feat: 接入 BGE-M3 本地嵌入默认配置 + 新增 Qwen3-Reranker 重排

- .env.example 嵌入默认改为本地 BGE-M3（EMBEDDING_PROVIDER=local, dim=1024），NAS 全新部署即用
- 新增 app/services/reranker.py：基于 Ollama /api/rerank 的 cross-encoder 重排服务
- retriever 在 chunk 候选阶段接入语义精排，调用失败优雅降级为 RRF 顺序
- docker-compose 拉取 qwen3-reranker:0.6b，OLLAMA_MAX_LOADED_MODELS 提至 3
- .gitignore 排除 .workbuddy/ 与 .trae/（防止误提交项目记忆与 IDE spec）
- 新增 reranker 单元与集成测试，全量测试 529 passed

tests/test_reranker.py

@@ -0,0 +1,128 @@
+"""RerankerService 单元测试（mock httpx，不发起真实网络请求）"""
+
+from typing import Any
+
+import httpx
+import pytest
+
+from app.config import settings
+from app.services.reranker import OllamaRerankerService, create_reranker_service
+
+
+def _make_response(status_code: int, data: dict | None = None) -> httpx.Response:
+    """构造带 request 上下文的 httpx.Response（raise_for_status 依赖 request）"""
+    request = httpx.Request("POST", "http://localhost:11434/api/rerank")
+    if data is None:
+        return httpx.Response(status_code, request=request)
+    return httpx.Response(status_code, json=data, request=request)
+
+
+class _FakeAsyncClient:
+    """httpx.AsyncClient 替代品：记录请求参数，返回预设响应或抛出预设异常"""
+
+    def __init__(
+        self,
+        captured: dict,
+        response: httpx.Response | None = None,
+        error: Exception | None = None,
+        **kwargs: Any,
+    ) -> None:
+        self._captured = captured
+        self._response = response
+        self._error = error
+        captured["timeout"] = kwargs.get("timeout")
+
+    async def __aenter__(self) -> "_FakeAsyncClient":
+        return self
+
+    async def __aexit__(self, *args: object) -> bool:
+        return False
+
+    async def post(self, url: str, json: dict | None = None) -> httpx.Response:
+        self._captured["url"] = url
+        self._captured["json"] = json
+        if self._error is not None:
+            raise self._error
+        assert self._response is not None
+        return self._response
+
+
+def _client(monkeypatch: pytest.MonkeyPatch, captured: dict, **kw: Any) -> None:
+    monkeypatch.setattr(httpx, "AsyncClient", lambda **k: _FakeAsyncClient(captured, **k, **kw))
+
+
+class TestRerank:
+    async def test_returns_scores_aligned_to_documents(self, monkeypatch: pytest.MonkeyPatch) -> None:
+        captured: dict = {}
+        _client(
+            monkeypatch,
+            captured,
+            response=_make_response(
+                200,
+                {
+                    "results": [
+                        {"index": 0, "relevance_score": 0.2},
+                        {"index": 1, "relevance_score": 0.9},
+                        {"index": 2, "relevance_score": 0.5},
+                    ]
+                },
+            ),
+        )
+        service = OllamaRerankerService(base_url="http://localhost:11434/", model="qwen3-reranker:0.6b")
+
+        scores = await service.rerank("查询", ["doc-a", "doc-b", "doc-c"])
+
+        assert scores == [0.2, 0.9, 0.5]
+        assert captured["url"] == "http://localhost:11434/api/rerank"
+        assert captured["json"] == {
+            "model": "qwen3-reranker:0.6b",
+            "query": "查询",
+            "documents": ["doc-a", "doc-b", "doc-c"],
+        }
+
+    async def test_empty_documents_returns_empty(self, monkeypatch: pytest.MonkeyPatch) -> None:
+        captured: dict = {}
+        _client(monkeypatch, captured, response=_make_response(200, {"results": []}))
+        service = OllamaRerankerService(base_url="http://localhost:11434", model="qwen3-reranker:0.6b")
+
+        scores = await service.rerank("查询", [])
+
+        assert scores == []
+        # 空文档不发起请求
+        assert "url" not in captured
+
+    async def test_partial_results_zero_fill_missing(self, monkeypatch: pytest.MonkeyPatch) -> None:
+        captured: dict = {}
+        # 仅返回 index 1，index 0/2 缺失 → 补 0
+        _client(
+            monkeypatch,
+            captured,
+            response=_make_response(200, {"results": [{"index": 1, "relevance_score": 0.7}]}),
+        )
+        service = OllamaRerankerService(base_url="http://localhost:11434", model="qwen3-reranker:0.6b")
+
+        scores = await service.rerank("查询", ["a", "b", "c"])
+
+        assert scores == [0.0, 0.7, 0.0]
+
+    async def test_http_error_raises(self, monkeypatch: pytest.MonkeyPatch) -> None:
+        _client(monkeypatch, {}, response=_make_response(500))
+        service = OllamaRerankerService(base_url="http://localhost:11434", model="qwen3-reranker:0.6b")
+
+        with pytest.raises(httpx.HTTPStatusError):
+            await service.rerank("查询", ["a", "b"])
+
+
+class TestCreateRerankerService:
+    def test_disabled_returns_none(self, monkeypatch: pytest.MonkeyPatch) -> None:
+        monkeypatch.setattr(settings, "reranker_enabled", False)
+        assert create_reranker_service() is None
+
+    def test_enabled_returns_service(self, monkeypatch: pytest.MonkeyPatch) -> None:
+        monkeypatch.setattr(settings, "reranker_enabled", True)
+        monkeypatch.setattr(settings, "reranker_model", "qwen3-reranker:0.6b")
+        monkeypatch.setattr(settings, "ollama_base_url", "http://ollama:11434")
+        monkeypatch.setattr(settings, "reranker_timeout", 30.0)
+        service = create_reranker_service()
+        assert isinstance(service, OllamaRerankerService)
+        assert service.model == "qwen3-reranker:0.6b"