pdf_download_csv_output

app.py

@@ -347,39 +347,37 @@ def convert_pdf_to_images(pdf_path):
 
 def process_pdf_to_data(password, pdf_file):
     if password != PASSWORD:
-        raise gr.Error("密码错误，请重新输入")
+        raise gr.Error("密碼错误，请重新输入")
 
     processed_data = []
     question_count = 0
 
     pdf_image_paths = convert_pdf_to_images(pdf_file.name)
-    for pdf_image_path in pdf_image_paths:
-        # pdf_image_to_text
-        image_url = upload_image_to_gcs(pdf_image_path, bucket)
-        text = pdf_image_to_text(image_url)
+    image_urls = [upload_image_to_gcs(pdf_image_path, bucket) for pdf_image_path in pdf_image_paths]
+    text = pdf_image_to_text(image_urls)
 
-        print("======pdf_image_to_text=====")
-        print(text)
-        print("========================")
+    print("======pdf_image_to_text=====")
+    print(text)
+    print("========================")
 
-        text = text.replace("```json", "").replace("```", "")
-        text_json = safe_json_loads(text)
-        for text_item in text_json:
-            print("======text_to_json=====")
-            print(text_item)
-            print("========================")
+    text = text.replace("```json", "").replace("```", "")
+    text_json = safe_json_loads(text)
+    for text_item in text_json:
+        print("======text_to_json=====")
+        print(text_item)
+        print("========================")
 
-            question_json = safe_json_loads(text_to_json(text_item))
-            perseus_json_str = build_perseus_json(question_json)
-            processed_data.append([image_url, text] + list(question_json.values()) + [perseus_json_str])
-            question_count += 1
+        question_json = safe_json_loads(text_to_json(text_item))
+        perseus_json_str = build_perseus_json(question_json)
+        processed_data.append(["", text] + list(question_json.values()) + [perseus_json_str])
+        question_count += 1
 
     result = f"PDF 處理完成，總共完成 {question_count} 道題目"
     csv_file_path = create_csv(processed_data)
 
     return processed_data, result, csv_file_path
 
-def pdf_image_to_text(image_url):
+def pdf_image_to_text(image_urls):
     user_prompt = """
         請解讀題目圖片:
         - 圖片請一定要用 zh-TW 解讀
@@ -405,25 +403,31 @@ def pdf_image_to_text(image_url):
         ]
     """
 
-    response = OPEN_AI_CLIENT.chat.completions.create(
-      model="gpt-4o",
-      messages=[
+    messages=[
         {
           "role": "user",
           "content": [
             {
                 "type": "text",
                 "text": user_prompt
-            },
+            }
+          ],
+        }
+      ]
+
+    for image_url in image_urls:
+        messages[0]["content"].append(
             {
               "type": "image_url",
               "image_url": {
                 "url": image_url,
               },
-            },
-          ],
-        }
-      ],
+            }
+        )
+
+    response = OPEN_AI_CLIENT.chat.completions.create(
+      model="gpt-4o",
+      messages=messages,
       max_tokens=4000,
     )
     return response.choices[0].message.content
@@ -442,23 +446,68 @@ def safe_json_loads(json_string):
             print(f"Second JSONDecodeError: {e2}")
             raise e2
 
+def show_multiple_questions_markdown(data):
+    if len(data) == 0:
+        return ""
+    
+    markdown = ""
+
+    for i in range(len(data)):
+        question_json = data.iloc[i].to_dict()  # 確保訪問的是 DataFrame 的第一行並轉換為字典
+        question = question_json['題目']
+        choice_1 = question_json['選項1']
+        choice_2 = question_json['選項2']
+        choice_3 = question_json['選項3']
+        choice_4 = question_json['選項4']
+        answer = question_json['答案']
+        
+        hints = []
+        for i in range(1, 6):
+            hint_key = question_json.get(f'提示{i}', None)
+            if hint_key:
+                hints.append(hint_key)
+            else:
+                break
+
+
+        markdown += f"""
+
+---
+        
+## 題目
+- {question}
+
+## 選項
+1. {choice_1}
+2. {choice_2}
+3. {choice_3}
+4. {choice_4}
+
+## 答案: {answer}
+
+## 提示
+
+"""
+        for i, hint in enumerate(hints):
+            markdown += f"{i+1}. {hint}\n"
+
+    return markdown
 
 # Gradio界面
 with gr.Blocks() as demo:
     with gr.Row():
-        password_input = gr.Textbox(label="密码", type="password")
+        password_input = gr.Textbox(label="密碼", type="password")
     
     with gr.Tab("批量處理"):
         with gr.Row():
-            gr.Markdown("## 批量图片處理 + Perseus JSON 生成")
+            gr.Markdown("## 批量圖片處理 + Perseus JSON 生成")
             image_input = gr.Files(label="選擇圖片", type="filepath")
-            # sheet_default_value = "https://docs.google.com/spreadsheets/d/1ygFGLxcnPad3LMVj4bZqfGZh1n2wqhs0-vOUjuVCkSY/edit#gid=0"
-            # sheet_input = gr.Textbox(label="Google Sheets URL", value=sheet_default_value)
             submit_button = gr.Button("開始��理圖片")
         with gr.Row():
             result_text = gr.Textbox(label="處理結果")
-        with gr.Row():
             download_csv_output = gr.File(label="下载 CSV")
+        with gr.Row():
+            batch_question_markdown = gr.Markdown(show_label=False, latex_delimiters=[{"left": "$", "right": "$", "display": False}])    
         with gr.Accordion(open=False):
             with gr.Row():
                 result_table = gr.Dataframe(
@@ -469,16 +518,15 @@ with gr.Blocks() as demo:
     
     with gr.Tab("單張處理"):
         with gr.Row():
-            gr.Markdown("## 單張图片處理")
-            single_image_input = gr.Image(sources=['upload', 'webcam', 'clipboard'])
+            gr.Markdown("## 單張圖片處理")
+            single_image_input =  gr.Files(label="選擇圖片", type="filepath")
             single_submit_button = gr.Button("開始處理圖片")
         with gr.Row():
             single_result_text = gr.Textbox(label="處理結果")
+            single_download_csv_output = gr.File(label="下载 CSV")
         with gr.Row():
             single_question_image = gr.Image()
-            single_question_markdown = gr.Markdown(show_label=False, latex_delimiters=[{"left": "$", "right": "$", "display": False}]) 
-        with gr.Row():
-            single_download_csv_output = gr.File(label="下载 CSV")
+            single_question_markdown = gr.Markdown(show_label=False, latex_delimiters=[{"left": "$", "right": "$", "display": False}])             
         with gr.Accordion(open=False):
             with gr.Row():
                 single_result_table = gr.Dataframe(
@@ -494,8 +542,9 @@ with gr.Blocks() as demo:
             pdf_submit_button = gr.Button("開始處理 PDF")
         with gr.Row():
             pdf_result_text = gr.Textbox(label="處理結果")
-        with gr.Row():
             pdf_download_csv_output = gr.File(label="下载 CSV")
+        with gr.Row():
+            pdf_question_markdown = gr.Markdown(show_label=False, latex_delimiters=[{"left": "$", "right": "$", "display": False}])
         with gr.Accordion(open=False):
             with gr.Row():
                 pdf_result_table = gr.Dataframe(
@@ -503,13 +552,15 @@ with gr.Blocks() as demo:
                     column_widths=[10, 10, 5, 20, 4, 4, 4, 4, 4,4,4,4,4,4, 10],
                     wrap=True
                 )
-                # pdf_result_table = gr.Textbox(label="處理結果")
-
 
     submit_button.click(
         fn=process_image_to_data,
         inputs=[password_input, image_input],
         outputs=[result_table, result_text, download_csv_output]
+    ).then(
+        fn=show_multiple_questions_markdown,
+        inputs=[result_table],
+        outputs=[batch_question_markdown]
     )
 
     single_submit_button.click(
@@ -530,6 +581,10 @@ with gr.Blocks() as demo:
         fn=process_pdf_to_data,
         inputs=[password_input, pdf_input],
         outputs=[pdf_result_table, pdf_result_text, pdf_download_csv_output]
+    ).then(
+        fn=show_multiple_questions_markdown,
+        inputs=[pdf_result_table],
+        outputs=[pdf_question_markdown]
     )
 
 demo.launch()