feat(app-instance): 集成Beaver后端并更新配置管理

集成新的Beaver后端服务到应用实例中，替换原有的nanobot实现。

主要变更包括：
- 在Dockerfile和环境配置中添加Beaver相关路径和配置变量
- 更新工作目录结构从.nanobot到.beaver
- 实现Beaver引擎加载器，支持配置文件加载和工具组装
- 添加内置工具如ListDirectoryTool、ReadFileTool、SearchFilesTool
- 更新消息处理流程，支持通道适配器和网关模式
- 重构技能系统，支持显式工具提示和嵌入式检索
- 改进错误处理和生命周期管理

此变更使应用实例能够使用统一的Beaver后端进行AI代理运行时管理。

app-instance/backend/tests/unit/test_tool_assembler.py

@@ -0,0 +1,149 @@
+from __future__ import annotations
+
+import asyncio
+from pathlib import Path
+import subprocess
+import sys
+from types import SimpleNamespace
+
+from beaver.engine.context import SkillContext
+from beaver.foundation.embedding import EmbeddingRetriever
+from beaver.skills.catalog.loader import SkillsLoader
+from beaver.tools import BaseTool, ToolAssembler, ToolContext, ToolExecutor, ToolRegistry, ToolResult, ToolSpec
+
+
+class DummyTool(BaseTool):
+    def __init__(
+        self,
+        name: str,
+        *,
+        description: str | None = None,
+        toolset: str = "test",
+        always_available: bool = False,
+    ) -> None:
+        self._spec = ToolSpec(
+            name=name,
+            description=description or name,
+            input_schema={"type": "object", "properties": {}},
+            toolset=toolset,
+            always_available=always_available,
+        )
+
+    @property
+    def spec(self) -> ToolSpec:
+        return self._spec
+
+    async def invoke(self, arguments: dict, context: ToolContext) -> ToolResult:
+        return ToolResult(success=True, content="ok", tool_name=self.spec.name)
+
+
+class StaticRetriever:
+    async def retrieve(self, **kwargs):
+        candidates = kwargs["candidates"]
+        top_k = kwargs["top_k"]
+        preferred = ["search_files", "echo"]
+        ordered = sorted(
+            candidates,
+            key=lambda item: preferred.index(item["name"]) if item["name"] in preferred else len(preferred),
+        )
+        return ordered[:top_k]
+
+
+def test_tool_spec_exports_mcp_and_provider_schema() -> None:
+    spec = ToolSpec(
+        name="read_file",
+        description="Read a file",
+        input_schema={"type": "object", "properties": {"path": {"type": "string"}}},
+        toolset="file",
+    )
+
+    assert spec.to_mcp_descriptor() == {
+        "name": "read_file",
+        "description": "Read a file",
+        "inputSchema": {"type": "object", "properties": {"path": {"type": "string"}}},
+    }
+    assert spec.to_provider_schema()["function"]["parameters"] == spec.input_schema
+
+
+def test_tool_assembler_merges_always_skill_hints_and_embedding(tmp_path: Path) -> None:
+    skill_dir = tmp_path / "skills" / "docker-debug"
+    skill_dir.mkdir(parents=True)
+    (skill_dir / "SKILL.md").write_text(
+        """---
+name: docker-debug
+description: Debug Docker issues.
+tools:
+  - terminal
+---
+
+# Docker Debug
+""",
+        encoding="utf-8",
+    )
+
+    registry = ToolRegistry()
+    registry.register(DummyTool("memory", toolset="memory", always_available=True))
+    registry.register(DummyTool("skill_view", toolset="skills", always_available=True))
+    registry.register(DummyTool("terminal", toolset="shell"))
+    registry.register(DummyTool("search_files", toolset="file"))
+    registry.register(DummyTool("echo", toolset="debug"))
+
+    assembler = ToolAssembler(retriever=StaticRetriever())
+    loader = SkillsLoader(tmp_path)
+    selected = asyncio.run(
+        assembler.assemble(
+            task_description="排查 Docker 容器日志",
+            registry=registry,
+            skills_loader=loader,
+            activated_skills=[SkillContext(name="docker-debug", content="")],
+            top_k=1,
+        )
+    )
+
+    assert [spec.name for spec in selected] == ["memory", "skill_view", "terminal", "search_files"]
+
+
+def test_embedding_fallback_can_return_all_or_top_k() -> None:
+    candidates = [{"name": f"tool_{index}", "description": "", "input_schema": "{}"} for index in range(3)]
+    retriever = EmbeddingRetriever(api_key_env="MISSING_EMBEDDING_KEY", api_base_env="MISSING_EMBEDDING_BASE")
+
+    all_candidates = asyncio.run(
+        retriever.retrieve(query="x", candidates=candidates, top_k=1, fallback_top_k=None)
+    )
+    top_candidate = asyncio.run(
+        retriever.retrieve(query="x", candidates=candidates, top_k=1, fallback_top_k=1)
+    )
+
+    assert [item["name"] for item in all_candidates] == ["tool_0", "tool_1", "tool_2"]
+    assert [item["name"] for item in top_candidate] == ["tool_0"]
+
+
+def test_beaver_tools_import_does_not_load_provider_stack_with_socks_proxy() -> None:
+    code = (
+        "import beaver.tools\n"
+        "from beaver.skills.catalog.loader import SkillsLoader\n"
+        "print('ok')"
+    )
+    result = subprocess.run(
+        [sys.executable, "-c", code],
+        check=False,
+        capture_output=True,
+        text=True,
+        env={
+            "PYTHONPATH": str(Path(__file__).resolve().parents[2]),
+            "HTTP_PROXY": "socks://127.0.0.1:7897/",
+            "HTTPS_PROXY": "socks://127.0.0.1:7897/",
+        },
+    )
+
+    assert result.returncode == 0, result.stderr
+    assert result.stdout.strip() == "ok"
+
+
+def test_tool_executor_parses_object_tool_call_string_arguments() -> None:
+    tool_call = SimpleNamespace(name="echo", arguments='{"text": "hello"}')
+
+    name, arguments = ToolExecutor._normalize_tool_call(tool_call)
+
+    assert name == "echo"
+    assert arguments == {"text": "hello"}