Update app.py

app.py

@@ -17,6 +17,7 @@ from langchain_community.vectorstores import FAISS
 TXT_FOLDER = "./out_texts"
 DB_PATH = "./faiss_db"
 os.makedirs(DB_PATH, exist_ok=True)
+os.makedirs(TXT_FOLDER, exist_ok=True)  # 避免沒有 txt 檔時錯誤
 
 HF_TOKEN = os.environ.get("HUGGINGFACEHUB_API_TOKEN")
 if not HF_TOKEN:
@@ -39,9 +40,7 @@ else:
     docs = []
     for filepath in txt_files:
         with open(filepath, "r", encoding="utf-8") as f:
-            docs.append(
-                Document(page_content=f.read(), metadata={"source": os.path.basename(filepath)})
-            )
+            docs.append(Document(page_content=f.read(), metadata={"source": os.path.basename(filepath)}))
     splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=100)
     split_docs = splitter.split_documents(docs)
     db = FAISS.from_documents(split_docs, embeddings_model)
@@ -54,7 +53,7 @@ retriever = db.as_retriever(search_type="similarity", search_kwargs={"k": 5})
 # -------------------------------
 llm = HuggingFaceEndpoint(
     repo_id="google/flan-t5-large",
-    task="text2text-generation",   # 明確指定 task
+    task="text2text-generation",
     huggingfacehub_api_token=HF_TOKEN,
     model_kwargs={"temperature": 0.7, "max_new_tokens": 512},
 )
@@ -74,10 +73,9 @@ def get_hf_rate_limit():
         r = requests.get("https://huggingface.co/api/whoami", headers=headers)
         r.raise_for_status()
         data = r.json()
-        used = data.get("rate_limit", {}).get("used", 0)
-        remaining = 300 - used if used is not None else "未知"
+        remaining = data.get("rate_limit", {}).get("remaining", "未知")
         return f"本小時剩餘 API 次數：約 {remaining}"
-    except:
+    except Exception:
         return "無法取得 API 速率資訊"
 
 # -------------------------------
@@ -94,7 +92,7 @@ def generate_article_with_rate(query, segments=5):
     for i in range(int(segments)):
         try:
             result = qa_chain({"query": prompt})
-            paragraph = result["result"].strip()
+            paragraph = result.get("result", "").strip()
             if not paragraph:
                 paragraph = "（本段生成失敗，請嘗試減少段落或改用較小模型。）"
         except Exception as e:
@@ -105,26 +103,23 @@ def generate_article_with_rate(query, segments=5):
 
     doc.save(docx_file)
     full_text = "\n\n".join(all_text)
-
     rate_info = get_hf_rate_limit()
     return f"{rate_info}\n\n{full_text}", docx_file
 
 # -------------------------------
 # 7. Gradio 介面
 # -------------------------------
-iface = gr.Interface(
-    fn=generate_article_with_rate,
-    inputs=[
-        gr.Textbox(lines=2, placeholder="請輸入文章主題", label="文章主題"),
-        gr.Slider(minimum=1, maximum=10, step=1, value=5, label="段落數")
-    ],
-    outputs=[
-        gr.Textbox(label="生成文章 + API 剩餘次數"),
-        gr.File(label="下載 DOCX")
-    ],
-    title="佛教經論 RAG 系統 (HF API)",
-    description="使用 Hugging Face Endpoint LLM + FAISS RAG，生成文章並提示 API 剩餘額度。"
-)
+with gr.Blocks() as demo:
+    gr.Markdown("# 佛教經論 RAG 系統 (HF API)")
+    gr.Markdown("使用 Hugging Face Endpoint LLM + FAISS RAG，生成文章並提示 API 剩餘額度。")
+    
+    query_input = gr.Textbox(lines=2, placeholder="請輸入文章主題", label="文章主題")
+    segments_input = gr.Slider(minimum=1, maximum=10, step=1, value=5, label="段落數")
+    output_text = gr.Textbox(label="生成文章 + API 剩餘次數")
+    output_file = gr.File(label="下載 DOCX")
+    
+    query_input.submit(generate_article_with_rate, [query_input, segments_input], [output_text, output_file])
+    segments_input.change(generate_article_with_rate, [query_input, segments_input], [output_text, output_file])
 
 if __name__ == "__main__":
-    iface.launch()
+    demo.launch()