refactor!: 删除 Web/CLI 应用层与内置 LLM 抽取，出库单功能下线（2b）

入口只剩 MCP 服务器；编排由宿主 Agent 负责，本项目只留工具与确定性内核。

删除：
- src/web/（Flask 路由、SSE 轮询、自研聊天前端、模板与静态资源，约 3400 行）
- src/main.py、src/pipeline.py、src/pipeline_core.py（CLI 全流程入口与两套重复管道）
- core/extraction/{extractor,llm_extractor}.py（内置 LLM 抽取、llm_stream.log、file_events.log、并行抽取）
- infra/documents/consumable.py 与 infra/documents/pdf.py：出库单功能下线（Word 模板 4fbb4c3 起已不在仓库，
  该路径此前一直静默失败），PDF 渲染由 core/extraction/render.py 取代
- mcp__arb__doc_consumable 工具；exceptions.ExtractionError；config 的 consumable_storage / get_llm_config /
  load_session_config / SESSION_CONFIG_KEYS（Web 专属，已无引用）
- 针对被删代码的 tests/test_extractor.py、tests/test_llm_extractor.py 与 scripts/ 下的 LLM 调试脚本
- 依赖：llama-index、llama-index-llms-openai-like、python-dotenv（连同其 50+ 传递依赖）
- 提交进来的调试截图 src/images/、根目录 images/ 与 .coverage（并补 .gitignore）

保留与新增：
- 材料入口改为独立进程 src/upload（手机扫码/PC 上传 → cases/<id>/inbox/）
- runner.py 自带 save_invoice_groups，不再依赖 pipeline_core
- scripts/mcp_smoke.py：stdio 自检（拉起自身、列工具、调 case_list、验证 instructions 下发）
- make mcp / mcp-smoke / upload 取代 make run；check 的 mypy 目标从 src/main.py 扩到 src

验证：mcp_smoke 通过（17 个工具、instructions 1449 字节、case_list 正常）；申购单实机生成此前已验证。
门禁：ruff check / format --check(43) / mypy src(36 文件) / deptry / pytest 107 passed。
注：仍用 --no-verify 提交并手动跑等价检查（原因见 2f08bbd 提交说明）。

scripts/debug_stream_fields.py

@@ -1,69 +0,0 @@
-"""诊断脚本：检查 stream_chat 返回对象的字段结构"""
-
-import sys
-from pathlib import Path
-
-sys.path.insert(0, str(Path(__file__).parent))
-
-# 加载 .env
-from dotenv import load_dotenv  # noqa: E402
-from llama_index.core.llms import ChatMessage  # noqa: E402
-
-from src.config import get_llm_config  # noqa: E402
-from src.core.extraction import _create_llm  # noqa: E402
-
-load_dotenv(Path(__file__).parent / ".env")
-
-llm_config = get_llm_config()
-print(f"LLM config: model={llm_config['model']}, api_base={llm_config['api_base']}")
-
-llm = _create_llm()
-messages = [
-    ChatMessage(role="system", content="你是一个助手"),
-    ChatMessage(role="user", content="1+1 等于几？"),
-]
-
-print("\n=== 检查 stream_chat 返回对象的字段 ===")
-count = 0
-try:
-    for resp in llm.stream_chat(messages, temperature=0.1):
-        count += 1
-        if count <= 3:
-            print(f"\n--- chunk #{count} ---")
-            print(f"  type: {type(resp).__name__}")
-            print(f"  delta: {repr(resp.delta)[:200]}")
-            if hasattr(resp, "additional_kwargs") and resp.additional_kwargs:
-                print(f"  additional_kwargs keys: {list(resp.additional_kwargs.keys())}")
-                for k, v in resp.additional_kwargs.items():
-                    val_preview = str(v)[:200]
-                    print(f"    additional_kwargs['{k}']: {val_preview}")
-            if hasattr(resp, "raw"):
-                raw = resp.raw
-                if isinstance(raw, dict):
-                    print(f"  raw keys: {list(raw.keys())}")
-                    for k, v in raw.items():
-                        val_preview = str(v)[:200]
-                        print(f"    raw['{k}']: {val_preview}")
-                else:
-                    print(f"  raw type: {type(raw)}")
-            if hasattr(resp, "message") and resp.message:
-                msg = resp.message
-                print(f"  message type: {type(msg).__name__}")
-                if hasattr(msg, "additional_kwargs") and msg.additional_kwargs:
-                    print(f"  message.additional_kwargs keys: {list(msg.additional_kwargs.keys())}")
-                    for k, v in msg.additional_kwargs.items():
-                        val_preview = str(v)[:200]
-                        print(f"    message.additional_kwargs['{k}']: {val_preview}")
-                if hasattr(msg, "reasoning_content"):
-                    rc = msg.reasoning_content
-                    print(f"  message.reasoning_content: {repr(rc)[:200]}")
-        elif count == 4:
-            print("\n... (更多 chunk 省略)")
-        if count >= 10:
-            break
-    print(f"\n=== 共收到 {count} 个 chunk ===")
-except Exception as e:
-    print(f"\n错误: {e}")
-    import traceback
-
-    traceback.print_exc()

scripts/mcp_smoke.py

@@ -0,0 +1,47 @@
+"""MCP 服务器自检：以 stdio 拉起自身，列出工具并调一次 case_list。
+
+用法：uv run python scripts/mcp_smoke.py
+验证要点：工具数量与名称、structured_content 可解析、instructions 已下发。
+"""
+
+from __future__ import annotations
+
+import asyncio
+import sys
+from pathlib import Path
+
+sys.path.insert(0, str(Path(__file__).resolve().parents[1]))
+
+from mcp.client.stdio import StdioServerParameters  # noqa: E402
+
+from mcp import Client  # noqa: E402
+
+
+async def main() -> int:
+    root = str(Path(__file__).resolve().parents[1])
+    params = StdioServerParameters(command=sys.executable, args=["-m", "src.mcp.server"], cwd=root)
+    async with Client(params) as client:
+        listed = await client.list_tools()
+        tools = list(getattr(listed, "tools", listed))
+        names = sorted(t.name for t in tools)
+        print("instructions bytes:", len(getattr(client, "instructions", "") or ""))
+        print("tools:", len(names))
+        for name in names:
+            print("  -", name)
+        result = await client.call_tool("case_list", {})
+        payload = getattr(result, "structured_content", None)
+        if not isinstance(payload, dict) or not payload.get("ok"):
+            print("case_list 调用异常:", payload)
+            return 1
+        print("case_list ok, cases:", len(payload.get("cases", [])))
+        missing = [
+            n
+            for n in ("case_status", "match_payments", "info_skeleton", "doc_requisition", "case_patch_extraction")
+            if n not in names
+        ]
+        print("关键工具缺失:", missing or "无")
+        return 1 if missing else 0
+
+
+if __name__ == "__main__":
+    sys.exit(asyncio.run(main()))

scripts/test_application_extract.py

@@ -1,208 +0,0 @@
-#!/usr/bin/env python3
-"""单独测试事前申请单的信息提取功能
-
-用于调试 LLM 对事前申请单的提取准确率。
-
-用法:
-    # 测试项目根目录的事前申请单.pdf
-    python scripts/test_application_extract.py
-
-    # 测试指定文件
-    python scripts/test_application_extract.py --file path/to/file.pdf
-
-    # 测试 scripts/data 目录下的事前申请单
-    python scripts/test_application_extract.py --dir scripts/data
-"""
-
-import argparse
-import io
-import json
-import sys
-from pathlib import Path
-
-# 加载 .env 环境变量（在导入 src 模块之前）
-from dotenv import load_dotenv
-
-ROOT = Path(__file__).resolve().parent.parent
-load_dotenv(ROOT / ".env")
-
-# Windows 终端强制 UTF-8
-sys.stdout = io.TextIOWrapper(sys.stdout.buffer, encoding="utf-8", errors="replace")
-
-sys.path.insert(0, str(ROOT))  # noqa: E402
-
-from src.core.extraction import extract_document  # noqa: E402
-
-
-def test_single_file(file_path: Path) -> None:
-    """测试单个文件的提取效果"""
-    print("=" * 60)
-    print(f"测试文件: {file_path.name}")
-    print("=" * 60)
-
-    if not file_path.exists():
-        print(f"文件不存在: {file_path}")
-        return
-
-    try:
-        # 调用统一提取接口
-        result = extract_document(file_path)
-
-        if not result:
-            print("[ERROR] 提取返回空结果")
-            return
-
-        # 检查类型判断
-        inv_type = result.get("invoice_type", "")
-        print(f"\n[类型判断] invoice_type = {inv_type}")
-
-        if inv_type != "application":
-            print(f"[WARNING] 类型判断错误！期望 'application'，实际得到 '{inv_type}'")
-        else:
-            print("[OK] 类型判断正确")
-
-        # 展示提取结果
-        print("\n[提取结果]")
-        print(json.dumps(result, ensure_ascii=False, indent=2))
-
-        # 字段完整性检查
-        print("\n[字段检查]")
-        expected_fields = {
-            "invoice_type": "类型标识",
-            "project_name": "项目名称",
-            "purpose": "出差事由",
-            "start_date": "开始日期",
-            "end_date": "结束日期",
-            "person_info": "人员信息",
-        }
-
-        for field, desc in expected_fields.items():
-            value = result.get(field)
-            if value is None:
-                print(f"  [MISSING] {desc} ({field}) - 字段缺失")
-            elif value == "" or value == []:
-                print(f"  [EMPTY]   {desc} ({field}) - 字段为空")
-            else:
-                print(f"  [OK]      {desc} ({field})")
-
-        # 日期格式检查
-        for date_field in ["start_date", "end_date"]:
-            date_value = result.get(date_field, "")
-            if date_value and len(date_value) == 10:
-                try:
-                    parts = date_value.split("-")
-                    if len(parts) == 3:
-                        int(parts[0])  # year
-                        int(parts[1])  # month
-                        int(parts[2])  # day
-                        print(f"  [OK]      {date_field} 格式正确 (YYYY-MM-DD)")
-                except (ValueError, IndexError):
-                    print(f"  [ERROR]   {date_field} 格式错误: {date_value}")
-            elif date_value:
-                print(f"  [ERROR]   {date_field} 格式错误: {date_value}")
-
-        # 人员信息结构检查
-        person_info = result.get("person_info")
-        if person_info:
-            if isinstance(person_info, list):
-                print(f"\n[人员信息] 共 {len(person_info)} 人")
-                for i, person in enumerate(person_info, 1):
-                    pid = person.get("person_id", "")
-                    pname = person.get("person_name", "")
-                    print(f"  人员 {i}: {pname} ({pid})")
-            elif isinstance(person_info, dict):
-                print("\n[人员信息] 单人格式")
-                print(f"  姓名: {person_info.get('person_name', '')}")
-                print(f"  编号: {person_info.get('person_id', '')}")
-
-    except Exception as e:
-        print(f"\n[EXCEPTION] 提取失败: {e}")
-        import traceback
-
-        traceback.print_exc()
-
-
-def test_cache_comparison(file_path: Path) -> None:
-    """对比原始提取和缓存结果"""
-    cache_dir = file_path.parent / ".invoice_cache"
-    cache_file = cache_dir / f"{file_path.stem}{file_path.suffix}.json"
-
-    if not cache_file.exists():
-        print(f"\n[INFO] 无缓存文件对比: {cache_file}")
-        return
-
-    print("\n" + "=" * 60)
-    print("[缓存对比]")
-    print("=" * 60)
-
-    try:
-        with open(cache_file, encoding="utf-8") as f:
-            cache_data = json.load(f)
-
-        cached_result = cache_data.get("extracted_data", {})
-        print("\n[缓存数据]")
-        print(json.dumps(cached_result, ensure_ascii=False, indent=2))
-
-        # 对比关键字段
-        print("\n[字段对比]")
-        for key in ["invoice_type", "project_name", "purpose", "start_date", "end_date"]:
-            cached_value = cached_result.get(key, "<缺失>")
-            print(f"  {key}: {cached_value}")
-
-    except Exception as e:
-        print(f"[ERROR] 读取缓存失败: {e}")
-
-
-def main() -> None:
-    parser = argparse.ArgumentParser(description="测试事前申请单信息提取")
-    parser.add_argument(
-        "--file",
-        type=str,
-        default=None,
-        help="要测试的文件路径 (默认: 扫描 scripts/data 目录)",
-    )
-    parser.add_argument(
-        "--dir",
-        type=str,
-        default=str(ROOT / "scripts" / "data"),
-        help="扫描目录下所有事前申请单文件 (默认: scripts/data)",
-    )
-    parser.add_argument(
-        "--no-cache-compare",
-        action="store_true",
-        help="不执行缓存对比",
-    )
-
-    args = parser.parse_args()
-
-    if args.dir:
-        # 目录模式：扫描所有PDF文件
-        dir_path = Path(args.dir)
-        if not dir_path.exists():
-            print(f"目录不存在: {dir_path}")
-            sys.exit(1)
-
-        pdf_files = sorted(dir_path.glob("*.pdf"))
-        if not pdf_files:
-            print(f"目录下没有找到PDF文件: {dir_path}")
-            sys.exit(1)
-
-        print(f"发现 {len(pdf_files)} 个PDF文件，开始逐个测试...\n")
-        for pdf in pdf_files:
-            if "事前申请" in pdf.name or "申请单" in pdf.name:
-                test_single_file(pdf)
-                if not args.no_cache_compare:
-                    test_cache_comparison(pdf)
-                print()
-    else:
-        # 单文件模式
-        file_path = Path(args.file)
-        test_single_file(file_path)
-        if not args.no_cache_compare:
-            test_cache_comparison(file_path)
-
-    print("\n测试完成!")
-
-
-if __name__ == "__main__":
-    main()

scripts/test_multimodal.py

@@ -1,81 +0,0 @@
-#!/usr/bin/env python3
-"""测试 PDF 多模态提取完整链路
-
-用法:
-    python scripts/test_multimodal.py
-"""
-
-import io
-import sys
-from pathlib import Path
-
-# Windows 终端强制 UTF-8
-sys.stdout = io.TextIOWrapper(sys.stdout.buffer, encoding="utf-8", errors="replace")
-
-ROOT = Path(__file__).resolve().parent.parent
-sys.path.insert(0, str(ROOT))  # noqa: E402
-
-from src.core.extraction import extract_document  # noqa: E402
-from src.infra.documents.pdf import render_pdf_to_images  # noqa: E402
-
-
-def test_render() -> None:
-    """测试 PDF 渲染"""
-    pdf_path = ROOT / "事前申请单.pdf"
-    if not pdf_path.exists():
-        print(f"跳过: {pdf_path.name} 不存在")
-        return
-
-    print("=" * 60)
-    print("测试 PDF 渲染")
-    print("=" * 60)
-    try:
-        images = render_pdf_to_images(pdf_path, dpi=150)
-        if images:
-            print(f"成功渲染 {len(images)} 页")
-            for i, img_b64 in enumerate(images):
-                print(f"  第 {i + 1} 页: base64 长度 {len(img_b64)} 字符")
-        else:
-            print("渲染返回空列表！")
-    except ImportError as e:
-        print(f"导入失败: {e}")
-    except Exception as e:
-        print(f"渲染异常: {e}")
-
-
-def test_multimodal_extract() -> None:
-    """测试完整的多模态提取链路"""
-    pdf_path = ROOT / "事前申请单.pdf"
-    if not pdf_path.exists():
-        print(f"跳过: {pdf_path.name} 不存在")
-        return
-
-    print()
-    print("=" * 60)
-    print("测试多模态提取")
-    print("=" * 60)
-    try:
-        result = extract_document(pdf_path)
-        if result:
-            print("提取成功:")
-            import json
-
-            print(json.dumps(result, ensure_ascii=False, indent=2))
-        else:
-            print("提取返回空字典！")
-    except Exception as e:
-        print(f"提取异常: {e}")
-        import traceback
-
-        traceback.print_exc()
-
-
-def main() -> None:
-    test_render()
-    test_multimodal_extract()
-    print()
-    print("测试完成!")
-
-
-if __name__ == "__main__":
-    main()

scripts/test_travel_info.py

@@ -1,54 +0,0 @@
-#!/usr/bin/env python3
-"""直接测试 extract_travel_info 函数
-
-所有数据从 .invoice_cache 缓存中自动加载，无需手动构造样本数据。
-
-用法:
-    python scripts/test_travel_info.py
-"""
-
-import json
-import sys
-from pathlib import Path
-
-# 项目根目录
-ROOT = Path(__file__).resolve().parent.parent
-sys.path.insert(0, str(ROOT))  # noqa: E402
-
-from src.core.extraction import extract_travel_info  # noqa: E402
-
-
-def main() -> None:
-    source_dir = ROOT / "scripts" / "data"
-
-    if not source_dir.exists():
-        print(f"源文件目录不存在: {source_dir}")
-        sys.exit(1)
-
-    print("=" * 60)
-    print("测试 extract_travel_info（数据来自缓存）")
-    print("=" * 60)
-    print(f"源文件目录: {source_dir}")
-    print()
-
-    try:
-        result = extract_travel_info(
-            source_dir=source_dir,
-        )
-        print()
-        print("=" * 60)
-        print("提取结果:")
-        print("=" * 60)
-        print(json.dumps(result, ensure_ascii=False, indent=2))
-        print()
-        print("测试通过!")
-    except Exception as e:
-        print(f"测试失败: {e}")
-        import traceback
-
-        traceback.print_exc()
-        sys.exit(1)
-
-
-if __name__ == "__main__":
-    main()