Update app.py

app.py

@@ -3,10 +3,8 @@
 from core.meta_extractor import extract_golden_metadata
 from core.rag_engine import initialize_rag
 from core.clone_architect import get_clone_architect
-
 import streamlit as st
 
-# 🚨 核心要求：必须在引入任何可能包含 Streamlit 动作的本地模块前，先设置 Page Config！
 st.set_page_config(page_title="Agent Article", page_icon="📊", layout="wide")
 
 import warnings
@@ -25,19 +23,18 @@ import pandas as pd
 import numpy as np
 import threading
 
-# 引入本地模块
 from config import APP_CONFIG, AMJ_PROMPTS, GENERIC_PROMPTS
 from utils.file_parser import extract_text_from_files, extract_data_from_files
 from utils.docx_builder import text_to_docx
 from core.pipeline_agent import SSciPipelineAgent
 
-# ==========================================
-# 0. 全局常量、工具函数与核心环境准备
-# ==========================================
 os.environ["STREAMLIT_SERVER_ENABLE_CORS"] = "false"
 os.environ["STREAMLIT_SERVER_ENABLE_XSRF_PROTECTION"] = "false"
 
-# 🚨 优化：去除重复硬编码，统一从 APP_CONFIG 读取黑名单
+WORKSPACE_DIR = APP_CONFIG.get("workspace_dir", ".workspace")
+os.makedirs(WORKSPACE_DIR, exist_ok=True)
+CHECKPOINT_FILE = os.path.join(WORKSPACE_DIR, "snapshot.json")
+
 META_SECTIONS_BLACKLIST = set(APP_CONFIG["blacklists"]["meta_sections"])
 
 def clean_trailing_references(text, keywords=("reference", "references", "reference list", "参考文献")):
@@ -60,65 +57,40 @@ if not global_s2_key:
     try: global_s2_key = st.secrets.get("S2_API_KEY", "")
     except: pass
 
-WORKSPACE_DIR = APP_CONFIG.get("workspace_dir", ".workspace")
-os.makedirs(WORKSPACE_DIR, exist_ok=True)
-CHECKPOINT_FILE = os.path.join(WORKSPACE_DIR, "snapshot.json")
-
-WECHAT_BLOCKER_JS = """
-<script>
-try {
-    var ua = navigator.userAgent.toLowerCase();
-    if (ua.indexOf('micromessenger') !== -1) {
-        var parentDoc = window.parent.document;
-        if (!parentDoc.getElementById('wx-blocker')) {
-            var overlay = parentDoc.createElement('div');
-            overlay.id = 'wx-blocker';
-            overlay.style.cssText = 'position:fixed;top:0;left:0;width:100vw;height:100vh;background:#1a365d;z-index:9999999;color:white;display:flex;flex-direction:column;align-items:center;justify-content:center;text-align:center;font-family:sans-serif;padding:30px;box-sizing:border-box;';
-            overlay.innerHTML = '<h2 style="font-size:26px;margin-bottom:15px;font-weight:bold;">环境受限 / Restricted</h2><p style="font-size:18px;line-height:1.6;color:#e2e8f0;max-width:100%;">为了保障最终双语 ZIP 压缩包正常下载，<b>本系统禁止在微信内直接运行</b>。<br><br>👉 <b>请点击右上角 [···] 菜单，选择“在浏览器打开” (Open in Browser)</b></p>';
-            parentDoc.body.appendChild(overlay);
-        }
-    }
-} catch (e) {}
-</script>
-"""
+WECHAT_BLOCKER_JS = """<script>try{var ua=navigator.userAgent.toLowerCase();if(ua.indexOf('micromessenger')!==-1){var parentDoc=window.parent.document;if(!parentDoc.getElementById('wx-blocker')){var overlay=parentDoc.createElement('div');overlay.id='wx-blocker';overlay.style.cssText='position:fixed;top:0;left:0;width:100vw;height:100vh;background:#1a365d;z-index:9999999;color:white;display:flex;flex-direction:column;align-items:center;justify-content:center;text-align:center;font-family:sans-serif;padding:30px;box-sizing:border-box;';overlay.innerHTML='<h2 style="font-size:26px;margin-bottom:15px;font-weight:bold;">环境受限 / Restricted</h2><p style="font-size:18px;line-height:1.6;color:#e2e8f0;max-width:100%;">为了保障最终双语 ZIP 压缩包正常下载，<b>本系统禁止在微信内直接运行</b>。<br><br>👉 <b>请点击右上角 [···] 菜单，选择“在浏览器打开” (Open in Browser)</b></p>';parentDoc.body.appendChild(overlay);}}}catch(e){}</script>"""
 components.html(WECHAT_BLOCKER_JS, height=0, width=0)
 
-# ==========================================
-# 🚨 访问控制与原生并发锁
-# ==========================================
 def check_password():
     expected_password = os.environ.get("APP_PASSWORD", "")
     if not expected_password:
         try: expected_password = st.secrets.get("APP_PASSWORD", "")
         except: pass
         
-    if not expected_password: return True
-
-    if "password_correct" not in st.session_state:
+    if not expected_password: 
+        return True
+        
+    if "password_correct" not in st.session_state: 
         st.session_state["password_correct"] = False
-
+        
     if not st.session_state["password_correct"]:
         st.markdown("<h3 style='text-align: center;'>🔒 系统已锁定 / System Locked</h3>", unsafe_allow_html=True)
         pwd = st.text_input("Access Token", type="password", key="pwd_input")
         if st.button("Unlock"):
-            if pwd == expected_password:
+            if pwd == expected_password: 
                 st.session_state["password_correct"] = True
                 st.rerun()
-            else:
+            else: 
                 st.error("❌ ���证错误，拒绝访问。")
         st.stop()
     return True
 
 @st.cache_resource
-def get_global_lock():
+def get_global_lock(): 
     return threading.Lock()
 
 check_password()
 global_lock = get_global_lock()
 
-# ==========================================
-# 1. 鲁棒的蒙特卡洛实证模拟引擎
-# ==========================================
 def generate_psd_corr_matrix(n_constructs, max_corr=0.5):
     if n_constructs == 1: return np.array([[1.0]])
     base_corr = np.random.uniform(0.2, max_corr - 0.1) 
@@ -150,14 +122,16 @@ def calculate_validity_metrics(df, constructs_info, true_loadings):
     return pd.DataFrame(metrics)
 
 def generate_empirical_data_robust(construct_names=None, n=300, scale_points=7):
-    if not construct_names or len(construct_names) < 2:
+    if not construct_names or len(construct_names) < 2: 
         construct_names = ["Independent_Var", "Mediator_Var", "Dependent_Var"]
+        
     n_constructs = len(construct_names)
     corr = generate_psd_corr_matrix(n_constructs, max_corr=0.55)
     latent_scores = np.random.multivariate_normal(mean=[0]*n_constructs, cov=corr, size=n)
     data, true_loadings, constructs_info = {}, {}, []
     target_mean = scale_points / 2.0 + 0.5
     target_std = scale_points / 5.0
+    
     for idx, c_name in enumerate(construct_names):
         true_score = latent_scores[:, idx]
         n_items = np.random.randint(3, 6) 
@@ -173,41 +147,30 @@ def generate_empirical_data_robust(construct_names=None, n=300, scale_points=7):
             data[col_name] = item_int
             item_names.append(col_name)
         constructs_info.append({"name": c_name, "items": item_names})
+        
     df = pd.DataFrame(data)
     validity_df = calculate_validity_metrics(df, constructs_info, true_loadings)
     mean_scores = pd.DataFrame({c['name']: df[c['items']].mean(axis=1) for c in constructs_info})
     fl_matrix = mean_scores.corr().astype(object) 
+    
     for i, row in validity_df.iterrows():
         c_name = row['Construct']
         sqrt_ave = np.sqrt(row['AVE'])
         fl_matrix.loc[c_name, c_name] = f"({sqrt_ave:.3f})"
+        
     for i in range(len(fl_matrix.columns)):
-        for j in range(i + 1, len(fl_matrix.columns)):
+        for j in range(i + 1, len(fl_matrix.columns)): 
             fl_matrix.iloc[i, j] = ""
+            
     desc_stats = mean_scores.mean().to_frame('Mean').join(mean_scores.std().to_frame('SD'))
     return df, validity_df, desc_stats.join(fl_matrix)
 
 def resolve_empirical_data(extracted_vars=None):
     df, validity_df, fl_matrix_df = generate_empirical_data_robust(extracted_vars)
-    simulated_md_prompt = f"""
-### Table 1: Reliability and Convergent Validity
-{validity_df.to_markdown(index=False)}
-
-### Table 2: Descriptive Statistics and Discriminant Validity (Fornell-Larcker Criterion)
-*Note: Diagonal elements in parentheses are the square roots of AVE.*
-{fl_matrix_df.to_markdown()}
-
-[STRICT EMPIRICAL INSTRUCTION]
-1. ZERO HALLUCINATION: You MUST report the exact metrics (Cronbach's α, CR, AVE, correlations) provided in the tables above.
-2. VALIDITY DEFENSE: Explicitly state that Convergent Validity is established because all AVEs > 0.5 and CRs > 0.7.
-3. DISCRIMINANT DEFENSE: Explicitly state that Discriminant Validity is satisfied per the Fornell-Larcker criterion, as the square root of AVE (diagonal values) exceed inter-construct correlations.
-"""
+    simulated_md_prompt = f"""### Table 1: Reliability and Convergent Validity\n{validity_df.to_markdown(index=False)}\n### Table 2: Descriptive Statistics and Discriminant Validity (Fornell-Larcker Criterion)\n*Note: Diagonal elements in parentheses are the square roots of AVE.*\n{fl_matrix_df.to_markdown()}\n[STRICT EMPIRICAL INSTRUCTION]\n1. ZERO HALLUCINATION: You MUST report the exact metrics (Cronbach's α, CR, AVE, correlations) provided in the tables above.\n2. VALIDITY DEFENSE: Explicitly state that Convergent Validity is established because all AVEs > 0.5 and CRs > 0.7.\n3. DISCRIMINANT DEFENSE: Explicitly state that Discriminant Validity is satisfied per the Fornell-Larcker criterion, as the square root of AVE (diagonal values) exceed inter-construct correlations."""
     status_msg = "✅ 已成功启用稳健信效度生成引擎，AVE与Fornell-Larcker准则均验证达标。"
     return df, simulated_md_prompt, status_msg
 
-# ==========================================
-# 2. UI 界面布局与初始化
-# ==========================================
 if "is_running" not in st.session_state: st.session_state.is_running = False
 if "start_pipeline" not in st.session_state: st.session_state.start_pipeline = False
 if "resume_pipeline" not in st.session_state: st.session_state.resume_pipeline = False
@@ -226,8 +189,10 @@ if os.path.exists(CHECKPOINT_FILE) and not (st.session_state.is_running or st.se
         st.session_state.start_pipeline = False
         st.rerun()
     if colB.button("🗑️ 清除临时数据 (Clear Garbage)", key="clear_task"):
-        try: os.remove(CHECKPOINT_FILE)
-        except OSError: pass
+        try: 
+            os.remove(CHECKPOINT_FILE)
+        except OSError: 
+            pass
         st.rerun()
     st.markdown("---")
 
@@ -239,14 +204,12 @@ tab_wizard, tab_clone = st.tabs(["✨ 智能向导模式 (Smart Wizard)", "🧬
 
 with tab_wizard:
     st.markdown("### 🛠️ 人在环路：研究资料配置")
-    prompt_style_ui = st.radio("1. 请选择期望的写作规范与提示词引擎：", 
-                               options=["实证研究 (AMJ Style)", "自适应 (Adaptive Style)"], 
-                               index=0, disabled=st.session_state.is_running)
-    uploaded_papers_wiz = st.file_uploader("2. 上传参考论文 (可选，支持 PDF/DOCX，建议1-20篇)", type=["pdf", "doc", "docx"], accept_multiple_files=True, disabled=st.session_state.is_running)
-    uploaded_data_wiz = st.file_uploader("3. 上传科研数据 (可选，支持 CSV/Excel，用于防幻觉)", type=["csv", "xlsx", "xls"], disabled=st.session_state.is_running)
+    prompt_style_ui = st.radio("1. 请选择期望的写作规范与提示词引擎：", options=["实证研究 (AMJ Style)", "自适应 (Adaptive Style)"], index=0, disabled=st.session_state.is_running)
+    uploaded_papers_wiz = st.file_uploader("2. 上传参考论文 (可选，支持 PDF/DOCX，建议1-5篇)", type=["pdf", "doc", "docx"], accept_multiple_files=True, disabled=st.session_state.is_running)
+    uploaded_data_wiz = st.file_uploader("3. 上传统计分析结果 (可选，支持 CSV/Excel，用于防幻觉)", type=["csv", "xlsx", "xls"], disabled=st.session_state.is_running)
     
     if st.button("🚀 启动智能向导生成", disabled=st.session_state.is_running, key="btn_wiz"):
-        if not global_lock.acquire(blocking=False):
+        if not global_lock.acquire(blocking=False): 
             st.warning("⏳ 有小伙伴正在创作，您在15-20 分钟后就能正常使用啦")
         else:
             global_lock.release()
@@ -256,32 +219,35 @@ with tab_wizard:
 
 with tab_clone:
     st.markdown("### 🧬 高级克隆：框架与风格复刻")
-    context_differences = st.text_area("1. 研究背景与核心差异（Context & Differences）（必填）", placeholder="简述您的新研究与参考论文的核心区别...", max_chars=2000, disabled=st.session_state.is_running)
+    context_differences = st.text_area("1. 研究背景与核心差异（Context & Differences）（必填，建议 1-2000 字）", placeholder="简述您的新研究与参考论文在数据、案例或背景上的核心区别...", max_chars=2000, disabled=st.session_state.is_running)
     
     if "extracted_sections" not in st.session_state: 
         st.session_state.extracted_sections = ["Introduction", "Literature Review", "Methodology", "Results", "Discussion"]
     if "last_uploaded_clone_ref" not in st.session_state: 
         st.session_state.last_uploaded_clone_ref = None
 
-    uploaded_reference_clone = st.file_uploader("2. 上传范例论文 (必填，单篇 PDF/DOCX)", type=["pdf", "doc", "docx"], key="clone_ref", disabled=st.session_state.is_running)
+    uploaded_reference_clone = st.file_uploader("2. 上传范例论文 (必填，单篇 PDF/DOCX，系统自动解构目录)", type=["pdf", "doc", "docx"], key="clone_ref", disabled=st.session_state.is_running)
     
     if uploaded_reference_clone and global_ds_key:
         if st.session_state.last_uploaded_clone_ref != uploaded_reference_clone.name:
-            with st.spinner("🔍 正在智能提取该范例论文的真实目录结构..."):
+            with st.spinner("🔍 正在智能提取该范例论文的真实目录结构，请稍候..."):
                 temp_agent = SSciPipelineAgent(global_ds_key, global_s2_key)
                 ref_text_raw = extract_text_from_files([uploaded_reference_clone])
                 dynamic_sections = temp_agent.extract_document_structure(ref_text_raw)
                 st.session_state.extracted_sections = dynamic_sections
                 st.session_state.last_uploaded_clone_ref = uploaded_reference_clone.name
-                st.success(f"✅ 目录提取成功：{', '.join(dynamic_sections)}")
+                st.success(f"✅ 目录提取成功，将自动全量克隆以下结构：{', '.join(dynamic_sections)}")
     
-    uploaded_core_papers_clone = st.file_uploader("3. 上传参考论文 (可选，支持 PDF/DOCX，建议1-20篇)", type=["pdf", "doc", "docx"], accept_multiple_files=True, key="clone_core_papers", disabled=st.session_state.is_running)
-    uploaded_data_clone = st.file_uploader("4. 上传科研数据 (可选)", type=["csv", "xlsx", "xls"], key="clone_data", disabled=st.session_state.is_running)
+    uploaded_core_papers_clone = st.file_uploader("3. 上传参考论文 (可选，用于提升理论深度，建议1-5篇)", type=["pdf", "doc", "docx"], accept_multiple_files=True, key="clone_core_papers", disabled=st.session_state.is_running)
+    uploaded_data_clone = st.file_uploader("4. 上传科研数据 (可选，支持 CSV/Excel，用于防幻觉)", type=["csv", "xlsx", "xls"], key="clone_data", disabled=st.session_state.is_running)
     
     if st.button("🚀 启动克隆级生成", disabled=st.session_state.is_running, key="btn_clone"):
-        if not context_differences.strip(): st.error("🚨 启动克隆模式必须填写研究背景与核心差异！")
-        elif not uploaded_reference_clone: st.error("🚨 启动克隆模式必须上传范例论文！")
-        elif not global_lock.acquire(blocking=False): st.warning("⏳ 有小伙伴正在创作，您在15-20 分钟后就能正常使用啦")
+        if not context_differences.strip(): 
+            st.error("🚨 启动克隆模式必须填写研究背景与核心差异！")
+        elif not uploaded_reference_clone: 
+            st.error("🚨 启动克隆模式必须上传范例论文！")
+        elif not global_lock.acquire(blocking=False): 
+            st.warning("⏳ 有小伙伴正在创作，您在15-20 分钟后就能正常使用啦")
         else:
             global_lock.release()
             st.session_state.run_mode = "clone"
@@ -306,11 +272,8 @@ def render_monitor(agent, phase, current_sec="", details=""):
 
 FLAG_MAP = {"Simplified Chinese (简体中文)": "🇨🇳", "Russian (Русский)": "🇷🇺", "Kazakh (Қазақша)": "🇰🇿"}
 
-# ==========================================
-# 3. 后台流水线执行
-# ==========================================
 if st.session_state.start_pipeline or st.session_state.resume_pipeline:
-    if not global_ds_key:
+    if not global_ds_key: 
         st.error("❌ DEEPSEEK_API_KEY missing. 请配置API KEY。")
         st.stop()
         
@@ -347,7 +310,7 @@ if st.session_state.start_pipeline or st.session_state.resume_pipeline:
                 is_resume = False
                 
         if is_resume and loaded_state:
-            pb = progress_bar.progress(50, text="🔄 正在从安全快照恢复状态...")
+            pb = progress_bar.progress(50, text="🔄 正在从本地安全快照恢复状态...")
             state = loaded_state
             topic = state["topic"]
             keywords = state["keywords"]
@@ -401,9 +364,9 @@ if st.session_state.start_pipeline or st.session_state.resume_pipeline:
                 if uploaded_data_clone: 
                     user_data_md = extract_data_from_files(uploaded_data_clone)
                 if uploaded_core_papers_clone:
-                    if len(uploaded_core_papers_clone) > 20:
-                        st.warning("⚠️ 核心参考论文超过 20 篇，系统仅截取前 20 篇。")
-                        uploaded_core_papers_clone = uploaded_core_papers_clone[:20]
+                    if len(uploaded_core_papers_clone) > 5:
+                        st.warning("⚠️ 核心参考论文超过 5 篇，系统仅截取前 5 篇。")
+                        uploaded_core_papers_clone = uploaded_core_papers_clone[:5]
                     user_lit_context = extract_text_from_files(uploaded_core_papers_clone)
 
             if not user_data_md:
@@ -577,28 +540,29 @@ if st.session_state.start_pipeline or st.session_state.resume_pipeline:
             adjustment = -accumulated_deviation / remaining_sections if remaining_sections > 0 else 0
             adjusted_target = min(int(SAFE_CHUNK_WORDS * 1.5), max(50, int(sec['target_words'] + adjustment)))
             
-            # 🚨 格式约束指令注入位置 🚨
-            lower_sec = sec_title.lower()
-            adaptive_heading_rule = ""
+            is_quarantined = any(kw in sec_title.lower() for kw in ["abstract", "keyword", "摘要", "关键字", "关键词", "conclusion", "结论", "заключение", "қорытынды", "references"])
             
-            # 1. 引言平铺规则
-            if any(kw in lower_sec for kw in ["introduction", "引言", "绪论"]):
-                adaptive_heading_rule = "[STRICT FORMAT RULE] DO NOT use any subheadings (## or ###) in this Introduction section. Write strictly in continuous paragraphs only."
+            if is_quarantined: 
+                quarantine_rule = "【CRITICAL】: NO IN-TEXT CITATIONS ALLOWED IN THIS SECTION."
             else:
-                # 2. 正文最高三级规则
-                if adjusted_target > 800:
-                    adaptive_heading_rule = "Use Level 2 (##) and Level 3 (###) subheadings to logically structure your arguments. ABSOLUTELY NEVER use Level 4 (####)."
-                elif adjusted_target > 400:
-                    adaptive_heading_rule = "Use Level 2 (##) subheadings to organize your content. ABSOLUTELY NEVER use Level 4 (####)."
-                else:
-                    adaptive_heading_rule = "NEVER use Level 4 (####) headings."
+                assigned_tags_str = ", ".join(allocation_map.get(display_name, [])) or "any relevant citations"
+                quarantine_rule = f"【CITATION GUIDANCE】: You are strongly encouraged to integrate citations: {assigned_tags_str}. <AVAILABLE_CITATIONS>{refs_text}</AVAILABLE_CITATIONS>"
+            
+            heading_instruction = f"Write the content for the section '{sec_title}'. DO NOT output the main heading '# {sec_title}' itself."
+            memory_block = f"\n【GLOBAL PAPER MEMORY】\n{agent._build_memory_context(sec_title, 2500)}\n" if agent._build_memory_context(sec_title, 2500) else ""
+
+            if hasattr(st.session_state, "rag_engine") and st.session_state.rag_engine is not None:
+                context_snippets = st.session_state.rag_engine.get_relevant_context(sec_title)
+                if context_snippets: 
+                    memory_block += f"\n【TARGETED RAG RESULTS】\n{context_snippets}\n"
 
-            # 3. 表头正文化规则
-            table_rule = ""
-            if sec["is_method_or_result"]:
-                table_rule = "[CRITICAL DATA FORMATTING] You MUST present ALL statistical data using STANDARD MARKDOWN TABLES. Also, all Table titles (e.g., Table 1: Analysis) MUST be plain text paragraphs. NEVER use heading markers (#, ##, ###) for table captions."
+            lang_hint = "ru" if "Русский" in target_lang else ("kk" if "Қазақша" in target_lang else "en")
+            num_paragraphs = agent._calc_paragraph_count(adjusted_target, lang_hint)
+            
+            adaptive_heading_rule = f"Use Level 2 (##) and Level 3 (###) subheadings." if adjusted_target > 800 else (f"Use Level 2 (##) subheadings." if adjusted_target > 400 else "")
+            table_rule = "[CRITICAL DATA FORMATTING] You MUST present ALL statistical data using STANDARD MARKDOWN TABLES." if sec["is_method_or_result"] else ""
 
-            length_directive = f"[LANGUAGE MANDATE] Output strictly in ENGLISH.\n[LENGTH MANDATE - STRICT] Write between {int(adjusted_target * 0.9)} and {int(adjusted_target * 1.1)} words (approx {agent._calc_paragraph_count(adjusted_target, 'en')} paragraphs).\n{adaptive_heading_rule}\n{table_rule}\n[FINAL CHECK] Do not output a word count statement."
+            length_directive = f"[LENGTH MANDATE - STRICT] Write between {int(adjusted_target * 0.9)} and {int(adjusted_target * 1.1)} words (approx {num_paragraphs} paragraphs).\n{adaptive_heading_rule}\n{table_rule}\n[FINAL CHECK] Do not output a word count statement."
 
             is_clone_target = False
             active_clone_key = None
@@ -606,8 +570,10 @@ if st.session_state.start_pipeline or st.session_state.resume_pipeline:
                 for target_key in clone_metadata.keys():
                     target_words = set(re.findall(r'[a-zA-Z]{3,}', target_key.lower()))
                     sec_words = set(re.findall(r'[a-zA-Z]{3,}', sec_title.lower()))
-                    if "result" in target_words: target_words.add("analysis")
-                    if "result" in sec_words: sec_words.add("analysis")
+                    if "result" in target_words: 
+                        target_words.add("analysis")
+                    if "result" in sec_words: 
+                        sec_words.add("analysis")
                     if target_words & sec_words: 
                         is_clone_target = True
                         active_clone_key = target_key
@@ -616,39 +582,58 @@ if st.session_state.start_pipeline or st.session_state.resume_pipeline:
             if is_clone_target:
                 skeleton_instruction = clone_metadata[active_clone_key]
                 sys_prompt_m3 = "You are a top-tier academic ghostwriter executing a structural and stylistic clone."
-                integrated_context = f"Core differences: {context_differences}\n{agent._build_memory_context(sec_title, 2500)}\nPrevious context tail: ...{context_tail}..."
+                integrated_context = f"Core differences: {context_differences}\n{memory_block}\nPrevious context tail: ...{context_tail}..."
                 wiz_data_injection = f"USER ACTUAL DATA:\n{user_data_md}" if user_data_md and ("method" in sec_title.lower() or "result" in sec_title.lower() or "analysis" in sec_title.lower()) else "No data required."
-                base_user = get_clone_architect(agent).build_rewrite_instruction(skeleton_instruction, sec_title, topic, integrated_context, wiz_data_injection) + f"\n\n{length_directive}"
+                base_user = clone_arch.build_rewrite_instruction(skeleton_instruction, sec_title, topic, integrated_context, wiz_data_injection) + f"\n\n{length_directive}\n{quarantine_rule}\n{heading_instruction}"
             else:
                 extra_directives = []
+                lower_sec = sec_title.lower()
                 golden_meta = agent.global_memory.get("core_golden_metadata", [])
 
                 if golden_meta:
-                    if any(kw in lower_sec for kw in ["introduction", "background"]): extra_directives.append("【TARGETED DIRECTIVE】 Response to Gaps in [TARGETED INJECTION].")
-                    elif any(kw in lower_sec for kw in ["literature", "theor", "hypothesis", "review"]): extra_directives.append("【TARGETED DIRECTIVE】 Use Theoretical Lens from [TARGETED INJECTION].")
-                    elif any(kw in lower_sec for kw in ["method", "measure", "data", "sample"]): extra_directives.append("【TARGETED DIRECTIVE】 Measurements adapted from [TARGETED INJECTION].")
-                    elif any(kw in lower_sec for kw in ["result", "analysis"]): extra_directives.append("【ZERO HALLUCINATION】 Report <USER_ACTUAL_DATA> exactly.")
-                    elif any(kw in lower_sec for kw in ["discussion", "conclusion", "implication"]): extra_directives.append("【TARGETED DIRECTIVE】 Structured comparison with [TARGETED INJECTION].")
+                    if any(kw in lower_sec for kw in ["introduction", "background"]): 
+                        extra_directives.append("【TARGETED DIRECTIVE】 Locate research gaps from [TARGETED INJECTION].")
+                    elif any(kw in lower_sec for kw in ["literature", "theor", "hypothesis", "review"]): 
+                        extra_directives.append("【TARGETED DIRECTIVE】 Base hypotheses on theories from [TARGETED INJECTION].")
+                    elif any(kw in lower_sec for kw in ["method", "measure", "data", "sample"]): 
+                        extra_directives.append("【TARGETED DIRECTIVE】 Adapt scale from [TARGETED INJECTION].")
+                    elif any(kw in lower_sec for kw in ["result", "analysis"]): 
+                        extra_directives.append("【ZERO HALLUCINATION】 Translate <USER_ACTUAL_DATA> strictly.")
+                    elif any(kw in lower_sec for kw in ["discussion", "conclusion", "implication"]): 
+                        extra_directives.append("【TARGETED DIRECTIVE】 Compare findings with [TARGETED INJECTION].")
+                    else: 
+                        extra_directives.append(APP_CONFIG["deep_refinement_directive"])
                 else:
-                    if sec["is_method_or_result"]: extra_directives.append(APP_CONFIG["methodological_defense_directive"]); extra_directives.append(APP_CONFIG["empirical_data_logic"])
-                    if any(kw in lower_sec for kw in ["theory", "theoretical", "hypothesis", "introduction"]): extra_directives.append(APP_CONFIG["deep_synthesis_directive"]); extra_directives.append(APP_CONFIG["hypothesis_development_directive"])
+                    if sec["is_method_or_result"]: 
+                        extra_directives.append(APP_CONFIG["methodological_defense_directive"])
+                        extra_directives.append(APP_CONFIG["empirical_data_logic"])
+                    if any(kw in lower_sec for kw in ["theory", "theoretical", "hypothesis", "introduction"]): 
+                        extra_directives.append(APP_CONFIG["deep_synthesis_directive"])
+                        extra_directives.append(APP_CONFIG["hypothesis_development_directive"])
+                    if "discussion" in lower_sec: 
+                        extra_directives.append(APP_CONFIG["discussion_depth_directive"])
+                    if "conclusion" in lower_sec: 
+                        extra_directives.append(APP_CONFIG["conclusion_anchor_directive"])
                 
                 theory_injection = "\n\n".join(extra_directives)
                 sys_prompt_m3 = APP_CONFIG["sys_academic_writer"] + "\n" + APP_CONFIG["deep_refinement_directive"]
                 wiz_data_injection = f"\n<USER_ACTUAL_DATA>\n{user_data_md}\n</USER_ACTUAL_DATA>\n" if user_data_md and sec["is_method_or_result"] else ""
-                
-                base_user = f"Generate polished section '{display_name}'.\n{length_directive}\n{theory_injection}\nGlobal Research Synthesis: {synthesis_result}\n{wiz_data_injection}\nPrevious Context Tail: ...{context_tail}...\nIMPORTANT: Output final academic text."
+                base_user = f"Generate section '{display_name}'.\n{length_directive}\n{quarantine_rule}\n{heading_instruction}\n{theory_injection}\n{memory_block}\nSynthesis: {synthesis_result}\n{wiz_data_injection}\nPrevious Tail: ...{context_tail}...\nIMPORTANT: Output final text."
 
-            render_monitor(agent, f"English Drafting ({current_step+1}/{total_steps})", display_name, f"目标: {adjusted_target}")
+            if agent._count_tokens(sys_prompt_m3) + agent._count_tokens(base_user) > 60000: 
+                base_user = agent._truncate_context(base_user, 60000 - agent._count_tokens(sys_prompt_m3))
+            
+            engine_label = "🧬 克隆引擎" if is_clone_target else ("🎯 靶向注入" if golden_meta else "📚 标准引擎")
+            render_monitor(agent, f"English Drafting ({current_step+1}/{total_steps})", display_name, f"目标: {adjusted_target} ({engine_label})")
+            en_ph.info(f"🔄 正在生成高质量章节...")
             
             try:
-                en_draft_raw = en_ph.write_stream(agent.stream_deepseek(sys_prompt_m3, base_user, max_tokens=6000))
-                en_draft_clean = re.sub(r'<think>.*?</think>', '', en_draft_raw, flags=re.DOTALL).strip()
+                en_draft_clean = re.sub(r'<think>.*?</think>', '', en_ph.write_stream(agent.stream_deepseek(sys_prompt_m3, base_user, max_tokens=6000)), flags=re.DOTALL).strip()
                 en_draft_clean = re.sub(rf'(?im)^#\s*{re.escape(sec_title)}\s*\n+', '', en_draft_clean).strip()
-                
-                if sec["is_first"]: en_draft_clean = f"# {sec_title}\n\n{en_draft_clean}"
+                if sec["is_first"]: 
+                    en_draft_clean = f"# {sec_title}\n\n{en_draft_clean}"
                     
-                en_resolved = agent.resolve_citations(en_draft_clean)
+                en_resolved = agent.resolve_citations(en_draft_clean, quarantine=is_quarantined)
                 en_ph.markdown(en_resolved)
                 full_draft_en += en_resolved + "\n\n"
                 
@@ -658,26 +643,29 @@ if st.session_state.start_pipeline or st.session_state.resume_pipeline:
                 
                 actual_words = len(en_draft_clean.split())
                 accumulated_deviation += (actual_words - adjusted_target)
+                st.caption(f"📊 本章节实际字数: {actual_words} (目标: {adjusted_target}), 累计偏差: {accumulated_deviation:+d}")
+                
                 current_step += 1
-                pb.progress(min(15 + int((current_step / total_steps) * 80), 99))
+                pb.progress(min(15 + int((current_step / total_steps) * 80), 99), text=f"🇬🇧 {display_name} 英文完成...")
             except Exception as e:
-                en_ph.error(f"节点异常: {e}")
+                en_ph.error("🚨 节点异常已熔断。")
+                st.error(f"异常: {e} | 💥 系统已保存断点，随时可恢复。")
                 generation_aborted = True
                 break
 
             render_monitor(agent, f"Translating ({current_step+1}/{total_steps})", display_name, "翻译中...")
             trans_ph.empty()
             try:
-                trans_chunk_raw = trans_ph.write_stream(agent.stream_deepseek(f"Translate into {target_lang}. Preserve formatting and [CITE-XX] tags.", f"TEXT:\n{en_draft_clean}", max_tokens=6000, model_name="deepseek-chat"))
-                trans_chunk_clean = re.sub(r'<think>.*?</think>', '', trans_chunk_raw, flags=re.DOTALL).strip()
-                trans_resolved = agent.resolve_citations(trans_chunk_clean)
+                trans_chunk_clean = re.sub(r'<think>.*?</think>', '', trans_ph.write_stream(agent.stream_deepseek(f"Translate into {target_lang}. Preserve formatting and [CITE-XX].", f"TEXT:\n{en_draft_clean}", max_tokens=6000, model_name="deepseek-chat")), flags=re.DOTALL).strip()
+                trans_resolved = agent.resolve_citations(trans_chunk_clean, quarantine=is_quarantined)
                 trans_ph.markdown(trans_resolved)
                 full_draft_translated += trans_resolved + "\n\n"
                 
                 current_step += 1
-                pb.progress(min(15 + int((current_step / total_steps) * 80), 99))
+                pb.progress(min(15 + int((current_step / total_steps) * 80), 99), text=f"{target_flag} {display_name} 翻译完成...")
             except Exception as e:
-                trans_ph.error(f"翻译熔断: {e}")
+                trans_ph.error("🚨 翻译熔断。")
+                st.error(f"异常: {e}")
                 generation_aborted = True
                 break
                 
@@ -687,31 +675,66 @@ if st.session_state.start_pipeline or st.session_state.resume_pipeline:
             state["context_tail"] = context_tail
             state["agent_state"] = agent.export_state()
             
+            os.makedirs(WORKSPACE_DIR, exist_ok=True)
             with open(CHECKPOINT_FILE, "w", encoding="utf-8") as f: 
                 json.dump(state, f, ensure_ascii=False)
 
         if not generation_aborted and full_draft_en.strip():
-            pb.progress(96, text="📦 正在打包 DOCX 文档...")
+            pb.progress(96, text="📦 打包DOCX...")
             full_draft_en = clean_trailing_references(full_draft_en, ["Reference", "References", "Reference List"])
             full_draft_translated = clean_trailing_references(full_draft_translated, ["参考文献", "Reference", "References", "Reference List"])
             
-            apa_refs = "\n\n".join([f"- {agent.citation_map[t]['ref']}" for t in sorted(list(agent.used_citations)) if t in agent.citation_map])
+            total_words_en = len(full_draft_en.split())
+            target_words = int(words)
+            lower_limit = int(target_words * 0.95)
+            upper_limit = int(target_words * 1.05)
+            st.info(f"📊 全文总字数: {total_words_en} (目标区间: {lower_limit} - {upper_limit})")
+            
+            if total_words_en > upper_limit:
+                conclusion_match = re.search(r'(#+\s*Conclusion.*?)(?=\n#+ |\Z)', full_draft_en, re.DOTALL | re.IGNORECASE)
+                if conclusion_match:
+                    compressed_en = agent.call_deepseek("Condense the following conclusion by 15-20% preserving citations.", conclusion_match.group(1), max_tokens=1500)
+                    if compressed_en:
+                        full_draft_en = full_draft_en.replace(conclusion_match.group(1), compressed_en.strip())
+                        compressed_trans = agent._translate_text(compressed_en, target_lang, max_tokens=2000)
+                        if compressed_trans:
+                            trans_match = re.search(rf'(#+\s*{"结论|Заключение|Қорытынды|Conclusion"}.*?)(?=\n#+ |\Z)', full_draft_translated, re.DOTALL | re.IGNORECASE)
+                            if trans_match: 
+                                full_draft_translated = full_draft_translated.replace(trans_match.group(1), compressed_trans.strip())
+                            else: 
+                                full_draft_translated += "\n\n" + compressed_trans.strip()
+            elif total_words_en < lower_limit:
+                expanded_en = agent.call_deepseek("Continue conclusion by adding 150 words (future research).", full_draft_en[-2000:], max_tokens=800)
+                if expanded_en:
+                    full_draft_en += "\n\n" + expanded_en.strip()
+                    expanded_trans = agent._translate_text(expanded_en, target_lang, max_tokens=1000)
+                    if expanded_trans: 
+                        full_draft_translated += "\n\n" + expanded_trans.strip()
+            
+            if len(agent.used_citations) < 50: 
+                for tag in [t for t in agent.citation_map if t not in agent.used_citations][:50 - len(agent.used_citations)]: 
+                    agent.used_citations.add(tag)
+            
+            apa_refs = "\n\n".join([f"- {ref}" for ref in sorted(list(set([agent.citation_map[t]['ref'] for t in sorted(list(agent.used_citations)) if t in agent.citation_map])))])
             full_draft_en += "\n\n# References\n" + apa_refs
             full_draft_translated += "\n\n# 参考文献\n" + apa_refs
 
             bj_time = datetime.now(timezone(timedelta(hours=8))).strftime("%Y%m%d_%H%M")
             zip_buffer = io.BytesIO()
             with zipfile.ZipFile(zip_buffer, "w", zipfile.ZIP_DEFLATED) as zf:
-                zf.writestr(f"Empirical_Research_EN_{bj_time}.docx", text_to_docx(full_draft_en).getvalue())
-                zf.writestr(f"Empirical_Research_ZH_{bj_time}.docx", text_to_docx(full_draft_translated).getvalue())
+                zf.writestr(f"Research_EN_{bj_time}.docx", text_to_docx(full_draft_en).getvalue())
+                zf.writestr(f"Research_Trans_{bj_time}.docx", text_to_docx(full_draft_translated).getvalue())
             zip_buffer.seek(0)
             st.session_state.zip_buffer = zip_buffer.getvalue()
-            st.session_state.zip_filename = f"Article_Draft_{bj_time}.zip"
+            st.session_state.zip_filename = f"Draft_{bj_time}.zip"
             
-            try: os.remove(CHECKPOINT_FILE)
-            except OSError: pass
+            try: 
+                os.remove(CHECKPOINT_FILE)
+            except OSError: 
+                pass
+                
             pb.progress(100, text="✅ 完成")
-            render_monitor(agent, "完成", "✅", "文档已生成并已清空本地快照")
+            render_monitor(agent, "完成", "✅", "文档已生成")
             
     except Exception as e: 
         st.error(f"🚨 启动/恢复异常: {e}")