feat(runner): 注入 tool_dispatch_factory/prompt_builder_factory + fail-fast

Runner.__init__ 新增 2 个可选参数:
- tool_dispatch_factory: 工具调度工厂
- prompt_builder_factory: prompt 构建工厂

infer/eval/train 模式缺少工厂时 fail-fast 抛 ValueError。
_make_tool_dispatch_fn/_make_prompt_builder 优先使用注入工厂。

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

app/harness/runner.py

@@ -449,6 +449,8 @@ class Runner:
         evolve_llm: 进化用 LLMProvider（thinking=True）。
         vlm: VLMProvider。
         telemetry: 遥测记录端口。
+        tool_dispatch_factory: 工具调度工厂（infer/eval/train 模式必传）。
+        prompt_builder_factory: prompt 构建工厂（infer/eval/train 模式必传）。
     """
 
     def __init__(
@@ -459,12 +461,28 @@ class Runner:
         evolve_llm: LLMProvider,
         vlm: VLMProvider,
         telemetry: TelemetryRecorder,
+        tool_dispatch_factory: Any | None = None,
+        prompt_builder_factory: Any | None = None,
     ) -> None:
         self._config = config
         self._llm = llm
         self._evolve_llm = evolve_llm
         self._vlm = vlm
         self._telemetry = telemetry
+        self._tool_dispatch_factory = tool_dispatch_factory
+        self._prompt_builder_factory = prompt_builder_factory
+
+        # fail-fast: 需要推理的模式必须注入工厂
+        if config.mode in {"infer", "eval", "train"} and (
+            tool_dispatch_factory is None or prompt_builder_factory is None
+        ):
+            raise ValueError(
+                f"mode={config.mode!r} 需要 tool_dispatch_factory 和 "
+                f"prompt_builder_factory，但收到 "
+                f"tool_dispatch_factory={tool_dispatch_factory!r}, "
+                f"prompt_builder_factory={prompt_builder_factory!r}"
+            )
+
         self._ensure_workspace()
         self._paths: ResolvedPaths = resolve_paths(config.workspace_dir)
 
@@ -2062,22 +2080,26 @@ class Runner:
     # -----------------------------------------------------------------------
 
     def _make_tool_dispatch_fn(self, *, skills_dir: Path | None = None):
-        """构造工具调度函数（由子类或 main.py 覆盖）。"""
+        """构造工具调度函数（优先使用注入的工厂，否则 noop 降级）。"""
+        if self._tool_dispatch_factory is not None:
+            return self._tool_dispatch_factory(skills_dir=skills_dir)
 
+        # noop fallback：diagnose/promote 等不需要推理的模式
         async def _noop_dispatch(tool_name: str, args: dict, *, context: dict) -> str:
-            raise NotImplementedError(
+            raise NotImplementedError(f"工具 {tool_name} 调度未配置")
-                f"工具 {tool_name} 调度未配置（需由 main.py 注入 tool_dispatch_fn）"
-            )
 
         return _noop_dispatch
 
     def _make_prompt_builder(
         self, *, skills_dir: Path | None = None, prompts_dir: Path | None = None
     ):
-        """构造 prompt 构建函数（由子类或 main.py 覆盖）。"""
+        """构造 prompt 构建函数（优先使用注入的工厂，否则 noop 降级）。"""
+        if self._prompt_builder_factory is not None:
+            return self._prompt_builder_factory(skills_dir=skills_dir, prompts_dir=prompts_dir)
 
+        # noop fallback：diagnose/promote 等不需要推理的模式
         def _noop_builder(qa: GeneratedQuestion) -> tuple[str, str]:
-            raise NotImplementedError("prompt_builder 未配置（需由 main.py 注入）")
+            raise NotImplementedError("prompt_builder 未配置")
 
         return _noop_builder
 

tests/unit/test_harness_runner.py

@@ -14,6 +14,7 @@ from unittest.mock import MagicMock, patch
 import pytest
 
 from app.harness.runner import (
+    Runner,
     _apply_batch_correctness,
     _batch_from_ids,
     _build_comparison_pairs,
@@ -680,3 +681,94 @@ class TestCooldownDecrement:
 
         cooldown = {t: n - 1 for t, n in cooldown.items() if n - 1 > 0}
         assert cooldown == {}
+
+
+# =========================================================================
+# factory 注入（Task 4）
+# =========================================================================
+
+
+class TestRunnerFactoryInjection:
+    """Runner 构造时 tool_dispatch_factory / prompt_builder_factory 注入检查。"""
+
+    @staticmethod
+    def _base_config(tmp_path: Path, *, mode: str = "infer", **overrides):
+        """构造 RunConfig，所有必填字段都给默认值。"""
+        from app.harness.config import RunConfig
+
+        defaults = {
+            "workspace_dir": tmp_path,
+            "store_dir": tmp_path,
+            "mode": mode,
+            "concurrency": 1,
+            "max_steps": 5,
+            "skill_mode": "none",
+            "n_samples": 0,
+            "questions": "benchmarks/Video-MME",
+            "skills_version": "v1",
+            "prompts_version": "v1",
+            "epochs": 1,
+            "diag_size": 10,
+            "diag_correct_ratio": 0.5,
+            "val_size": 24,
+            "val_correct_ratio": 0.5,
+            "edit_budget_start": 5,
+            "edit_budget_end": 2,
+            "batch_size": 5,
+            "min_class_per_batch": 2,
+            "eval_min_per_class": 2,
+            "early_stop_patience": 3,
+            "test_size": 10,
+            "use_slow_momentum": False,
+            "gate_e_confirm": 20.0,
+            "gate_e_provisional": 3.0,
+            "gate_w_net_min": 2,
+            "gate_delta_min": 0.02,
+            "gate_lambda_dir": -0.642,
+            "gate_e_rollback": 10.0,
+            "gate_block": 8,
+            "gate_n_max": 40,
+            "gate_p_low": 0.05,
+            "gate_p_high": 0.95,
+            "gate_probe_quota": 0.2,
+            "gate_gamma_decay": 0.9,
+            "gate_cooldown_steps": 2,
+            "gate_guard_err": 0.10,
+            "skill_update_mode": "patch",
+            "appendix_consolidate_threshold": 6,
+        }
+        defaults.update(overrides)
+        return RunConfig(**defaults)
+
+    def test_infer_mode_missing_factory_raises(self, tmp_path: Path) -> None:
+        """infer 模式缺少工厂时抛出 ValueError。"""
+        config = self._base_config(tmp_path, mode="infer")
+        with pytest.raises(ValueError, match="tool_dispatch_factory"):
+            Runner(
+                config,
+                llm=MagicMock(),
+                evolve_llm=MagicMock(),
+                vlm=MagicMock(),
+                telemetry=MagicMock(),
+            )
+
+    def test_diagnose_mode_allows_none_factory(self, tmp_path: Path) -> None:
+        """diagnose 模式不需要工厂，允许 None。"""
+        ws = tmp_path / "ws"
+        ws.mkdir()
+        (ws / "manifest.json").write_text(
+            '{"name":"ws","created_at":"","store":"../store",'
+            '"current":{"videos":"v","questions":"q","skills":"s","prompts":"p"},'
+            '"history":[]}'
+        )
+        config = self._base_config(tmp_path, mode="diagnose", workspace_dir=ws, run_id="test_run")
+        # 不应抛出 ValueError
+        runner = Runner(
+            config,
+            llm=MagicMock(),
+            evolve_llm=MagicMock(),
+            vlm=MagicMock(),
+            telemetry=MagicMock(),
+        )
+        assert runner._tool_dispatch_factory is None
+        assert runner._prompt_builder_factory is None