feat: 多功能改进（导入/项目管理/笔记/错题推荐/UI修复）

backend/agents/note/agent.py

@@ -4,9 +4,12 @@
 支持：
   - 重试机制（指数退避，最多 3 次）
   - 多页分批处理（超过阈值时分批整理再合并）
+  - 多模式结构化输出（function_calling → prompt JSON fallback）
 """
 
+import json
 import logging
+import re
 import time
 
 from langchain_core.messages import SystemMessage, HumanMessage
@@ -22,49 +25,79 @@
 BATCH_CHAR_LIMIT = 6000
 MAX_RETRIES = 3
 
+_messages = lambda ocr_text: [
+    SystemMessage(content=NOTE_ORGANIZE_PROMPT),
+    HumanMessage(
+        content=f"以下是 OCR 识别出的课堂笔记原始文本，请整理为结构化笔记：\n\n{ocr_text}"
+    ),
+]
+
+
+def _extract_json_from_text(text: str) -> dict | None:
+    """从 LLM 输出文本中提取 JSON 对象（兼容 ```json ... ``` 包裹）"""
+    # 尝试 ```json ... ``` 代码块
+    m = re.search(r"```json\s*\n?(.*?)\n?\s*```", text, re.DOTALL)
+    if m:
+        try:
+            return json.loads(m.group(1))
+        except json.JSONDecodeError:
+            pass
+    # 尝试裸 JSON 对象
+    m = re.search(r"\{[\s\S]*\}", text)
+    if m:
+        try:
+            return json.loads(m.group(0))
+        except json.JSONDecodeError:
+            pass
+    return None
+
 
 def _invoke_once(
     model,
     ocr_text: str,
     provider: str = "openai",
     supports_function_calling: bool = True,
 ) -> OrganizedNote:
-    """单次 LLM 调用，返回结构化笔记"""
-    from core.config import settings
-
-    # 百度千帆等特殊平台/模型，即使 supports_function_calling=False
-    # 如果直接用 with_structured_output 仍可能报错 "暂不支持该模型"（如果 Langchain 默认使用了 JSON Schema mode）
-    # 为保证最大兼容性，我们通过 prompt 要求输出 JSON，并手动解析
-    if not supports_function_calling:
-        from langchain_core.output_parsers import PydanticOutputParser
-
-        parser = PydanticOutputParser(pydantic_object=OrganizedNote)
-        format_instructions = parser.get_format_instructions()
-
-        response = model.invoke(
-            [
-                SystemMessage(
-                    content=NOTE_ORGANIZE_PROMPT
-                    + f"\n\n你必须以 JSON 格式输出，且遵循以下结构：\n{format_instructions}"
-                ),
-                HumanMessage(
-                    content=f"以下是 OCR 识别出的课堂笔记原始文本，请整理为结构化笔记：\n\n{ocr_text}"
-                ),
-            ]
-        )
+    """单次 LLM 调用，返回结构化笔记。按优先级尝试多种输出模式。"""
+    messages = _messages(ocr_text)
+
+    # 模式 1: with_structured_output（function_calling / json_schema，取决于模型）
+    if supports_function_calling:
+        try:
+            structured_model = model.with_structured_output(OrganizedNote)
+            return structured_model.invoke(messages)
+        except Exception as exc:
+            logger.warning(
+                "笔记整理: with_structured_output 失败 (%s)，fallback 到 prompt JSON", exc
+            )
+
+    # 模式 2: prompt JSON — 在 system prompt 中要求输出 JSON，并手动解析
+    from langchain_core.output_parsers import PydanticOutputParser
+
+    parser = PydanticOutputParser(pydantic_object=OrganizedNote)
+    format_instructions = parser.get_format_instructions()
+
+    prompt_messages = [
+        SystemMessage(
+            content=NOTE_ORGANIZE_PROMPT
+            + f"\n\n你必须以 JSON 格式输出，且遵循以下结构：\n{format_instructions}"
+        ),
+        messages[1],  # HumanMessage
+    ]
+    response = model.invoke(prompt_messages)
+
+    # 先尝试 PydanticOutputParser（依赖严格格式）
+    try:
         return parser.parse(response.content)
-    else:
-        structured_model = model.with_structured_output(
-            OrganizedNote, method="function_calling"
-        )
-        return structured_model.invoke(
-            [
-                SystemMessage(content=NOTE_ORGANIZE_PROMPT),
-                HumanMessage(
-                    content=f"以下是 OCR 识别出的课堂笔记原始文本，请整理为结构化笔记：\n\n{ocr_text}"
-                ),
-            ]
-        )
+    except Exception:
+        pass
+
+    # 再尝试从文本中提取 JSON 并手动构建
+    data = _extract_json_from_text(response.content)
+    if data:
+        return OrganizedNote.model_validate(data)
+
+    raise RuntimeError("LLM 未返回合法的 JSON 结构")
 
 
 def _invoke_with_retry(

backend/core/config.py

@@ -316,10 +316,6 @@ def build_provider_config(
     ) -> LLMProviderConfig:
         key = self._normalize_provider(name)
 
-        # 处理可能包含多个模型的字符串（取第一个作为默认）
-        if model_name and "," in model_name:
-            model_name = [m.strip() for m in model_name.split(",") if m.strip()][0]
-
         if key == "openai":
             return OpenAICompatibleConfig(
                 api_key=api_key,
@@ -508,7 +504,7 @@ def load_providers_from_db(
         if owns_db:
             db = SessionLocal()
         try:
-            for category in [("openai"), ("anthropic")]:
+            for category in ("openai", "anthropic"):
                 provider = get_active_provider(db, user_id, category)
                 if provider and provider.api_key:
                     cfg = self.build_provider_config(

backend/core/model_selection.py

@@ -22,7 +22,9 @@ class LLMSelectionError(Exception):
 
 
 def split_models(model_name: str | None) -> list[str]:
-    return [item.strip() for item in (model_name or "").split(",") if item.strip()]
+    """返回模型名称列表。现在只支持单个模型，但保留列表形式以兼容调用方。"""
+    name = (model_name or "").strip()
+    return [name] if name else []
 
 
 def build_managed_provider_context(db):

backend/db/crud/projects.py

@@ -4,7 +4,10 @@
 
 from sqlalchemy.orm import Session
 
-from db.models import Note, Project, Question, UploadBatch
+from db.models import (
+    ChatSession, Note, NoteTagMapping, Project, Question,
+    QuestionEmbedding, QuestionTagMapping, UploadBatch,
+)
 
 
 VALID_PROJECT_TYPES = {"question", "note"}
@@ -135,14 +138,18 @@ def delete_project(db: Session, project_id: int, user_id=None) -> bool:
     if project.is_default:
         raise ValueError("DEFAULT_PROJECT_IMMUTABLE")
 
-    # 检查是否有题目或笔记
-    has_questions = db.query(Question.id).filter(Question.project_id == project.id).first()
-    has_notes = db.query(Note.id).filter(Note.project_id == project.id).first()
-
-    if has_questions or has_notes:
-        raise ValueError("PROJECT_NOT_EMPTY")
-
-    # 如果没有题目和笔记了，自动清理关联的空批次（UploadBatch）
+    # 先删除题目和笔记的关联子表，再删除题目/笔记本身
+    question_ids = [q.id for q in db.query(Question.id).filter(Question.project_id == project.id).all()]
+    if question_ids:
+        db.query(QuestionEmbedding).filter(QuestionEmbedding.question_id.in_(question_ids)).delete(synchronize_session=False)
+        db.query(ChatSession).filter(ChatSession.question_id.in_(question_ids)).delete(synchronize_session=False)
+        db.query(QuestionTagMapping).filter(QuestionTagMapping.question_id.in_(question_ids)).delete(synchronize_session=False)
+        db.query(Question).filter(Question.id.in_(question_ids)).delete(synchronize_session=False)
+
+    note_ids = [n.id for n in db.query(Note.id).filter(Note.project_id == project.id).all()]
+    if note_ids:
+        db.query(NoteTagMapping).filter(NoteTagMapping.note_id.in_(note_ids)).delete(synchronize_session=False)
+        db.query(Note).filter(Note.id.in_(note_ids)).delete(synchronize_session=False)
     db.query(UploadBatch).filter(UploadBatch.project_id == project.id).delete()
 
     db.delete(project)

backend/routes/projects.py

@@ -113,8 +113,6 @@ def delete_project(project_id):
             except ValueError as exc:
                 if str(exc) == "DEFAULT_PROJECT_IMMUTABLE":
                     return jsonify({"success": False, "error": "默认项目不能删除"}), 400
-                if str(exc) == "PROJECT_NOT_EMPTY":
-                    return jsonify({"success": False, "error": "项目里还有内容，暂时不能删除"}), 400
                 raise
             if not deleted:
                 return jsonify({"success": False, "error": "项目不存在"}), 404

backend/routes/questions.py

@@ -504,7 +504,10 @@ def save_to_db():
 
     Request Body:
         {
-            "selected_ids": ["q_0", "q_1", ...]   # 选中的题目 ID 列表
+            "selected_ids": ["q_0", "q_1", ...],  # 选中的题目 ID 列表
+            "run_id": "xxx",                       # 可选，从 WorkflowRun 读取题目
+            "record_id": 123,                      # 可选，从 SplitRecord 读取题目（历史记录导入）
+            "project_id": 456                      # 目标错题库 ID
         }
     """
     try:
@@ -516,14 +519,39 @@ def save_to_db():
             return jsonify({'success': False, 'error': '请选择至少一道题目'}), 400
 
         run_id = data.get('run_id')
-        if not run_id:
-            return jsonify({'success': False, 'code': 'MISSING_RUN_ID', 'error': '缺少 run_id，请重新分割题目'}), 400
+        record_id = data.get('record_id')
         user_id = session.get('user_id')
-        with SessionLocal() as db:
-            run = run_store.get_split_run(db, run_id, user_id=user_id)
-            if not run or run.status != run_store.STATUS_SUCCEEDED:
-                return jsonify({'success': False, 'error': '请先分割题目'}), 400
-            questions = run_store.read_questions(run)
+
+        # 优先从 SplitRecord 读取（历史记录导入场景）
+        if record_id:
+            with SessionLocal() as db:
+                record = crud.get_split_record_by_id(db, record_id, user_id=user_id)
+                if not record:
+                    return jsonify({'success': False, 'error': '分割记录不存在'}), 404
+                questions = json.loads(record.questions_json) if record.questions_json else []
+                subject = record.subject or ''
+                file_names = json.loads(record.file_names_json) if record.file_names_json else []
+                batch_info = {
+                    "original_filename": ", ".join(file_names) or "Unknown",
+                    "subject": subject,
+                    "file_path": "",
+                }
+        # 从 WorkflowRun 读取（正常分割后导入场景）
+        elif run_id:
+            with SessionLocal() as db:
+                run = run_store.get_split_run(db, run_id, user_id=user_id)
+                if not run or run.status != run_store.STATUS_SUCCEEDED:
+                    return jsonify({'success': False, 'error': '请先分割题目'}), 400
+                questions = run_store.read_questions(run)
+                subject = run_store.read_subject(run)
+                file_names = json.loads(run.file_names_json) if run.file_names_json else []
+                batch_info = {
+                    "original_filename": ", ".join(file_names) or "Unknown",
+                    "subject": subject,
+                    "file_path": run.result_dir,
+                }
+        else:
+            return jsonify({'success': False, 'code': 'MISSING_SOURCE', 'error': '缺少 run_id 或 record_id'}), 400
 
         uid_set = set(selected_uids)
         selected_questions = [q for q in questions if q.get('uid') in uid_set]
@@ -540,16 +568,6 @@ def save_to_db():
                 if 'user_answer' in answers_map[uid]:
                     sq['user_answer'] = answers_map[uid]['user_answer']
 
-        # 读取科目信息
-        subject = run_store.read_subject(run)
-
-        file_names = json.loads(run.file_names_json) if run.file_names_json else []
-        batch_info = {
-            "original_filename": ", ".join(file_names) or "Unknown",
-            "subject": subject,
-            "file_path": run.result_dir,
-        }
-
         with SessionLocal() as db:
             try:
                 project_id = (
@@ -572,7 +590,7 @@ def save_to_db():
         return jsonify({
             'success': True,
             'message': f'已导入 {result["created"]} 道题目（跳过 {result["duplicates"]} 道重复）',
-            'run_id': run.public_id,
+            'run_id': run_id,
             'created': result['created'],
             'duplicates': result['duplicates'],
         })

backend/routes/settings.py

@@ -109,11 +109,7 @@ def get_status():
                     crud.get_active_provider(db, user_id, category) if user_id else None
                 )
                 if provider and provider.api_key:
-                    models = (
-                        [m.strip() for m in provider.model_name.split(",")]
-                        if provider.model_name
-                        else []
-                    )
+                    models = [provider.model_name] if provider.model_name else []
                     available_models.append(
                         {
                             "value": category,
@@ -128,7 +124,7 @@ def get_status():
                 else:
                     managed_cfg = managed_llm.get(category)
                     managed_models = (
-                        [m.strip() for m in managed_cfg.model_name.split(",")]
+                        [managed_cfg.model_name]
                         if managed_cfg
                         and managed_cfg.configured
                         and managed_cfg.model_name
@@ -302,7 +298,7 @@ def list_models():
                     else:
                         provider = crud.get_active_provider(db, user_id, provider_type)
                     if provider:
-                        api_key = api_key or provider.api_key or ""
+                        api_key = provider.api_key or ""
                         base_url = base_url or provider.base_url or ""
                 if not api_key and provider_type in ("openai", "anthropic"):
                     system_provider = crud.get_active_system_provider(db, provider_type)
@@ -413,7 +409,7 @@ def test_paddleocr():
                     else:
                         provider = crud.get_active_provider(db, user_id, "paddleocr")
                     if provider:
-                        api_token = api_token or provider.api_key or ""
+                        api_token = provider.api_key or ""
                         api_url = api_url or provider.base_url or ""
                 if not api_token or not api_url:
                     system_provider = crud.get_active_system_provider(db, "paddleocr")

frontend/src/api/upload.js

@@ -83,9 +83,10 @@ export async function exportQuestions(selectedIds, runId) {
 }
 
 /** 将选中的分割题目和答案保存到错题库。 */
-export async function saveToDb(selectedIds, answers = [], runId, projectId) {
+export async function saveToDb(selectedIds, answers = [], runId, projectId, recordId) {
   const body = { selected_ids: selectedIds, answers }
   if (runId) body.run_id = runId
+  if (recordId) body.record_id = recordId
   if (projectId) body.project_id = projectId
   const resp = await fetch('/api/save-to-db', {
     method: 'POST',

frontend/src/components/base/BaseModal.vue

@@ -11,6 +11,7 @@ const props = defineProps({
   bodyClass: { type: String, default: 'px-6 py-5' },
   blurBackdrop: { type: Boolean, default: true },
   sidebarOffset: { type: Number, default: null },
+  zIndex: { type: Number, default: null },
 })
 
 const emit = defineEmits(['close'])
@@ -28,7 +29,7 @@ const backdropStyle = computed(() => ({
       <div
         v-if="open"
         class="dialog-backdrop fixed inset-0 z-[100] bg-black/40 transition-all duration-300"
-        :style="backdropStyle"
+        :style="{ ...backdropStyle, ...(zIndex ? { zIndex: zIndex - 1 } : {}) }"
         @click="emit('close')"
       ></div>
     </Transition>
@@ -37,6 +38,7 @@ const backdropStyle = computed(() => ({
       <div
         v-if="open"
         class="fixed inset-0 z-[101] flex items-center justify-center p-4 transition-all duration-300"
+        :style="zIndex ? { zIndex } : undefined"
         @click.self="emit('close')"
       >
         <div