Update app.py

app.py

@@ -1,75 +1,115 @@
 import gradio as gr
 import os
-import yaml
 import shutil
 import subprocess
 import sys
 import queue
 import threading
+import uuid
 from datetime import datetime
 from concurrent.futures import ThreadPoolExecutor, as_completed
 from typing import Iterable
 from gradio.themes import Soft
 from gradio.themes.utils import colors, fonts, sizes
 
-# 自动安装 Playwright 浏览器（如果不存在）
-try:
-    import playwright
-    # 检查是否已经安装了浏览器，没有则下载
-    subprocess.run(["playwright", "install", "chromium"], check=True)
-except Exception as e:
-    print(f"Playwright setup failed: {e}")
 
-# Initialize environment paths
+import threading
+import subprocess
+# ==========================================
+# --- 🌐 异步安装 Playwright 浏览器 ---
+# ==========================================
+def setup_playwright():
+    """在后台静默安装 Playwright，防止阻塞 Gradio 启动导致 HF 500 超时"""
+    try:
+        import playwright
+        print("⏳ [System] Downloading Playwright Chromium in background...")
+        # 增加 --with-deps 尝试安装系统级依赖 (虽然在非 root 容器可能失效，但有备无患)
+        subprocess.run(["playwright", "install", "chromium"], check=True)
+        print("✅ [System] Playwright browsers ready.")
+    except Exception as e:
+        print(f"❌ [System] Playwright setup failed: {e}")
+
+# 这一步非常关键：启动一个后台守护线程去下载，主进程直接往下走！
+threading.Thread(target=setup_playwright, daemon=True).start()
+
+# ==========================================
+# --- 📁 全局目录配置 (修改为 Session 基础目录) ---
+# ==========================================
 BASE_DIR = os.path.dirname(os.path.abspath(__file__))
-PAPERS_DIR = os.path.join(BASE_DIR, "papers")
-CONFIG_PATH = os.path.join(BASE_DIR, "config.yaml")
-OUTPUT_DIR = os.path.join(BASE_DIR, "mineru_outputs")
-ZIP_OUTPUT_PATH = os.path.join(BASE_DIR, "mineru_results.zip")
+SESSIONS_BASE_DIR = os.path.join(BASE_DIR, "user_sessions")
+os.makedirs(SESSIONS_BASE_DIR, exist_ok=True)
+
+def get_user_dirs(session_id):
+    """根据 Session ID 生成用户专属的隔离目录"""
+    user_base = os.path.join(SESSIONS_BASE_DIR, session_id)
+    papers_dir = os.path.join(user_base, "papers")
+    output_dir = os.path.join(user_base, "mineru_outputs")
+    zip_path = os.path.join(user_base, "mineru_results.zip")
+    
+    os.makedirs(papers_dir, exist_ok=True)
+    os.makedirs(output_dir, exist_ok=True)
+    return papers_dir, output_dir, zip_path
+
+
+import time
+
+# ==========================================
+# --- 🧹 垃圾回收 (后台清理过期 Session) ---
+# ==========================================
+# 设定 Session 过期时间（例如：2 小时 = 7200 秒）
+SESSION_MAX_AGE_SECONDS = 2 * 60 * 60
+# 设定清理器执行间隔（例如：每 30 分钟扫描一次 = 1800 秒）
+CLEANUP_INTERVAL_SECONDS = 30 * 60
+
+def cleanup_expired_sessions():
+    """后台运行的垃圾回收任务"""
+    while True:
+        try:
+            if os.path.exists(SESSIONS_BASE_DIR):
+                current_time = time.time()
+                for session_folder in os.listdir(SESSIONS_BASE_DIR):
+                    folder_path = os.path.join(SESSIONS_BASE_DIR, session_folder)
+                    
+                    # 确保只处理目录
+                    if os.path.isdir(folder_path):
+                        # 获取文件夹的最后修改时间
+                        folder_mtime = os.path.getmtime(folder_path)
+                        
+                        # 判断是否超过了最大存活时间
+                        if (current_time - folder_mtime) > SESSION_MAX_AGE_SECONDS:
+                            try:
+                                shutil.rmtree(folder_path)
+                                print(f"🧹 [Garbage Collector] Deleted expired session: {session_folder}")
+                            except Exception as e:
+                                print(f"⚠️ [Garbage Collector] Failed to delete {session_folder}: {e}")
+        except Exception as e:
+            print(f"⚠️ [Garbage Collector] Error during cleanup scan: {e}")
+        
+        # 休眠到下一次扫描时间
+        time.sleep(CLEANUP_INTERVAL_SECONDS)
+
+def start_garbage_collector():
+    """启动后台守护线程"""
+    gc_thread = threading.Thread(target=cleanup_expired_sessions, daemon=True)
+    gc_thread.start()
+    print("🚀 [Garbage Collector] Background cleanup service started.")
 
-os.makedirs(PAPERS_DIR, exist_ok=True)
-os.makedirs(OUTPUT_DIR, exist_ok=True)
 
 # ==========================================
 # --- 🎨 Custom Purple Theme Definition ---
 # ==========================================
 colors.purple = colors.Color(
-    name="purple",
-    c50="#FAF5FF",
-    c100="#F3E8FF",
-    c200="#E9D5FF",
-    c300="#DAB2FF",
-    c400="#C084FC",
-    c500="#A855F7",
-    c600="#9333EA",
-    c700="#7E22CE",
-    c800="#6B21A8",
-    c900="#581C87",
-    c950="#3B0764",
+    name="purple", c50="#FAF5FF", c100="#F3E8FF", c200="#E9D5FF",
+    c300="#DAB2FF", c400="#C084FC", c500="#A855F7", c600="#9333EA",
+    c700="#7E22CE", c800="#6B21A8", c900="#581C87", c950="#3B0764",
 )
 
 class PurpleTheme(Soft):
-    def __init__(
-        self,
-        *,
-        primary_hue: colors.Color | str = colors.gray,
-        secondary_hue: colors.Color | str = colors.purple,
-        neutral_hue: colors.Color | str = colors.slate,
-        text_size: sizes.Size | str = sizes.text_lg,
-        font: fonts.Font | str | Iterable[fonts.Font | str] = (
-            fonts.GoogleFont("Outfit"), "Arial", "sans-serif",
-        ),
-        font_mono: fonts.Font | str | Iterable[fonts.Font | str] = (
-            fonts.GoogleFont("IBM Plex Mono"), "ui-monospace", "monospace",
-        ),
-    ):
+    def __init__(self, **kwargs):
         super().__init__(
-            primary_hue=primary_hue,
-            secondary_hue=secondary_hue,
-            neutral_hue=neutral_hue,
-            text_size=text_size,
-            font=font,
-            font_mono=font_mono,
+            primary_hue=colors.gray, secondary_hue=colors.purple, neutral_hue=colors.slate,
+            font=(fonts.GoogleFont("Outfit"), "Arial", "sans-serif"),
+            font_mono=(fonts.GoogleFont("IBM Plex Mono"), "ui-monospace", "monospace"),
         )
         super().set(
             background_fill_primary="*primary_50",
@@ -77,48 +117,53 @@ class PurpleTheme(Soft):
             body_background_fill="linear-gradient(135deg, *primary_200, *primary_100)",
             body_background_fill_dark="linear-gradient(135deg, *primary_900, *primary_800)",
             button_primary_text_color="white",
-            button_primary_text_color_hover="white",
             button_primary_background_fill="linear-gradient(90deg, *secondary_500, *secondary_600)",
             button_primary_background_fill_hover="linear-gradient(90deg, *secondary_600, *secondary_700)",
-            button_primary_background_fill_dark="linear-gradient(90deg, *secondary_600, *secondary_700)",
-            button_primary_background_fill_hover_dark="linear-gradient(90deg, *secondary_500, *secondary_600)",
             button_secondary_text_color="black",
-            button_secondary_text_color_hover="white",
             button_secondary_background_fill="linear-gradient(90deg, *primary_300, *primary_300)",
-            button_secondary_background_fill_hover="linear-gradient(90deg, *primary_400, *primary_400)",
-            button_secondary_background_fill_dark="linear-gradient(90deg, *primary_500, *primary_600)",
-            button_secondary_background_fill_hover_dark="linear-gradient(90deg, *primary_500, *primary_500)",
             slider_color="*secondary_500",
-            slider_color_dark="*secondary_600",
-            block_title_text_weight="600",
             block_border_width="3px",
             block_shadow="*shadow_drop_lg",
             button_primary_shadow="*shadow_drop_lg",
-            button_large_padding="11px",
-            color_accent_soft="*primary_100",
-            block_label_background_fill="*primary_200",
         )
 
 purple_theme = PurpleTheme()
 
+# ==========================================
+# --- 🚀 HTML Progress Bar Components ---
+# ==========================================
+def empty_progress_html(text="Waiting for action..."):
+    return f"""
+    <div class="custom-progress-container" style="background-color: transparent; border: 2px dashed rgba(168, 85, 247, 0.4);">
+        <div class="custom-progress-text" style="color: #A855F7;">{text}</div>
+    </div>
+    """
+
+def create_progress_html(percent, text, status="active"):
+    """
+    status: "active" (紫色滚动条纹), "success" (绿色), "error" (红色)
+    """
+    return f"""
+    <div class="custom-progress-container">
+        <div class="custom-progress-bar {status}" style="width: {percent}%;"></div>
+        <div class="custom-progress-text">{text} ({percent}%)</div>
+    </div>
+    """
+
 # ==========================================
 # --- ⚙️ Backend Logic & Functions ---
 # ==========================================
 
 def get_tree_html(dir_path):
-    """递归生成可折叠的 HTML 目录树"""
     if not os.path.exists(dir_path):
         return "<div style='margin-left: 15px; color: #888;'><i>Directory missing</i></div>"
         
     def build_html(current_path):
         html = ""
-        try:
-            items = sorted(os.listdir(current_path))
-        except Exception:
-            return ""
+        try: items = sorted(os.listdir(current_path))
+        except Exception: return ""
         
-        if not items:
-            return ""
+        if not items: return ""
             
         for item in items:
             item_path = os.path.join(current_path, item)
@@ -134,13 +179,14 @@ def get_tree_html(dir_path):
     content = build_html(dir_path)
     return content if content else "<div style='margin-left: 15px; color: #888;'><i>Empty directory</i></div>"
 
-def get_debug_info():
-    """返回渲染好的目录树 HTML 容器"""
-    papers_tree = get_tree_html(PAPERS_DIR)
-    output_tree = get_tree_html(OUTPUT_DIR)
+def get_debug_info(session_id):
+    papers_dir, output_dir, _ = get_user_dirs(session_id)
+    papers_tree = get_tree_html(papers_dir)
+    output_tree = get_tree_html(output_dir)
     
     html = f"""
     <div style="font-family: 'IBM Plex Mono', monospace; font-size: 13px; background-color: #1e1e1e; border: 1px solid #C084FC; border-radius: 8px; padding: 16px; max-height: 400px; overflow-y: auto;">
+        <div style="color: #888; margin-bottom: 8px;">Session ID: {session_id[:8]}...</div>
         <details open style="margin-bottom: 12px; cursor: pointer;">
             <summary style="outline: none; font-size: 15px; color: #A855F7;">📁 <b>papers/</b></summary>
             {papers_tree}
@@ -153,164 +199,133 @@ def get_debug_info():
     """
     return html
 
-def save_api_settings(api_key, api_base_url=None):
+def save_api_settings(api_key, api_base_url, session_id):
     if not api_key: 
-        return "❌ Key cannot be empty", get_debug_info(), False
+        return "❌ Key cannot be empty", get_debug_info(session_id), False, "", ""
+    success_msg = "✅ Key saved securely in session memory"
+    if api_base_url: success_msg += ", Base URL updated"
+    return success_msg, get_debug_info(session_id), True, api_key, api_base_url
+
+def save_pdf(file, session_id):
+    if file is None: 
+        return gr.update(visible=False), get_debug_info(session_id), 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(), False
-    try:
-        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)
+        papers_dir, _, _ = get_user_dirs(session_id)
+        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)
                 
-        file_path = os.path.join(PAPERS_DIR, os.path.basename(file.name))
+        file_path = os.path.join(papers_dir, os.path.basename(file.name))
         shutil.copy(file.name, file_path)
-        # 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
+        return gr.update(value=create_progress_html(100, f"✅ PDF Uploaded: {os.path.basename(file.name)}", "success"), visible=True), get_debug_info(session_id), True
     except Exception as e:
-        return f"❌ Error: {str(e)}", get_debug_info(), False
+        return gr.update(value=create_progress_html(0, f"❌ Error: {str(e)}", "error"), visible=True), get_debug_info(session_id), False
 
-def clear_pdf():
-    """清空 PDF 并锁定后续所有步骤"""
+def clear_pdf(session_id):
     try:
-        deleted_files = []
-        deleted_dirs = []
-        
-        pdf_names = [f for f in os.listdir(PAPERS_DIR) if f.endswith('.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)
-                
-        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(" ", "_")
-                
-                for item in os.listdir(OUTPUT_DIR):
-                    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)
-                            
-        if os.path.exists(ZIP_OUTPUT_PATH):
-            os.remove(ZIP_OUTPUT_PATH)
+        user_base = os.path.join(SESSIONS_BASE_DIR, session_id)
+        if os.path.exists(user_base):
+            shutil.rmtree(user_base) 
             
         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", 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
+        return gr.update(visible=False), gr.update(visible=False), get_debug_info(session_id), False, disable_btn, disable_btn, disable_btn, disable_btn
     except Exception as e:
-        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(progress=gr.Progress()):
+        return gr.update(value=create_progress_html(0, f"❌ Clear Error: {str(e)}", "error"), visible=True), gr.update(), get_debug_info(session_id), False, gr.update(), gr.update(), gr.update(), gr.update()
+
+def build_user_env(api_key, api_base_url, papers_dir, output_dir):
+    env = os.environ.copy()
+    env["MINERU_FORMULA_ENABLE"] = "false"
+    env["MINERU_TABLE_ENABLE"] = "false"
+    env["MINERU_DEVICE_MODE"] = "cpu"
+    env["MINERU_VIRTUAL_VRAM_SIZE"] = "8"
+    if api_key: env["GEMINI_API_KEY"] = api_key
+    if api_base_url: env["GEMINI_API_BASE_URL"] = api_base_url
+    env["USER_PAPERS_DIR"] = papers_dir
+    env["USER_OUTPUT_DIR"] = output_dir
+    return env
+
+def run_mineru_parsing_and_dag_gen(session_id, api_key, api_base_url, progress=gr.Progress()):
     no_change = gr.update()
     disable_btn = gr.update(interactive=False)
+    papers_dir, output_dir, _ = get_user_dirs(session_id)
     
-    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.", no_change, no_change, no_change, no_change
+    if not os.path.exists(papers_dir) or not any(f.endswith('.pdf') for f in os.listdir(papers_dir)):
+        yield gr.update(value=create_progress_html(0, "❌ No PDF file found", "error"), visible=True), get_debug_info(session_id), "No execution logs.", no_change, no_change, no_change, no_change
         return
 
     full_log = ""
     try:
-        env = os.environ.copy()
-        env["MINERU_FORMULA_ENABLE"] = "false"
-        env["MINERU_TABLE_ENABLE"] = "false"
-        env["MINERU_DEVICE_MODE"] = "cpu"
-        env["MINERU_VIRTUAL_VRAM_SIZE"] = "8"
-
-        command_mineru = ["mineru", "-p", PAPERS_DIR, "-o", OUTPUT_DIR]
-        full_log += "--- Mineru Executing ---\n"
+        env = build_user_env(api_key, api_base_url, papers_dir, output_dir)
+        command_mineru = ["mineru", "-p", papers_dir, "-o", output_dir]
+        full_log += f"--- Mineru Executing (Session: {session_id[:8]}) ---\n"
         
-        # 初始化进度条 (10%)
+        # 10%
         progress(0.1, desc="启动 Mineru 解析...")
-        yield "⏳ Executing Mineru parsing...", get_debug_info(), full_log, no_change, no_change, no_change, no_change
+        yield gr.update(value=create_progress_html(10, "⏳ Starting Mineru parsing...", "active"), visible=True), get_debug_info(session_id), 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
         )
         
-        # 维持在 30%，显示解析中
+        # 30%
         progress(0.3, desc="Mineru 正在解析 PDF...")
         for line in iter(process_mineru.stdout.readline, ''):
             full_log += line
-            yield "⏳ Executing Mineru parsing...", get_debug_info(), full_log, no_change, no_change, no_change, no_change
+            # ==========================================
+            # 🔔 双端输出 1：Mineru 阶段
+            # ==========================================
+            print(f"[Mineru | {session_id[:6]}] {line}", end="", flush=True)
+            yield gr.update(value=create_progress_html(30, "⏳ Mineru parsing PDF...", "active"), visible=True), get_debug_info(session_id), full_log, no_change, no_change, no_change, no_change
         process_mineru.stdout.close()
         returncode_mineru = process_mineru.wait()
 
         if returncode_mineru != 0:
             progress(1.0, desc="Mineru 解析失败")
-            yield f"❌ Mineru parsing failed (Exit Code: {returncode_mineru})", get_debug_info(), full_log, disable_btn, disable_btn, disable_btn, disable_btn
+            yield gr.update(value=create_progress_html(30, f"❌ Mineru failed (Code {returncode_mineru})", "error"), visible=True), get_debug_info(session_id), 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"
         
-        # Mineru 完成，进入 DAG 生成 (60%)
+        # 60%
         progress(0.6, desc="执行 DAG 生成...")
-        yield "⏳ Mineru parsing complete, executing DAG generation...", get_debug_info(), full_log, no_change, no_change, no_change, no_change
+        yield gr.update(value=create_progress_html(60, "⏳ Executing DAG generation...", "active"), visible=True), get_debug_info(session_id), 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
+            command_dag, env=env, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, text=True, bufsize=1
         )
         
-        # 维持在 80%
+        # 80%
         progress(0.8, desc="构建图结构中...")
         for line in iter(process_dag.stdout.readline, ''):
             full_log += line
-            yield "⏳ Executing DAG generation...", get_debug_info(), full_log, no_change, no_change, no_change, no_change
+            # ==========================================
+            # 🔔 双端输出 2：DAG 生成阶段
+            # ==========================================
+            print(f"[DAG | {session_id[:6]}] {line}", end="", flush=True)
+            yield gr.update(value=create_progress_html(80, "⏳ Building DAG...", "active"), visible=True), get_debug_info(session_id), full_log, no_change, no_change, no_change, no_change
         process_dag.stdout.close()
         returncode_dag = process_dag.wait()
         
-        # 解析完全成功 (100%)
         if returncode_dag == 0:
             progress(1.0, desc="解析与构建完成！")
-            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
+            yield gr.update(value=create_progress_html(100, "✅ Fully completed", "success"), visible=True), get_debug_info(session_id), full_log, enable_btn, enable_btn, enable_btn, enable_btn
         else:
             progress(1.0, desc="DAG 生成失败")
-            status = f"❌ DAG generation failed (Exit Code: {returncode_dag})"
-            yield status, get_debug_info(), full_log, disable_btn, disable_btn, disable_btn, disable_btn
+            yield gr.update(value=create_progress_html(80, "❌ DAG generation failed", "error"), visible=True), get_debug_info(session_id), full_log, disable_btn, disable_btn, disable_btn, disable_btn
 
     except Exception as e:
         progress(1.0, desc="发生异常")
-        error_log = full_log + f"\n[Global Exception] Exception occurred:\n{str(e)}"
-        yield "❌ Execution Exception", get_debug_info(), error_log, disable_btn, disable_btn, disable_btn, disable_btn
+        error_log = full_log + f"\n[Global Exception]:\n{str(e)}"
+        print(f"[Exception | {session_id[:6]}] {str(e)}", flush=True)
+        yield gr.update(value=create_progress_html(0, "❌ Execution Exception", "error"), visible=True), get_debug_info(session_id), error_log, disable_btn, disable_btn, disable_btn, disable_btn
 
-
-def run_final_generation(task_type="all", progress=gr.Progress()):
-    """
-    使用队列和多线程实时获取子进程的流式输出，并通过 yield 推送到前端界面。
-    引入 progress 参数更新进度条。
-    """
-    if not os.path.exists(OUTPUT_DIR):
-        yield "❌ Please run the parsing step first", get_debug_info(), "No output folder found.", gr.update(visible=False)
+def run_final_generation(task_type, session_id, api_key, api_base_url, progress=gr.Progress()):
+    papers_dir, output_dir, zip_path = get_user_dirs(session_id)
+    
+    if not os.path.exists(output_dir):
+        yield gr.update(value=create_progress_html(0, "❌ Please run parsing first", "error"), visible=True), get_debug_info(session_id), "No output folder found.", gr.update(visible=False)
         return
 
     scripts_to_run = []
@@ -318,18 +333,16 @@ def run_final_generation(task_type="all", progress=gr.Progress()):
     elif task_type == "poster": scripts_to_run = ["gen_poster.py"]
     elif task_type == "pr": scripts_to_run = ["gen_pr.py"]
     elif task_type == "all": scripts_to_run = ["gen_ppt.py", "gen_poster.py", "gen_pr.py"]
-    else: 
-        yield "❌ Unknown task type", get_debug_info(), "Invalid task_type.", gr.update(visible=False)
-        return
-
-    full_log = f"🚀 Preparing to start {len(scripts_to_run)} tasks...\n"
     
-    # 启动阶段 (10%)
+    full_log = f"🚀 Starting {len(scripts_to_run)} tasks for session {session_id[:8]}...\n"
+    print(f"[GEN Start | {session_id[:6]}] Starting {task_type.upper()}", flush=True)
+    
     progress(0.1, desc=f"启动 {task_type.upper()} 生成任务...")
-    yield f"⏳ Starting {task_type.upper()} generation...", get_debug_info(), full_log, gr.update(visible=False)
+    yield gr.update(value=create_progress_html(10, f"⏳ Starting {task_type.upper()}...", "active"), visible=True), get_debug_info(session_id), full_log, gr.update(visible=False)
 
     q = queue.Queue()
     processes = []
+    env = build_user_env(api_key, api_base_url, papers_dir, output_dir)
     
     def enqueue_output(out, script_name):
         for line in iter(out.readline, ''):
@@ -339,11 +352,7 @@ def run_final_generation(task_type="all", progress=gr.Progress()):
     try:
         for script in scripts_to_run:
             p = subprocess.Popen(
-                [sys.executable, script], 
-                stdout=subprocess.PIPE, 
-                stderr=subprocess.STDOUT, 
-                text=True, 
-                bufsize=1
+                [sys.executable, script], env=env, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, text=True, bufsize=1
             )
             processes.append((script, p))
             t = threading.Thread(target=enqueue_output, args=(p.stdout, script))
@@ -351,48 +360,43 @@ def run_final_generation(task_type="all", progress=gr.Progress()):
             t.start()
 
         active_processes = len(processes)
+        progress(0.5, desc=f"正在并行生成 {task_type.upper()}...")
         
-        # 生成阶段维持在 (50%)
-        progress(0.5, desc=f"正在并行生成 {task_type.upper()} 资产...")
         while active_processes > 0 or not q.empty():
             try:
                 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)
+                # ==========================================
+                # 🔔 双端输出 3：生成阶段
+                # ==========================================
+                print(f"[{script_name.upper()} | {session_id[:6]}] {line}", end="", flush=True)
+                yield gr.update(value=create_progress_html(50, f"⏳ Generating {task_type.upper()}...", "active"), visible=True), get_debug_info(session_id), full_log, gr.update(visible=False)
             except queue.Empty:
                 active_processes = sum(1 for _, p in processes if p.poll() is None)
 
-        success = True
-        for script, p in processes:
-            if p.returncode != 0:
-                success = False
-                full_log += f"\n❌ [Error] {script} returned non-zero exit code (Exit Code: {p.returncode})\n"
+        success = all(p.returncode == 0 for _, p in processes)
 
         if not success: 
             progress(1.0, desc="生成失败")
-            yield f"❌ {task_type.upper()} contains failed tasks, please check logs", get_debug_info(), full_log, gr.update(visible=False)
+            yield gr.update(value=create_progress_html(50, "❌ Tasks failed", "error"), visible=True), get_debug_info(session_id), full_log, gr.update(visible=False)
             return
 
         full_log += "\n📦 Zipping output directory...\n"
-        
-        # 压缩阶段 (90%)
         progress(0.9, desc="打包压缩结果...")
-        yield f"⏳ Zipping outputs...", get_debug_info(), full_log, gr.update(visible=False)
-        
-        zip_base_name = ZIP_OUTPUT_PATH.replace(".zip", "")
-        shutil.make_archive(zip_base_name, 'zip', OUTPUT_DIR)
+        yield gr.update(value=create_progress_html(90, "⏳ Zipping outputs...", "active"), visible=True), get_debug_info(session_id), full_log, gr.update(visible=False)
         
-        full_log += "✅ All tasks and zipping completed successfully.\n"
+        zip_base_name = zip_path.replace(".zip", "")
+        shutil.make_archive(zip_base_name, 'zip', output_dir)
         
-        # 完成阶段 (100%)
+        full_log += "✅ All tasks completed successfully.\n"
         progress(1.0, desc="全部完成！")
-        yield f"✅ {task_type.upper()} generated and zipped successfully", get_debug_info(), full_log, gr.update(value=ZIP_OUTPUT_PATH, visible=True)
+        yield gr.update(value=create_progress_html(100, f"✅ {task_type.upper()} Generated", "success"), visible=True), get_debug_info(session_id), full_log, gr.update(value=zip_path, visible=True)
         
     except Exception as e:
         progress(1.0, desc="发生全局异常")
-        error_log = full_log + f"\n[Global Exception] Exception occurred:\n{str(e)}"
-        yield "❌ Global exception during final generation", get_debug_info(), error_log, gr.update(visible=False)
-
+        error_log = full_log + f"\n[Global Exception]:\n{str(e)}"
+        print(f"[Exception | {session_id[:6]}] {str(e)}", flush=True)
+        yield gr.update(value=create_progress_html(0, "❌ Global exception", "error"), visible=True), get_debug_info(session_id), error_log, gr.update(visible=False)
 
 # ==========================================
 # --- 🚀 UI Configuration & Advanced CSS ---
@@ -452,8 +456,42 @@ body, .gradio-container {
     border-color: rgba(168, 85, 247, 0.3) !important; 
 }
 
-.gradio-group > div, 
-.gradio-group > .form { border-radius: 12px !important; }
+/* ================= 进度条自定义 CSS ================= */
+@keyframes progress-bar-stripes {
+    from { background-position: 1rem 0; }
+    to { background-position: 0 0; }
+}
+.custom-progress-container {
+    width: 100%; 
+    background-color: rgba(233, 213, 255, 0.3); 
+    border-radius: 12px; 
+    overflow: hidden; 
+    position: relative; 
+    height: 40px; 
+    border: 1px solid rgba(168, 85, 247, 0.3); 
+    box-shadow: inset 0 2px 4px rgba(0,0,0,0.05);
+}
+.custom-progress-bar {
+    height: 100%; 
+    border-radius: 12px; 
+    transition: width 0.4s ease;
+}
+.custom-progress-bar.active { 
+    background-color: #A855F7;
+    background-image: linear-gradient(45deg, rgba(255, 255, 255, .15) 25%, transparent 25%, transparent 50%, rgba(255, 255, 255, .15) 50%, rgba(255, 255, 255, .15) 75%, transparent 75%, transparent);
+    background-size: 1rem 1rem;
+    animation: progress-bar-stripes 1s linear infinite; 
+    box-shadow: 0 0 10px rgba(168, 85, 247, 0.5);
+}
+.custom-progress-bar.success { background-image: none; background-color: #10B981; box-shadow: 0 0 10px rgba(16, 185, 129, 0.5);}
+.custom-progress-bar.error { background-image: none; background-color: #EF4444; box-shadow: 0 0 10px rgba(239, 68, 68, 0.5);}
+.custom-progress-text {
+    position: absolute; top: 0; left: 0; width: 100%; height: 100%; 
+    display: flex; align-items: center; justify-content: center; 
+    font-weight: 600; color: #581C87; font-size: 14px; 
+    text-shadow: 0 0 4px rgba(255,255,255,0.8);
+}
+.dark .custom-progress-text { color: #E9D5FF; text-shadow: 0 0 4px rgba(0,0,0,0.8); }
 
 #pdf-upload-box {
     border: 2px dashed rgba(192, 132, 252, 0.6) !important;
@@ -503,7 +541,7 @@ body, .gradio-container {
     margin-right: auto !important; 
     margin-top: 10px !important;
     margin-bottom: 10px !important;
-    display: block !important;      
+    display: block !important;     
     transition: transform 0.3s cubic-bezier(0.175, 0.885, 0.32, 1.275), box-shadow 0.3s ease !important;
     box-shadow: 0 4px 15px rgba(126, 34, 206, 0.3) !important;
     cursor: pointer !important; 
@@ -543,7 +581,6 @@ 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;
@@ -565,13 +602,15 @@ body, .gradio-container {
 ::-webkit-scrollbar-track { background: rgba(168, 85, 247, 0.05); border-radius: 4px; }
 ::-webkit-scrollbar-thumb { background: linear-gradient(135deg, #A855F7, #C084FC); border-radius: 4px; }
 ::-webkit-scrollbar-thumb:hover { background: linear-gradient(135deg, #9333EA, #A855F7); }
-
 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
+    session_id_state = gr.State("")
+    user_api_key_state = gr.State("")
+    user_api_base_state = gr.State("")
+    
     api_saved_state = gr.State(False)
     pdf_ready_state = gr.State(False)
 
@@ -580,126 +619,126 @@ with gr.Blocks(theme=purple_theme, css=custom_css) as demo:
         gr.Markdown("One-click parsing of academic PDFs, DAG structuring, and multi-modal asset generation.", elem_id="subtitle")
         
         with gr.Row():
-            # ================= LEFT COLUMN: SETTINGS & ACTIONS =================
             with gr.Column(scale=1):
-                
                 # 1. API Configuration
                 with gr.Group(elem_classes="gradio-group"):
                     gr.Markdown("### ⚙️ 1. Global API Configuration")
                     with gr.Row():
-                        key_input = gr.Textbox(label="API Key", type="password", placeholder="sk-...", scale=1)
+                        key_input = gr.Textbox(label="Gemini 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"
-                    )
-                
+                    api_status = gr.Textbox(show_label=False, interactive=False, elem_classes="status-text")
+
                 # 2. Document Parsing
                 with gr.Group(elem_classes="gradio-group"):
                     gr.Markdown("### 📄 2. Document Parsing")
-                    
-                    pdf_input = gr.File(
-                        label="Upload Document", 
-                        file_types=[".pdf"], 
-                        elem_id="pdf-upload-box"
-                    )
-                    
-                    # 默认锁定
+                    pdf_input = gr.File(label="Upload Document", file_types=[".pdf"], elem_id="pdf-upload-box")
                     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"
-                    )
+                    # 默认隐藏进度条
+                    parse_progress = gr.HTML(visible=False)
 
                 # 3. Asset Generation
                 with gr.Group(elem_classes="gradio-group"):
                     gr.Markdown("### 🎯 3. Asset Generation")
-                    gr.Markdown("Generate final formats based on DAG structure:")
-                    
                     with gr.Row(elem_classes="action-row"):
-                        # 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)
-                    
-                    # 默认锁定
                     gen_all_btn = gr.Button("✨ Generate All Assets (ALL)", elem_classes="primary-action-btn", interactive=False)
+                    
+                    # 默认隐藏进度条
+                    gen_progress = gr.HTML(visible=False)
 
-            # ================= RIGHT COLUMN: OUTPUTS & LOGS =================
             with gr.Column(scale=1):
-                
                 # 4. Results & Downloads
                 with gr.Group(elem_classes="gradio-group"):
                     gr.Markdown("### 📦 Generation Results & Download")
-                    gen_status = gr.Textbox(
-                        show_label=False, placeholder="No generation task currently...", lines=2, interactive=False, elem_classes="status-text"
+                    
+                    download_placeholder = gr.HTML(
+                        '''
+                        <div style="display: flex; flex-direction: column; align-items: center; justify-content: center; height: 160px; border: 2px dashed rgba(192, 132, 252, 0.6); border-radius: 16px; background-color: rgba(250, 245, 255, 0.5); color: #9333EA; text-align: center; transition: all 0.3s ease;">
+                            <span style="font-size: 32px; margin-bottom: 8px;">📦</span>
+                            <span style="font-weight: 600; font-size: 16px; margin-bottom: 4px;">Awaiting Generation</span>
+                            <span style="font-size: 13px; color: #A855F7; opacity: 0.8;">Generated assets will appear here as a downloadable ZIP archive.</span>
+                        </div>
+                        '''
                     )
+                    
                     download_file = gr.File(label="📥 Get Final Zip Archive", interactive=False, visible=False)
 
-                # 5. Debugging & Terminal
+                # 5. Debugging
                 with gr.Group(elem_classes="gradio-group"):
                     gr.Markdown("### 🛠️ Developer Monitoring (Debug Only)")
                     with gr.Tabs():
                         with gr.Tab("📜 Terminal Stream"):
-                            cmd_logs = gr.Textbox(
-                                label="Stdout / Stderr", placeholder="Waiting for task to start...", lines=14, interactive=False, elem_classes="log-box" 
-                            )
-                        
+                            cmd_logs = gr.Textbox(show_label=False, lines=14, interactive=False, elem_classes="log-box")
                         with gr.Tab("🔍 System Snapshot"):
                             refresh_btn = gr.Button("🔄 Refresh Directory Tree") 
-                            debug_view = gr.HTML(
-                                value=get_debug_info()
-                            )
+                            debug_view = gr.HTML()
+
+# ================= LOGIC BINDINGS =================
+
+    def init_app_for_user():
+        new_session_id = str(uuid.uuid4())
+        debug_html = get_debug_info(new_session_id)
+        return new_session_id, debug_html
+
+    demo.load(fn=init_app_for_user, inputs=None, outputs=[session_id_state, debug_view])
 
-    # ================= LOGIC BINDINGS =================
-    
-    # [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]
+        inputs=[key_input, api_base_url_input, session_id_state], 
+        outputs=[api_status, debug_view, api_saved_state, user_api_key_state, user_api_base_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])
+    pdf_input.upload(
+        fn=save_pdf, 
+        inputs=[pdf_input, session_id_state], 
+        outputs=[parse_progress, 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]
+        inputs=[session_id_state],
+        outputs=[parse_progress, gen_progress, debug_view, pdf_ready_state, gen_ppt_btn, gen_poster_btn, gen_pr_btn, gen_all_btn]
     )
     
-    # [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]
+        inputs=[session_id_state, user_api_key_state, user_api_base_state],
+        outputs=[parse_progress, debug_view, cmd_logs, gen_ppt_btn, gen_poster_btn, gen_pr_btn, gen_all_btn]
     )
 
-    # [步骤 3] 资产生成
-    # 将 progress 对象注入到包装函数中
-    def trigger_gen_ppt(progress=gr.Progress()): yield from run_final_generation("ppt", progress)
-    def trigger_gen_poster(progress=gr.Progress()): yield from run_final_generation("poster", progress)
-    def trigger_gen_pr(progress=gr.Progress()): yield from run_final_generation("pr", progress)
-    def trigger_gen_all(progress=gr.Progress()): yield from run_final_generation("all", progress)
-
-    gen_ppt_btn.click(fn=trigger_gen_ppt, outputs=[gen_status, debug_view, cmd_logs, download_file])
-    gen_poster_btn.click(fn=trigger_gen_poster, outputs=[gen_status, debug_view, cmd_logs, download_file])
-    gen_pr_btn.click(fn=trigger_gen_pr, outputs=[gen_status, debug_view, cmd_logs, download_file])
-    gen_all_btn.click(fn=trigger_gen_all, outputs=[gen_status, debug_view, cmd_logs, download_file])
+    def trigger_gen_ppt(sid, ak, ab, progress=gr.Progress()): yield from run_final_generation("ppt", sid, ak, ab, progress)
+    def trigger_gen_poster(sid, ak, ab, progress=gr.Progress()): yield from run_final_generation("poster", sid, ak, ab, progress)
+    def trigger_gen_pr(sid, ak, ab, progress=gr.Progress()): yield from run_final_generation("pr", sid, ak, ab, progress)
+    def trigger_gen_all(sid, ak, ab, progress=gr.Progress()): yield from run_final_generation("all", sid, ak, ab, progress)
+
+    gen_ppt_btn.click(fn=trigger_gen_ppt, inputs=[session_id_state, user_api_key_state, user_api_base_state], outputs=[gen_progress, debug_view, cmd_logs, download_file])
+    gen_poster_btn.click(fn=trigger_gen_poster, inputs=[session_id_state, user_api_key_state, user_api_base_state], outputs=[gen_progress, debug_view, cmd_logs, download_file])
+    gen_pr_btn.click(fn=trigger_gen_pr, inputs=[session_id_state, user_api_key_state, user_api_base_state], outputs=[gen_progress, debug_view, cmd_logs, download_file])
+    gen_all_btn.click(fn=trigger_gen_all, inputs=[session_id_state, user_api_key_state, user_api_base_state], outputs=[gen_progress, debug_view, cmd_logs, download_file])
     
-    refresh_btn.click(fn=get_debug_info, outputs=debug_view)
+    refresh_btn.click(fn=get_debug_info, inputs=[session_id_state], outputs=debug_view)
+
+    def toggle_empty_placeholder(file_val):
+        return gr.update(visible=(file_val is None))
+
+    download_file.change(
+        fn=toggle_empty_placeholder,
+        inputs=[download_file],
+        outputs=[download_placeholder]
+    )
 
 if __name__ == "__main__":
-    demo.launch(debug=True)
+    start_garbage_collector()
+    # 并发放宽至 5
+    # demo.queue(default_concurrency_limit=5).launch()
+    demo.launch(server_name="0.0.0.0", server_port=7860)