From acec8ff74325430b9eae15afb8556c9b9e326671 Mon Sep 17 00:00:00 2001
From: chiguyong <chiguyong@beyondsoft.com>
Date: Sat, 6 Jun 2026 12:05:56 +0800
Subject: [PATCH] feat(evolution): Phase A - lifecycle hooks + EvolutionConfig

U11: EvolutionMixin integrated into ConfigDrivenAgent lifecycle
  - on_task_complete triggers evolve_after_task
  - on_task_failed records failure patterns
  - Evolution errors never break main task flow

U12: EvolutionConfig added to SkillConfig
  - enabled, reflect_on_failure, auto_apply, min_quality_threshold
  - Backward compatible: defaults to enabled=False

21 new tests passing, no regression.
---
 src/agentkit/core/config_driven.py       |  64 +++-
 src/agentkit/skills/base.py              |  19 ++
 tests/unit/test_evolution_integration.py | 368 +++++++++++++++++++++++
 3 files changed, 450 insertions(+), 1 deletion(-)
 create mode 100644 tests/unit/test_evolution_integration.py

diff --git a/src/agentkit/core/config_driven.py b/src/agentkit/core/config_driven.py
index 7de51d6..d683ea0 100644
--- a/src/agentkit/core/config_driven.py
+++ b/src/agentkit/core/config_driven.py
@@ -16,6 +16,8 @@ import yaml
 from agentkit.core.base import BaseAgent
 from agentkit.core.exceptions import ConfigValidationError
 from agentkit.core.protocol import AgentCapability, TaskMessage
+from agentkit.evolution.lifecycle import EvolutionMixin
+from agentkit.evolution.reflector import Reflector
 from agentkit.prompts.section import PromptSection
 from agentkit.prompts.template import PromptTemplate
 from agentkit.tools.base import Tool
@@ -153,7 +155,7 @@ class AgentConfig:
         return d
 
 
-class ConfigDrivenAgent(BaseAgent):
+class ConfigDrivenAgent(BaseAgent, EvolutionMixin):
     """配置驱动的 Agent
 
     从 YAML/Dict 配置自动组装，支持三种任务模式：
@@ -247,6 +249,28 @@ class ConfigDrivenAgent(BaseAgent):
         from agentkit.quality.gate import QualityGate
         self._quality_gate = QualityGate()
 
+        # v2: Initialize Evolution if configured
+        evolution_config = getattr(config, 'evolution', None)
+        if evolution_config is not None:
+            # Support both dict and EvolutionConfig
+            if isinstance(evolution_config, dict):
+                is_enabled = evolution_config.get("enabled", False)
+            else:
+                is_enabled = getattr(evolution_config, 'enabled', False)
+        else:
+            is_enabled = False
+
+        if is_enabled:
+            reflector = Reflector()
+            EvolutionMixin.__init__(
+                self,
+                reflector=reflector,
+            )
+            self._evolution_enabled = True
+        else:
+            EvolutionMixin.__init__(self)  # Initialize with no components
+            self._evolution_enabled = False
+
         # v2: Initialize Output Standardizer
         from agentkit.quality.output import OutputStandardizer
         self._output_standardizer = OutputStandardizer()
@@ -278,6 +302,44 @@ class ConfigDrivenAgent(BaseAgent):
     def prompt_template(self) -> PromptTemplate | None:
         return self._prompt_template
 
+    async def on_task_complete(self, task: TaskMessage, output: dict) -> None:
+        """Task complete hook - trigger evolution if enabled"""
+        if self._evolution_enabled:
+            try:
+                from agentkit.core.protocol import TaskResult, TaskStatus
+                from datetime import datetime, timezone
+                result = TaskResult(
+                    task_id=task.task_id,
+                    agent_name=self.name,
+                    status=TaskStatus.COMPLETED,
+                    output_data=output,
+                    error_message=None,
+                    started_at=datetime.now(timezone.utc),
+                    completed_at=datetime.now(timezone.utc),
+                )
+                await self.evolve_after_task(task, result)
+            except Exception as e:
+                logger.warning(f"Evolution after task failed: {e}")
+
+    async def on_task_failed(self, task: TaskMessage, error: Exception) -> None:
+        """Task failed hook - record failure for evolution"""
+        if self._evolution_enabled:
+            try:
+                from agentkit.core.protocol import TaskResult, TaskStatus
+                from datetime import datetime, timezone
+                result = TaskResult(
+                    task_id=task.task_id,
+                    agent_name=self.name,
+                    status=TaskStatus.FAILED,
+                    output_data=None,
+                    error_message=str(error),
+                    started_at=datetime.now(timezone.utc),
+                    completed_at=datetime.now(timezone.utc),
+                )
+                await self.evolve_after_task(task, result)
+            except Exception as e:
+                logger.warning(f"Evolution after task failure failed: {e}")
+
     def _bind_tools(self) -> None:
         """根据配置绑定工具"""
         for tool_name in self._config.tools:
diff --git a/src/agentkit/skills/base.py b/src/agentkit/skills/base.py
index 6e95ecb..919ff8f 100644
--- a/src/agentkit/skills/base.py
+++ b/src/agentkit/skills/base.py
@@ -11,6 +11,16 @@ from agentkit.tools.base import Tool
 logger = logging.getLogger(__name__)
 
 
+@dataclass
+class EvolutionConfig:
+    """Evolution configuration"""
+
+    enabled: bool = False
+    reflect_on_failure: bool = True  # Whether to reflect on failed tasks
+    auto_apply: bool = False  # Whether to auto-apply optimizations (without AB test)
+    min_quality_threshold: float = 0.5  # Minimum quality score to trigger optimization
+
+
 @dataclass
 class IntentConfig:
     """意图配置"""
@@ -59,6 +69,7 @@ class SkillConfig(AgentConfig):
         quality_gate: dict[str, Any] | None = None,
         execution_mode: str = "react",
         max_steps: int = 5,
+        evolution: dict[str, Any] | None = None,
     ):
         super().__init__(
             name=name,
@@ -80,6 +91,7 @@ class SkillConfig(AgentConfig):
         self.quality_gate = QualityGateConfig(**(quality_gate or {}))
         self.execution_mode = execution_mode
         self.max_steps = max_steps
+        self.evolution = EvolutionConfig(**(evolution or {}))
         self._validate_v2()
 
     def _validate_v2(self) -> None:
@@ -116,6 +128,7 @@ class SkillConfig(AgentConfig):
             quality_gate=data.get("quality_gate"),
             execution_mode=data.get("execution_mode", "react"),
             max_steps=data.get("max_steps", 5),
+            evolution=data.get("evolution"),
         )
 
     @classmethod
@@ -149,6 +162,12 @@ class SkillConfig(AgentConfig):
         }
         d["execution_mode"] = self.execution_mode
         d["max_steps"] = self.max_steps
+        d["evolution"] = {
+            "enabled": self.evolution.enabled,
+            "reflect_on_failure": self.evolution.reflect_on_failure,
+            "auto_apply": self.evolution.auto_apply,
+            "min_quality_threshold": self.evolution.min_quality_threshold,
+        }
         return d
 
 
diff --git a/tests/unit/test_evolution_integration.py b/tests/unit/test_evolution_integration.py
new file mode 100644
index 0000000..737efc9
--- /dev/null
+++ b/tests/unit/test_evolution_integration.py
@@ -0,0 +1,368 @@
+"""U11+U12 测试: Evolution 生命周期集成 + EvolutionConfig
+
+覆盖：
+- EvolutionConfig 默认值与自定义值
+- SkillConfig 的 evolution 字段
+- ConfigDrivenAgent 集成 EvolutionMixin
+- 生命周期钩子触发进化
+- 进化失败不影响主任务流程
+"""
+
+from datetime import datetime, timezone
+from unittest.mock import AsyncMock, MagicMock, patch
+
+import pytest
+
+from agentkit.core.protocol import TaskMessage, TaskResult, TaskStatus
+
+
+# ── Helpers ──────────────────────────────────────────────
+
+
+def _make_task(**overrides) -> TaskMessage:
+    defaults = dict(
+        task_id="test-task-001",
+        agent_name="test_agent",
+        task_type="generate",
+        priority=1,
+        input_data={"query": "hello"},
+        callback_url=None,
+        created_at=datetime.now(timezone.utc),
+    )
+    defaults.update(overrides)
+    return TaskMessage(**defaults)
+
+
+def _make_task_result(**overrides) -> TaskResult:
+    defaults = dict(
+        task_id="test-task-001",
+        agent_name="test_agent",
+        status=TaskStatus.COMPLETED,
+        output_data={"result": "ok"},
+        error_message=None,
+        started_at=datetime.now(timezone.utc),
+        completed_at=datetime.now(timezone.utc),
+    )
+    defaults.update(overrides)
+    return TaskResult(**defaults)
+
+
+# ── EvolutionConfig 测试 ──────────────────────────────────
+
+
+class TestEvolutionConfig:
+    """U12: EvolutionConfig 数据类测试"""
+
+    def test_default_values(self):
+        """默认 EvolutionConfig — enabled=False"""
+        from agentkit.skills.base import EvolutionConfig
+
+        config = EvolutionConfig()
+        assert config.enabled is False
+        assert config.reflect_on_failure is True
+        assert config.auto_apply is False
+        assert config.min_quality_threshold == 0.5
+
+    def test_from_dict_all_fields(self):
+        """EvolutionConfig 从字典创建 — 所有字段设置"""
+        from agentkit.skills.base import EvolutionConfig
+
+        config = EvolutionConfig(
+            enabled=True,
+            reflect_on_failure=False,
+            auto_apply=True,
+            min_quality_threshold=0.8,
+        )
+        assert config.enabled is True
+        assert config.reflect_on_failure is False
+        assert config.auto_apply is True
+        assert config.min_quality_threshold == 0.8
+
+    def test_from_dict_partial(self):
+        """EvolutionConfig 部分字段 — 缺失字段使用默认值"""
+        from agentkit.skills.base import EvolutionConfig
+
+        config = EvolutionConfig(enabled=True)
+        assert config.enabled is True
+        assert config.reflect_on_failure is True  # default
+        assert config.auto_apply is False  # default
+        assert config.min_quality_threshold == 0.5  # default
+
+
+# ── SkillConfig evolution 字段测试 ─────────────────────────
+
+
+class TestSkillConfigEvolution:
+    """U12: SkillConfig 的 evolution 字段"""
+
+    def test_skill_config_without_evolution(self):
+        """SkillConfig 无 evolution — 默认 enabled=False"""
+        from agentkit.skills.base import SkillConfig
+
+        config = SkillConfig(
+            name="test_agent",
+            agent_type="test",
+            task_mode="llm_generate",
+            prompt={"identity": "test", "instructions": "test"},
+        )
+        assert config.evolution.enabled is False
+
+    def test_skill_config_with_evolution(self):
+        """SkillConfig 有 evolution 配置 — 正确解析"""
+        from agentkit.skills.base import SkillConfig
+
+        config = SkillConfig(
+            name="test_agent",
+            agent_type="test",
+            task_mode="llm_generate",
+            prompt={"identity": "test", "instructions": "test"},
+            evolution={"enabled": True, "auto_apply": True, "min_quality_threshold": 0.7},
+        )
+        assert config.evolution.enabled is True
+        assert config.evolution.auto_apply is True
+        assert config.evolution.min_quality_threshold == 0.7
+
+    def test_skill_config_to_dict_includes_evolution(self):
+        """SkillConfig.to_dict 包含 evolution 字段"""
+        from agentkit.skills.base import SkillConfig
+
+        config = SkillConfig(
+            name="test_agent",
+            agent_type="test",
+            task_mode="llm_generate",
+            prompt={"identity": "test", "instructions": "test"},
+            evolution={"enabled": True},
+        )
+        d = config.to_dict()
+        assert "evolution" in d
+        assert d["evolution"]["enabled"] is True
+        assert d["evolution"]["reflect_on_failure"] is True
+        assert d["evolution"]["auto_apply"] is False
+        assert d["evolution"]["min_quality_threshold"] == 0.5
+
+    def test_skill_config_from_dict_with_evolution(self):
+        """SkillConfig.from_dict 正确解析 evolution"""
+        from agentkit.skills.base import SkillConfig
+
+        data = {
+            "name": "test_agent",
+            "agent_type": "test",
+            "task_mode": "llm_generate",
+            "prompt": {"identity": "test", "instructions": "test"},
+            "evolution": {"enabled": True, "reflect_on_failure": False},
+        }
+        config = SkillConfig.from_dict(data)
+        assert config.evolution.enabled is True
+        assert config.evolution.reflect_on_failure is False
+
+
+# ── ConfigDrivenAgent evolution 集成测试 ──────────────────
+
+
+class TestConfigDrivenAgentEvolution:
+    """U11: ConfigDrivenAgent 集成 EvolutionMixin"""
+
+    def _make_agent_config(self, evolution=None):
+        from agentkit.core.config_driven import AgentConfig
+
+        config = AgentConfig(
+            name="test_agent",
+            agent_type="test",
+            task_mode="llm_generate",
+            prompt={"identity": "test", "instructions": "test"},
+        )
+        if evolution is not None:
+            config.evolution = evolution
+        return config
+
+    def _make_skill_config(self, evolution=None):
+        from agentkit.skills.base import SkillConfig
+
+        return SkillConfig(
+            name="test_agent",
+            agent_type="test",
+            task_mode="llm_generate",
+            prompt={"identity": "test", "instructions": "test"},
+            evolution=evolution,
+        )
+
+    def test_agent_without_evolution_config(self):
+        """Agent 无 evolution 配置 — _evolution_enabled=False"""
+        from agentkit.core.config_driven import ConfigDrivenAgent
+
+        config = self._make_agent_config()
+        agent = ConfigDrivenAgent(config=config)
+        assert agent._evolution_enabled is False
+
+    def test_agent_with_evolution_enabled(self):
+        """Agent 有 evolution 且 enabled=True — _evolution_enabled=True"""
+        from agentkit.core.config_driven import ConfigDrivenAgent
+
+        config = self._make_agent_config(evolution={"enabled": True})
+        agent = ConfigDrivenAgent(config=config)
+        assert agent._evolution_enabled is True
+
+    def test_agent_with_evolution_disabled(self):
+        """Agent 有 evolution 但 enabled=False — _evolution_enabled=False"""
+        from agentkit.core.config_driven import ConfigDrivenAgent
+
+        config = self._make_agent_config(evolution={"enabled": False})
+        agent = ConfigDrivenAgent(config=config)
+        assert agent._evolution_enabled is False
+
+    async def test_on_task_complete_evolution_disabled(self):
+        """on_task_complete 进化禁用 — 不调用 evolve_after_task"""
+        from agentkit.core.config_driven import ConfigDrivenAgent
+
+        config = self._make_agent_config()
+        agent = ConfigDrivenAgent(config=config)
+
+        task = _make_task()
+        output = {"result": "ok"}
+
+        # Should not raise and should not call evolve_after_task
+        await agent.on_task_complete(task, output)
+
+    async def test_on_task_complete_evolution_enabled(self):
+        """on_task_complete 进化启用 — 调用 evolve_after_task"""
+        from agentkit.core.config_driven import ConfigDrivenAgent
+
+        config = self._make_agent_config(evolution={"enabled": True})
+        agent = ConfigDrivenAgent(config=config)
+
+        task = _make_task()
+        output = {"result": "ok"}
+
+        with patch.object(agent, "evolve_after_task", new_callable=AsyncMock) as mock_evolve:
+            await agent.on_task_complete(task, output)
+            mock_evolve.assert_called_once()
+            # Verify the TaskResult passed to evolve_after_task
+            call_args = mock_evolve.call_args
+            result_arg = call_args[0][1]  # second positional arg is TaskResult
+            assert result_arg.status == TaskStatus.COMPLETED
+            assert result_arg.output_data == output
+
+    async def test_on_task_failed_evolution_enabled(self):
+        """on_task_failed 进化启用 — 调用 evolve_after_task"""
+        from agentkit.core.config_driven import ConfigDrivenAgent
+
+        config = self._make_agent_config(evolution={"enabled": True})
+        agent = ConfigDrivenAgent(config=config)
+
+        task = _make_task()
+        error = ValueError("test error")
+
+        with patch.object(agent, "evolve_after_task", new_callable=AsyncMock) as mock_evolve:
+            await agent.on_task_failed(task, error)
+            mock_evolve.assert_called_once()
+            # Verify the TaskResult passed to evolve_after_task
+            call_args = mock_evolve.call_args
+            result_arg = call_args[0][1]  # second positional arg is TaskResult
+            assert result_arg.status == TaskStatus.FAILED
+            assert result_arg.error_message == "test error"
+
+    async def test_evolution_failure_does_not_break_task(self):
+        """进化失败不影响任务完成"""
+        from agentkit.core.config_driven import ConfigDrivenAgent
+
+        config = self._make_agent_config(evolution={"enabled": True})
+        agent = ConfigDrivenAgent(config=config)
+
+        task = _make_task()
+        output = {"result": "ok"}
+
+        with patch.object(agent, "evolve_after_task", new_callable=AsyncMock, side_effect=RuntimeError("evolution crashed")):
+            # Should NOT raise — evolution failure is caught
+            await agent.on_task_complete(task, output)
+
+    async def test_evolution_failure_on_task_failed_does_not_break(self):
+        """进化失败不影响 on_task_failed"""
+        from agentkit.core.config_driven import ConfigDrivenAgent
+
+        config = self._make_agent_config(evolution={"enabled": True})
+        agent = ConfigDrivenAgent(config=config)
+
+        task = _make_task()
+        error = ValueError("task error")
+
+        with patch.object(agent, "evolve_after_task", new_callable=AsyncMock, side_effect=RuntimeError("evolution crashed")):
+            # Should NOT raise
+            await agent.on_task_failed(task, error)
+
+    def test_skill_config_evolution_propagated(self):
+        """SkillConfig 的 evolution 配置传递到 ConfigDrivenAgent"""
+        from agentkit.core.config_driven import ConfigDrivenAgent
+
+        config = self._make_skill_config(evolution={"enabled": True})
+        agent = ConfigDrivenAgent(config=config)
+        assert agent._evolution_enabled is True
+
+
+# ── EvolutionMixin 集成测试 ───────────────────────────────
+
+
+class TestEvolutionMixinIntegration:
+    """U11: EvolutionMixin 方法集成到 ConfigDrivenAgent"""
+
+    def _make_agent_with_evolution(self):
+        from agentkit.core.config_driven import AgentConfig, ConfigDrivenAgent
+
+        config = AgentConfig(
+            name="test_agent",
+            agent_type="test",
+            task_mode="llm_generate",
+            prompt={"identity": "test", "instructions": "test"},
+        )
+        config.evolution = {"enabled": True}
+        return ConfigDrivenAgent(config=config)
+
+    def test_agent_has_get_evolution_history(self):
+        """Agent 继承 get_evolution_history 方法"""
+        from agentkit.core.config_driven import ConfigDrivenAgent
+        from agentkit.evolution.lifecycle import EvolutionMixin
+
+        agent = self._make_agent_with_evolution()
+        assert hasattr(agent, "get_evolution_history")
+        assert callable(agent.get_evolution_history)
+
+    def test_agent_has_set_current_module(self):
+        """Agent 继承 set_current_module 方法"""
+        from agentkit.core.config_driven import ConfigDrivenAgent
+        from agentkit.evolution.lifecycle import EvolutionMixin
+
+        agent = self._make_agent_with_evolution()
+        assert hasattr(agent, "set_current_module")
+        assert callable(agent.set_current_module)
+
+    def test_get_evolution_history_empty_initially(self):
+        """get_evolution_history 初始返回空列表"""
+        agent = self._make_agent_with_evolution()
+        history = agent.get_evolution_history()
+        assert history == []
+
+    def test_set_current_module_works(self):
+        """set_current_module 正常工作"""
+        from agentkit.evolution.prompt_optimizer import Module, Signature
+
+        agent = self._make_agent_with_evolution()
+        signature = Signature(
+            input_fields={"query": "user query"},
+            output_fields={"result": "result"},
+            instruction="test instructions",
+        )
+        module = Module(name="test_module", signature=signature)
+        agent.set_current_module(module)
+        assert agent._current_module is not None
+        assert agent._current_module.name == "test_module"
+
+    def test_mro_correct(self):
+        """MRO 正确: ConfigDrivenAgent → BaseAgent → EvolutionMixin"""
+        from agentkit.core.config_driven import ConfigDrivenAgent
+        from agentkit.core.base import BaseAgent
+        from agentkit.evolution.lifecycle import EvolutionMixin
+
+        mro = ConfigDrivenAgent.__mro__
+        # BaseAgent should come before EvolutionMixin in MRO
+        base_idx = mro.index(BaseAgent)
+        mixin_idx = mro.index(EvolutionMixin)
+        assert base_idx < mixin_idx, f"BaseAgent (idx={base_idx}) should come before EvolutionMixin (idx={mixin_idx}) in MRO"