Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
|
@@ -27,14 +27,11 @@ load_dotenv()
|
|
| 27 |
class NotebookLMTool:
|
| 28 |
def __init__(self):
|
| 29 |
self.api_key = os.getenv("GEMINI_API_KEY")
|
| 30 |
-
|
| 31 |
-
if self.api_key:
|
| 32 |
-
self.client = genai.Client(api_key=self.api_key)
|
| 33 |
|
| 34 |
def set_key(self, user_key):
|
| 35 |
if user_key and user_key.strip():
|
| 36 |
self.api_key = user_key.strip()
|
| 37 |
-
self.client = genai.Client(api_key=self.api_key)
|
| 38 |
return "✅ API Key 已更新!"
|
| 39 |
return "⚠️ Key 無效"
|
| 40 |
|
|
@@ -49,15 +46,21 @@ class NotebookLMTool:
|
|
| 49 |
except:
|
| 50 |
return []
|
| 51 |
|
| 52 |
-
def
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 53 |
"""
|
| 54 |
-
封裝 Gemini 呼叫,加入指數退避重試機制
|
| 55 |
"""
|
| 56 |
delay = 5 # 初始等待秒數
|
| 57 |
|
| 58 |
for attempt in range(retries):
|
| 59 |
try:
|
| 60 |
-
response =
|
| 61 |
model=model_name,
|
| 62 |
contents=contents,
|
| 63 |
config=config
|
|
@@ -65,14 +68,14 @@ class NotebookLMTool:
|
|
| 65 |
return response
|
| 66 |
except Exception as e:
|
| 67 |
error_str = str(e)
|
| 68 |
-
# 檢查是否為 Rate Limit 相關錯誤
|
| 69 |
if "429" in error_str or "RESOURCE_EXHAUSTED" in error_str or "503" in error_str:
|
| 70 |
wait_time = delay + random.uniform(0, 3)
|
| 71 |
print(f"⚠️ API 忙碌 (Attempt {attempt+1}/{retries}),休息 {wait_time:.1f} 秒...", flush=True)
|
| 72 |
time.sleep(wait_time)
|
| 73 |
-
delay *= 1.5
|
| 74 |
else:
|
| 75 |
-
raise e
|
| 76 |
|
| 77 |
raise Exception("API 重試多次失敗,請檢查配額。")
|
| 78 |
|
|
@@ -81,6 +84,13 @@ class NotebookLMTool:
|
|
| 81 |
"""處理單一頁面的:去字(背景) + 文字分析(Layout)"""
|
| 82 |
print(f"🚀 [Page {page_index+1}] 啟動處理...", flush=True)
|
| 83 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 84 |
result = {
|
| 85 |
"index": page_index,
|
| 86 |
"bg_path": None,
|
|
@@ -107,8 +117,9 @@ class NotebookLMTool:
|
|
| 107 |
3. Output ONLY the image.
|
| 108 |
"""
|
| 109 |
|
| 110 |
-
#
|
| 111 |
resp_img = self._call_gemini_with_retry(
|
|
|
|
| 112 |
model_name="gemini-2.5-flash-image",
|
| 113 |
contents=[clean_prompt, img],
|
| 114 |
config=types.GenerateContentConfig(response_modalities=["IMAGE"])
|
|
@@ -139,7 +150,7 @@ class NotebookLMTool:
|
|
| 139 |
except Exception as e:
|
| 140 |
print(f"❌ [Page {page_index+1}] Clean Error: {e}", flush=True)
|
| 141 |
|
| 142 |
-
# 失敗回退原圖
|
| 143 |
if not bg_success:
|
| 144 |
img.save(final_bg_path)
|
| 145 |
result["bg_path"] = final_bg_path
|
|
@@ -155,8 +166,9 @@ class NotebookLMTool:
|
|
| 155 |
Each item: {"text": string, "box_2d": [ymin, xmin, ymax, xmax] (0-1000), "font_size": int, "color": hex, "is_bold": bool}
|
| 156 |
"""
|
| 157 |
|
| 158 |
-
#
|
| 159 |
resp_layout = self._call_gemini_with_retry(
|
|
|
|
| 160 |
model_name="gemini-2.5-flash",
|
| 161 |
contents=[layout_prompt, img],
|
| 162 |
config=types.GenerateContentConfig(response_mime_type="application/json")
|
|
@@ -177,7 +189,7 @@ class NotebookLMTool:
|
|
| 177 |
return result
|
| 178 |
|
| 179 |
def process_pdf(self, pdf_file, progress=gr.Progress()):
|
| 180 |
-
if not self.
|
| 181 |
raise ValueError("請先輸入 Google API Key!")
|
| 182 |
|
| 183 |
if pdf_file is None:
|
|
@@ -206,8 +218,7 @@ class NotebookLMTool:
|
|
| 206 |
except Exception as e:
|
| 207 |
raise ValueError(f"PDF 轉換失敗: {str(e)}")
|
| 208 |
|
| 209 |
-
# 3. 平行處理 (Parallel Execution
|
| 210 |
-
# 降低併發數以適應免費版 API
|
| 211 |
max_workers = 2
|
| 212 |
results_map = {}
|
| 213 |
|
|
@@ -216,16 +227,17 @@ class NotebookLMTool:
|
|
| 216 |
with concurrent.futures.ThreadPoolExecutor(max_workers=max_workers) as executor:
|
| 217 |
future_to_page = {}
|
| 218 |
for i, img in enumerate(images):
|
| 219 |
-
time.sleep(1.5) #
|
| 220 |
future = executor.submit(self.process_single_page, i, img, img_output_dir)
|
| 221 |
future_to_page[future] = i
|
| 222 |
|
| 223 |
for future in concurrent.futures.as_completed(future_to_page):
|
| 224 |
try:
|
| 225 |
res = future.result()
|
| 226 |
-
|
| 227 |
-
|
| 228 |
-
|
|
|
|
| 229 |
except Exception as exc:
|
| 230 |
print(f"Page processing generated an exception: {exc}")
|
| 231 |
|
|
@@ -331,7 +343,7 @@ with gr.Blocks(title="NotebookLM Slide Restorer,PPT.404", theme=gr.themes.Soft
|
|
| 331 |
|
| 332 |
gr.Markdown("---")
|
| 333 |
pdf_input = gr.File(label="上傳 PDF")
|
| 334 |
-
btn_process = gr.Button("🚀 開始還原 PPTX (
|
| 335 |
|
| 336 |
with gr.Column():
|
| 337 |
out_zip = gr.File(label="📦 下載完整包")
|
|
|
|
| 27 |
class NotebookLMTool:
|
| 28 |
def __init__(self):
|
| 29 |
self.api_key = os.getenv("GEMINI_API_KEY")
|
| 30 |
+
# 移除全域 client,改為動態建立以確保執行緒安全
|
|
|
|
|
|
|
| 31 |
|
| 32 |
def set_key(self, user_key):
|
| 33 |
if user_key and user_key.strip():
|
| 34 |
self.api_key = user_key.strip()
|
|
|
|
| 35 |
return "✅ API Key 已更新!"
|
| 36 |
return "⚠️ Key 無效"
|
| 37 |
|
|
|
|
| 46 |
except:
|
| 47 |
return []
|
| 48 |
|
| 49 |
+
def _create_client(self):
|
| 50 |
+
"""為每個執行緒建立獨立的 Client"""
|
| 51 |
+
if not self.api_key:
|
| 52 |
+
raise ValueError("API Key 未設定")
|
| 53 |
+
return genai.Client(api_key=self.api_key)
|
| 54 |
+
|
| 55 |
+
def _call_gemini_with_retry(self, client, model_name, contents, config=None, retries=5):
|
| 56 |
"""
|
| 57 |
+
封裝 Gemini 呼叫,加入指數退避重試機制
|
| 58 |
"""
|
| 59 |
delay = 5 # 初始等待秒數
|
| 60 |
|
| 61 |
for attempt in range(retries):
|
| 62 |
try:
|
| 63 |
+
response = client.models.generate_content(
|
| 64 |
model=model_name,
|
| 65 |
contents=contents,
|
| 66 |
config=config
|
|
|
|
| 68 |
return response
|
| 69 |
except Exception as e:
|
| 70 |
error_str = str(e)
|
| 71 |
+
# 檢查是否為 Rate Limit 相關錯誤
|
| 72 |
if "429" in error_str or "RESOURCE_EXHAUSTED" in error_str or "503" in error_str:
|
| 73 |
wait_time = delay + random.uniform(0, 3)
|
| 74 |
print(f"⚠️ API 忙碌 (Attempt {attempt+1}/{retries}),休息 {wait_time:.1f} 秒...", flush=True)
|
| 75 |
time.sleep(wait_time)
|
| 76 |
+
delay *= 1.5
|
| 77 |
else:
|
| 78 |
+
raise e
|
| 79 |
|
| 80 |
raise Exception("API 重試多次失敗,請檢查配額。")
|
| 81 |
|
|
|
|
| 84 |
"""處理單一頁面的:去字(背景) + 文字分析(Layout)"""
|
| 85 |
print(f"🚀 [Page {page_index+1}] 啟動處理...", flush=True)
|
| 86 |
|
| 87 |
+
# 關鍵修改:在此處建立獨立的 Client,避免執行緒衝突
|
| 88 |
+
try:
|
| 89 |
+
local_client = self._create_client()
|
| 90 |
+
except Exception as e:
|
| 91 |
+
print(f"❌ [Page {page_index+1}] Client Init Error: {e}")
|
| 92 |
+
return None
|
| 93 |
+
|
| 94 |
result = {
|
| 95 |
"index": page_index,
|
| 96 |
"bg_path": None,
|
|
|
|
| 117 |
3. Output ONLY the image.
|
| 118 |
"""
|
| 119 |
|
| 120 |
+
# 使用 gemini-2.5-flash-image (支援繪圖)
|
| 121 |
resp_img = self._call_gemini_with_retry(
|
| 122 |
+
client=local_client,
|
| 123 |
model_name="gemini-2.5-flash-image",
|
| 124 |
contents=[clean_prompt, img],
|
| 125 |
config=types.GenerateContentConfig(response_modalities=["IMAGE"])
|
|
|
|
| 150 |
except Exception as e:
|
| 151 |
print(f"❌ [Page {page_index+1}] Clean Error: {e}", flush=True)
|
| 152 |
|
| 153 |
+
# 失敗回退原圖
|
| 154 |
if not bg_success:
|
| 155 |
img.save(final_bg_path)
|
| 156 |
result["bg_path"] = final_bg_path
|
|
|
|
| 166 |
Each item: {"text": string, "box_2d": [ymin, xmin, ymax, xmax] (0-1000), "font_size": int, "color": hex, "is_bold": bool}
|
| 167 |
"""
|
| 168 |
|
| 169 |
+
# 使用一般的 flash 模型做文字分析
|
| 170 |
resp_layout = self._call_gemini_with_retry(
|
| 171 |
+
client=local_client,
|
| 172 |
model_name="gemini-2.5-flash",
|
| 173 |
contents=[layout_prompt, img],
|
| 174 |
config=types.GenerateContentConfig(response_mime_type="application/json")
|
|
|
|
| 189 |
return result
|
| 190 |
|
| 191 |
def process_pdf(self, pdf_file, progress=gr.Progress()):
|
| 192 |
+
if not self.api_key:
|
| 193 |
raise ValueError("請先輸入 Google API Key!")
|
| 194 |
|
| 195 |
if pdf_file is None:
|
|
|
|
| 218 |
except Exception as e:
|
| 219 |
raise ValueError(f"PDF 轉換失敗: {str(e)}")
|
| 220 |
|
| 221 |
+
# 3. 平行處理 (Parallel Execution)
|
|
|
|
| 222 |
max_workers = 2
|
| 223 |
results_map = {}
|
| 224 |
|
|
|
|
| 227 |
with concurrent.futures.ThreadPoolExecutor(max_workers=max_workers) as executor:
|
| 228 |
future_to_page = {}
|
| 229 |
for i, img in enumerate(images):
|
| 230 |
+
time.sleep(1.5) # 錯開請求
|
| 231 |
future = executor.submit(self.process_single_page, i, img, img_output_dir)
|
| 232 |
future_to_page[future] = i
|
| 233 |
|
| 234 |
for future in concurrent.futures.as_completed(future_to_page):
|
| 235 |
try:
|
| 236 |
res = future.result()
|
| 237 |
+
if res:
|
| 238 |
+
results_map[res["index"]] = res
|
| 239 |
+
total_input_tokens += res["tokens_in"]
|
| 240 |
+
total_output_tokens += res["tokens_out"]
|
| 241 |
except Exception as exc:
|
| 242 |
print(f"Page processing generated an exception: {exc}")
|
| 243 |
|
|
|
|
| 343 |
|
| 344 |
gr.Markdown("---")
|
| 345 |
pdf_input = gr.File(label="上傳 PDF")
|
| 346 |
+
btn_process = gr.Button("🚀 開始還原 PPTX (穩定版)", variant="primary")
|
| 347 |
|
| 348 |
with gr.Column():
|
| 349 |
out_zip = gr.File(label="📦 下載完整包")
|