Update app.py

app.py

@@ -123,13 +123,11 @@ def get_tree_html(dir_path):
         for item in items:
             item_path = os.path.join(current_path, item)
             if os.path.isdir(item_path):
-                # 文件夹节点 (可折叠)
                 html += f'<details style="margin-left: 15px; cursor: pointer; margin-top: 4px;"><summary style="outline: none; color: #C084FC;">📁 <b>{item}</b></summary>'
                 inner_html = build_html(item_path)
                 html += inner_html if inner_html else "<div style='margin-left: 20px; color: #888; font-size: 12px;'><i>Empty</i></div>"
                 html += '</details>'
             else:
-                # 文件节点
                 html += f'<div style="margin-left: 18px; padding-left: 12px; border-left: 1px dotted #A855F7; margin-top: 4px; color: #DAB2FF;">📄 {item}</div>'
         return html
 
@@ -155,10 +153,32 @@ def get_debug_info():
     """
     return html
 
+def save_api_settings(api_key, api_base_url=None):
+    if not api_key: 
+        return "❌ Key cannot be empty", get_debug_info(), False
+    try:
+        config = {}
+        if os.path.exists(CONFIG_PATH):
+            with open(CONFIG_PATH, "r", encoding="utf-8") as f:
+                config = yaml.safe_load(f) or {}
+        config.setdefault("api_keys", {})["gemini_api_key"] = api_key
+        if api_base_url:
+            config["api_base_url"] = api_base_url
+        with open(CONFIG_PATH, "w", encoding="utf-8") as f:
+            yaml.dump(config, f, allow_unicode=True)
+            
+        success_msg = "✅ Key saved"
+        if api_base_url:
+            success_msg += ", Base URL updated"
+            
+        # 返回成功消息和 True (表示 API 已就绪)
+        return success_msg, get_debug_info(), True
+    except Exception as e:
+        return f"❌ Error: {str(e)}", get_debug_info(), False
+
 def save_pdf(file):
-    if file is None: return "❌ Please select a PDF first", get_debug_info()
+    if file is None: return "❌ Please select a PDF first", get_debug_info(), False
     try:
-        # 先清空原有 PDF，确保只有最新的一份
         for f in os.listdir(PAPERS_DIR):
             file_to_del = os.path.join(PAPERS_DIR, f)
             if os.path.isfile(file_to_del):
@@ -166,72 +186,54 @@ def save_pdf(file):
                 
         file_path = os.path.join(PAPERS_DIR, os.path.basename(file.name))
         shutil.copy(file.name, file_path)
-        return f"✅ Saved: {os.path.basename(file.name)}", get_debug_info()
+        # 返回成功消息和 True (表示 PDF 已就绪)
+        return f"✅ Saved: {os.path.basename(file.name)}", get_debug_info(), True
     except Exception as e:
-        return f"❌ Error: {str(e)}", get_debug_info()
+        return f"❌ Error: {str(e)}", get_debug_info(), False
 
 def clear_pdf():
-    """Delete PDF and its corresponding mineru output directory"""
+    """清空 PDF 并锁定后续所有步骤"""
     try:
         deleted_files = []
         deleted_dirs = []
         
-        # 1. 记录要被删除的 PDF 文件名（用于匹配输出目录）
         pdf_names = [f for f in os.listdir(PAPERS_DIR) if f.endswith('.pdf')]
         
-        # 2. 删除 papers 目录下的 PDF
         for f in os.listdir(PAPERS_DIR):
             file_to_del = os.path.join(PAPERS_DIR, f)
             if os.path.isfile(file_to_del):
                 os.remove(file_to_del)
                 deleted_files.append(f)
                 
-        # 3. 删除 mineru_outputs 下的同名文件夹
         if os.path.exists(OUTPUT_DIR):
             for pdf_name in pdf_names:
                 base_name = os.path.splitext(pdf_name)[0]
-                base_name_us = base_name.replace(" ", "_") # 兼容 Mineru 可能会把空格转为下划线
+                base_name_us = base_name.replace(" ", "_")
                 
                 for item in os.listdir(OUTPUT_DIR):
-                    # 匹配严格同名，或者下划线名称，或者带 .pdf 后缀的文件夹
                     if item in [base_name, base_name_us, pdf_name]:
                         dir_to_del = os.path.join(OUTPUT_DIR, item)
                         if os.path.isdir(dir_to_del):
                             shutil.rmtree(dir_to_del)
                             deleted_dirs.append(item)
                             
-        # 4. 删除可能遗留的打包压缩文件
         if os.path.exists(ZIP_OUTPUT_PATH):
             os.remove(ZIP_OUTPUT_PATH)
             
+        disable_btn = gr.update(interactive=False)
+        # 返回：状态文本, debug视窗, pdf_ready=False, 以及锁定4个生成按钮
         if deleted_files or deleted_dirs:
-            return f"🗑️ Workspace cleared (Deleted: {len(deleted_files)} PDF, {len(deleted_dirs)} Output Folder)", get_debug_info()
-        return "ℹ️ Workspace is already empty", get_debug_info()
+            return f"🗑️ Workspace cleared", get_debug_info(), False, disable_btn, disable_btn, disable_btn, disable_btn
+        return "ℹ️ Workspace is already empty", get_debug_info(), False, disable_btn, disable_btn, disable_btn, disable_btn
     except Exception as e:
-        return f"❌ Error deleting file: {str(e)}", get_debug_info()
-
-def save_api_settings(api_key, api_base_url=None):
-    if not api_key: return "❌ Key cannot be empty", get_debug_info()
-    try:
-        config = {}
-        if os.path.exists(CONFIG_PATH):
-            with open(CONFIG_PATH, "r", encoding="utf-8") as f:
-                config = yaml.safe_load(f) or {}
-        config.setdefault("api_keys", {})["gemini_api_key"] = api_key
-        if api_base_url:
-            config["api_base_url"] = api_base_url
-        with open(CONFIG_PATH, "w", encoding="utf-8") as f:
-            yaml.dump(config, f, allow_unicode=True)
-        success_msg = "✅ Key saved"
-        if api_base_url:
-            success_msg += ", Base URL updated"
-        return success_msg, get_debug_info()
-    except Exception as e:
-        return f"❌ Error: {str(e)}", get_debug_info()
+        return f"❌ Error deleting file: {str(e)}", get_debug_info(), False, gr.update(), gr.update(), gr.update(), gr.update()
 
 def run_mineru_parsing_and_dag_gen():
+    no_change = gr.update()
+    disable_btn = gr.update(interactive=False)
+    
     if not os.path.exists(PAPERS_DIR) or not any(f.endswith('.pdf') for f in os.listdir(PAPERS_DIR)):
-        yield "❌ No PDF file found", get_debug_info(), "No execution logs."
+        yield "❌ No PDF file found", get_debug_info(), "No execution logs.", no_change, no_change, no_change, no_change
         return
 
     full_log = ""
@@ -244,40 +246,47 @@ def run_mineru_parsing_and_dag_gen():
 
         command_mineru = ["mineru", "-p", PAPERS_DIR, "-o", OUTPUT_DIR]
         full_log += "--- Mineru Executing ---\n"
-        yield "⏳ Executing Mineru parsing...", get_debug_info(), full_log
+        # 运行中保持按钮状态不变
+        yield "⏳ Executing Mineru parsing...", get_debug_info(), full_log, no_change, no_change, no_change, no_change
 
         process_mineru = subprocess.Popen(
             command_mineru, env=env, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, text=True, bufsize=1
         )
         for line in iter(process_mineru.stdout.readline, ''):
             full_log += line
-            yield "⏳ Executing Mineru parsing...", get_debug_info(), full_log
+            yield "⏳ Executing Mineru parsing...", get_debug_info(), full_log, no_change, no_change, no_change, no_change
         process_mineru.stdout.close()
         returncode_mineru = process_mineru.wait()
 
         if returncode_mineru != 0:
-            yield f"❌ Mineru parsing failed (Exit Code: {returncode_mineru})", get_debug_info(), full_log
+            yield f"❌ Mineru parsing failed (Exit Code: {returncode_mineru})", get_debug_info(), full_log, disable_btn, disable_btn, disable_btn, disable_btn
             return
 
         command_dag = [sys.executable, "gen_dag.py"]
         full_log += "\n--- DAG Gen Executing ---\n"
-        yield "⏳ Mineru parsing complete, executing DAG generation...", get_debug_info(), full_log
+        yield "⏳ Mineru parsing complete, executing DAG generation...", get_debug_info(), full_log, no_change, no_change, no_change, no_change
         
         process_dag = subprocess.Popen(
             command_dag, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, text=True, bufsize=1
         )
         for line in iter(process_dag.stdout.readline, ''):
             full_log += line
-            yield "⏳ Executing DAG generation...", get_debug_info(), full_log
+            yield "⏳ Executing DAG generation...", get_debug_info(), full_log, no_change, no_change, no_change, no_change
         process_dag.stdout.close()
         returncode_dag = process_dag.wait()
         
-        status = "✅ PDF parsing & DAG generation fully completed" if returncode_dag == 0 else f"❌ DAG generation failed (Exit Code: {returncode_dag})"
-        yield status, get_debug_info(), full_log
+        # 解析完全成功，解锁第三部分的按钮
+        if returncode_dag == 0:
+            status = "✅ PDF parsing & DAG generation fully completed"
+            enable_btn = gr.update(interactive=True)
+            yield status, get_debug_info(), full_log, enable_btn, enable_btn, enable_btn, enable_btn
+        else:
+            status = f"❌ DAG generation failed (Exit Code: {returncode_dag})"
+            yield status, get_debug_info(), full_log, disable_btn, disable_btn, disable_btn, disable_btn
 
     except Exception as e:
         error_log = full_log + f"\n[Global Exception] Exception occurred:\n{str(e)}"
-        yield "❌ Execution Exception", get_debug_info(), error_log
+        yield "❌ Execution Exception", get_debug_info(), error_log, disable_btn, disable_btn, disable_btn, disable_btn
 
 
 def run_final_generation(task_type="all"):
@@ -303,14 +312,12 @@ def run_final_generation(task_type="all"):
     q = queue.Queue()
     processes = []
     
-    # 将标准输出推送到队列的工作线程
     def enqueue_output(out, script_name):
         for line in iter(out.readline, ''):
             q.put((script_name, line))
         out.close()
 
     try:
-        # 并发启动所有脚本
         for script in scripts_to_run:
             p = subprocess.Popen(
                 [sys.executable, script], 
@@ -324,18 +331,15 @@ def run_final_generation(task_type="all"):
             t.daemon = True
             t.start()
 
-        # 从队列中实时读取日志并更新 UI
         active_processes = len(processes)
         while active_processes > 0 or not q.empty():
             try:
-                # 设定 timeout，这样即便没有日志产生也能循环检查进程是否结束
                 script_name, line = q.get(timeout=0.1)
                 full_log += f"[{script_name}] {line}"
                 yield f"⏳ Generating {task_type.upper()}...", get_debug_info(), full_log, gr.update(visible=False)
             except queue.Empty:
                 active_processes = sum(1 for _, p in processes if p.poll() is None)
 
-        # 检查最终执行结果
         success = True
         for script, p in processes:
             if p.returncode != 0:
@@ -346,7 +350,6 @@ def run_final_generation(task_type="all"):
             yield f"❌ {task_type.upper()} contains failed tasks, please check logs", get_debug_info(), full_log, gr.update(visible=False)
             return
 
-        # 全部成功后进行压缩
         full_log += "\n📦 Zipping output directory...\n"
         yield f"⏳ Zipping outputs...", get_debug_info(), full_log, gr.update(visible=False)
         
@@ -510,6 +513,21 @@ body, .gradio-container {
 }
 .action-btn:active { transform: translateY(2px) scale(0.98) !important; box-shadow: 0 2px 10px rgba(147, 51, 234, 0.2) !important; }
 
+/* ======== CSS FOR DISABLED BUTTONS ======== */
+.primary-action-btn:disabled, .action-btn:disabled {
+    background: #e5e7eb !important;
+    color: #9ca3af !important;
+    box-shadow: none !important;
+    transform: none !important;
+    cursor: not-allowed !important;
+    border: 1px solid #d1d5db !important;
+}
+.dark .primary-action-btn:disabled, .dark .action-btn:disabled {
+    background: #374151 !important;
+    color: #6b7280 !important;
+    border: 1px solid #4b5563 !important;
+}
+
 .log-box textarea { font-family: 'IBM Plex Mono', monospace !important; font-size: 13px !important; background-color: #1e1e1e !important; color: #DAB2FF !important; border: 1px solid #C084FC !important; border-radius: 8px !important; }
 .status-text textarea { background-color: transparent !important; border: none !important; box-shadow: none !important; font-weight: 600 !important; color: #6B21A8 !important; }
 .dark .status-text textarea { color: #C084FC !important; }
@@ -518,12 +536,15 @@ body, .gradio-container {
 ::-webkit-scrollbar-thumb { background: linear-gradient(135deg, #A855F7, #C084FC); border-radius: 4px; }
 ::-webkit-scrollbar-thumb:hover { background: linear-gradient(135deg, #9333EA, #A855F7); }
 
-/* 使折叠树中的 details 和 summary 更加平滑 */
 details > summary { transition: color 0.2s ease; }
 details > summary:hover { color: #E9D5FF !important; }
 """
 
 with gr.Blocks(theme=purple_theme, css=custom_css) as demo:
+    # 定义全局状态变量，用于实现依赖解锁
+    api_saved_state = gr.State(False)
+    pdf_ready_state = gr.State(False)
+
     with gr.Column(elem_id="col-container"):
         gr.Markdown("# **PaperX Platform**", elem_id="main-title")
         gr.Markdown("One-click parsing of academic PDFs, DAG structuring, and multi-modal asset generation.", elem_id="subtitle")
@@ -539,6 +560,10 @@ with gr.Blocks(theme=purple_theme, css=custom_css) as demo:
                         key_input = gr.Textbox(label="API Key", type="password", placeholder="sk-...", scale=1)
                         api_base_url_input = gr.Textbox(label="Base URL (Optional)", placeholder="https://api.example.com", scale=1)
                     key_btn = gr.Button("💾 Save API Configuration")
+                    
+                    api_status = gr.Textbox(
+                        show_label=False, placeholder="Waiting for API configuration...", lines=1, interactive=False, elem_classes="status-text"
+                    )
                 
                 # 2. Document Parsing
                 with gr.Group(elem_classes="gradio-group"):
@@ -550,7 +575,8 @@ with gr.Blocks(theme=purple_theme, css=custom_css) as demo:
                         elem_id="pdf-upload-box"
                     )
                     
-                    parse_btn = gr.Button("🚀 Start Mineru & DAG Extraction", elem_classes="primary-action-btn")
+                    # 默认锁定
+                    parse_btn = gr.Button("🚀 Start Mineru & DAG Extraction", elem_classes="primary-action-btn", interactive=False)
                     
                     parse_status = gr.Textbox(
                         show_label=False, placeholder="Waiting for document upload...", lines=1, interactive=False, elem_classes="status-text"
@@ -562,11 +588,13 @@ with gr.Blocks(theme=purple_theme, css=custom_css) as demo:
                     gr.Markdown("Generate final formats based on DAG structure:")
                     
                     with gr.Row(elem_classes="action-row"):
-                        gen_ppt_btn = gr.Button("📊 Gen PPT", elem_classes="action-btn")
-                        gen_poster_btn = gr.Button("🖼️ Gen Poster", elem_classes="action-btn")
-                        gen_pr_btn = gr.Button("📰 Gen PR", elem_classes="action-btn")
+                        # 默认锁定
+                        gen_ppt_btn = gr.Button("📊 Gen PPT", elem_classes="action-btn", interactive=False)
+                        gen_poster_btn = gr.Button("🖼️ Gen Poster", elem_classes="action-btn", interactive=False)
+                        gen_pr_btn = gr.Button("📰 Gen PR", elem_classes="action-btn", interactive=False)
                     
-                    gen_all_btn = gr.Button("✨ Generate All Assets (ALL)", elem_classes="primary-action-btn")
+                    # 默认锁定
+                    gen_all_btn = gr.Button("✨ Generate All Assets (ALL)", elem_classes="primary-action-btn", interactive=False)
 
             # ================= RIGHT COLUMN: OUTPUTS & LOGS =================
             with gr.Column(scale=1):
@@ -590,21 +618,46 @@ with gr.Blocks(theme=purple_theme, css=custom_css) as demo:
                         
                         with gr.Tab("🔍 System Snapshot"):
                             refresh_btn = gr.Button("🔄 Refresh Directory Tree") 
-                            # 这里改为 HTML 组件，以便渲染我们的 details/summary 折叠树
                             debug_view = gr.HTML(
                                 value=get_debug_info()
                             )
 
     # ================= LOGIC BINDINGS =================
-    key_btn.click(fn=save_api_settings, inputs=[key_input, api_base_url_input], outputs=[parse_status, debug_view])
     
-    pdf_input.upload(fn=save_pdf, inputs=pdf_input, outputs=[parse_status, debug_view])
-    # 彻底清除：不仅清除 PDF，还清除 mineru output 中的对应文件夹
-    pdf_input.clear(fn=clear_pdf, outputs=[parse_status, debug_view])
+    # [步骤 1] 配置 API
+    # 成功后不仅更新 UI 文本，还会将 api_saved_state 设置为 True
+    key_btn.click(
+        fn=save_api_settings, 
+        inputs=[key_input, api_base_url_input], 
+        outputs=[api_status, debug_view, api_saved_state]
+    )
+    
+    # [步骤 2.1] 上传 PDF
+    # 成功后更新文本，并将 pdf_ready_state 设置为 True
+    pdf_input.upload(fn=save_pdf, inputs=pdf_input, outputs=[parse_status, debug_view, pdf_ready_state])
+    
+    # [步骤 2.2] 清除 PDF
+    # 会锁定第四步的所有生成按钮，同时将 pdf_ready_state 重置为 False
+    pdf_input.clear(
+        fn=clear_pdf, 
+        outputs=[parse_status, debug_view, pdf_ready_state, gen_ppt_btn, gen_poster_btn, gen_pr_btn, gen_all_btn]
+    )
     
-    parse_btn.click(fn=run_mineru_parsing_and_dag_gen, outputs=[parse_status, debug_view, cmd_logs])
+    # [依赖检查] 当 api 状态或 pdf 状态发生变化时，检查是否解锁"开始解析"按钮
+    def check_parse_btn_ready(api_ready, pdf_ready):
+        return gr.update(interactive=(api_ready and pdf_ready))
+        
+    api_saved_state.change(fn=check_parse_btn_ready, inputs=[api_saved_state, pdf_ready_state], outputs=parse_btn)
+    pdf_ready_state.change(fn=check_parse_btn_ready, inputs=[api_saved_state, pdf_ready_state], outputs=parse_btn)
+
+    # [步骤 2.3] 执行解析
+    # 如果完全成功，函数最后的一个 yield 会连带输出 interactive=True 给这4个生成按钮
+    parse_btn.click(
+        fn=run_mineru_parsing_and_dag_gen, 
+        outputs=[parse_status, debug_view, cmd_logs, gen_ppt_btn, gen_poster_btn, gen_pr_btn, gen_all_btn]
+    )
 
-    # 为了确保 Gradio 生成器模式正确工作，拆分为显式的封装函数
+    # [步骤 3] 资产生成
     def trigger_gen_ppt(): yield from run_final_generation("ppt")
     def trigger_gen_poster(): yield from run_final_generation("poster")
     def trigger_gen_pr(): yield from run_final_generation("pr")
@@ -618,4 +671,4 @@ with gr.Blocks(theme=purple_theme, css=custom_css) as demo:
     refresh_btn.click(fn=get_debug_info, outputs=debug_view)
 
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch(debug=True)