庫
 # -------------------------------
 EMBEDDINGS_MODEL_NAME = "sentence-transformers/all-MiniLM-L6-v2"
-from langchain.embeddings import HuggingFaceEmbeddings
 embeddings_model = HuggingFaceEmbeddings(model_name=EMBEDDINGS_MODEL_NAME)
 
 if os.path.exists(os.path.join(DB_PATH, "index.faiss")):
+    print("載入現有向量資料庫...")
     db = FAISS.load_local(DB_PATH, embeddings_model, allow_dangerous_deserialization=True)
 else:
+    print("沒有資料庫，開始建立新向量資料庫...")
     txt_files = glob.glob(f"{TXT_FOLDER}/*.txt")
     docs = []
     for filepath in txt_files:
@@ -38,17 +44,15 @@ else:
     db = FAISS.from_documents(split_docs, embeddings_model)
     db.save_local(DB_PATH)
 
-retriever = db.as_retriever(search_type="similarity", search_kwargs={"k":5})
+retriever = db.as_retriever(search_type="similarity", search_kwargs={"k": 5})
 
 # -------------------------------
-# Hugging Face Hub LLM
+# 4. LLM 設定（Hugging Face Hub）
 # -------------------------------
-HUGGINGFACEHUB_API_TOKEN = os.environ.get("HUGGINGFACEHUB_API_TOKEN")
-
 llm = HuggingFaceHub(
     repo_id="google/flan-t5-large",
-    model_kwargs={"temperature":0.7, "max_new_tokens":512},
-    huggingfacehub_api_token=HUGGINGFACEHUB_API_TOKEN
+    model_kwargs={"temperature": 0.7, "max_new_tokens": 512},
+    huggingfacehub_api_token=HF_TOKEN
 )
 
 qa_chain = RetrievalQA.from_chain_type(
@@ -58,15 +62,14 @@ qa_chain = RetrievalQA.from_chain_type(
 )
 
 # -------------------------------
-# 查剩餘額度
+# 5. 查詢 API 剩餘額度
 # -------------------------------
 def get_hf_rate_limit():
-    headers = {"Authorization": f"Bearer {HUGGINGFACEHUB_API_TOKEN}"}
+    headers = {"Authorization": f"Bearer {HF_TOKEN}"}
     try:
         r = requests.get("https://huggingface.co/api/whoami", headers=headers)
         r.raise_for_status()
         data = r.json()
-        # free plan 每小時 300 次
         used = data.get("rate_limit", {}).get("used", 0)
         remaining = 300 - used if used is not None else "未知"
         return f"本小時剩餘 API 次數：約 {remaining}"
@@ -74,16 +77,16 @@ def get_hf_rate_limit():
         return "無法取得 API 速率資訊"
 
 # -------------------------------
-# 文章生成
+# 6. 生成文章
 # -------------------------------
 def generate_article_with_rate(query, segments=5):
     docx_file = "/tmp/generated_article.docx"
     doc = DocxDocument()
     doc.add_heading(query, level=1)
-    
+
     all_text = []
     prompt = f"請依據下列主題生成段落：{query}\n\n每段約150-200字。"
-    
+
     for i in range(int(segments)):
         try:
             result = qa_chain({"query": prompt})
@@ -95,17 +98,16 @@ def generate_article_with_rate(query, segments=5):
         all_text.append(paragraph)
         doc.add_paragraph(paragraph)
         prompt = f"請接續上一段生成下一段：\n{paragraph}\n\n下一段："
-    
+
     doc.save(docx_file)
     full_text = "\n\n".join(all_text)
-    
+
     # 取得 API 剩餘次數
     rate_info = get_hf_rate_limit()
-    
     return f"{rate_info}\n\n{full_text}", docx_file
 
 # -------------------------------
-# Gradio 介面
+# 7. Gradio 介面
 # -------------------------------
 iface = gr.Interface(
     fn=generate_article_with_rate,
@@ -121,4 +123,6 @@ iface = gr.Interface(
     description="使用 Hugging Face Hub LLM + FAISS RAG，生成文章並提示 API 剩餘額度。"
 )
 
-iface.launch()
+if __name__ == "__main__":
+    iface.launch()
+