feat(engine): 添加MCP连接管理和工具集成功能
- 集成MCP连接管理器，支持MCP服务器连接 - 添加多种内置工具：ClarifyTool、CronTool、DelegateTool、ExecuteCodeTool、 PatchFileTool、ProcessTool、SendMessageTool、SpawnTool、TerminalTool、 TodoTool、WebFetchTool、WebSearchTool、WriteFileTool等 - 实现工具注册和装配功能 - 添加技能选择上下文参数 - 支持思考模式控制参数thinking_enabled feat(coordinator): 重构任务执行计划器参数命名 - 将learning_candidate_enabled重命名为allow_candidate_generation - 更新TeamGraphScheduler中的参数传递 - 修改LocalAgentRunner中的相关参数处理 - 更新README文档中的相应描述 refactor(context): 标准化工具调用参数格式 - 添加_json导入用于参数序列化 - 实现_provider_tool_calls方法标准化OpenAI兼容的工具调用载荷 - 修复工具调用中参数非字符串类型的序列化问题 refactor(session): 优化消息历史记录过滤逻辑 - 修改get_messages_as_conversation为基于运行状态过滤消息 - 排除未完成、失败或错误结束的运行记录 - 改进对话历史的可见性控制机制 fix(store): 修复FTS索引重建逻辑 - 添加异常处理防止FTS索引创建失败 - 实现_rebuild_fts_index方法重新构建全文搜索索引 - 优化索引触发器和表的维护流程
This commit is contained in:
@ -12,6 +12,7 @@ from beaver.engine.context.builder import ContextBuilder, ContextBuildInput
|
||||
from beaver.engine.providers.base import LLMProvider, LLMResponse
|
||||
from beaver.engine.providers.factory import ProviderBundle
|
||||
from beaver.services.agent_service import AgentService
|
||||
from beaver.skills.assembler import SkillAssemblyResult
|
||||
from beaver.tasks import TaskExecutionPlan, TaskService, ValidationResult, ValidationService
|
||||
|
||||
|
||||
@ -67,7 +68,25 @@ class FakeLearningCandidate:
|
||||
return {"candidate_id": "candidate-1", "kind": "new_skill", "status": "open"}
|
||||
|
||||
|
||||
def _bundle(*responses: str) -> ProviderBundle:
|
||||
class RecordingSkillAssembler:
|
||||
def __init__(self) -> None:
|
||||
self.task_descriptions: list[str] = []
|
||||
|
||||
async def assemble(self, **kwargs) -> SkillAssemblyResult:
|
||||
self.task_descriptions.append(kwargs["task_description"])
|
||||
return SkillAssemblyResult()
|
||||
|
||||
|
||||
def _route_response(action: str = "new_task", short_title: str = "Test task") -> LLMResponse:
|
||||
return LLMResponse(
|
||||
content=f'{{"action":"{action}","reason":"test route","short_title":"{short_title}"}}',
|
||||
finish_reason="stop",
|
||||
provider_name="stub",
|
||||
model="stub-model",
|
||||
)
|
||||
|
||||
|
||||
def _bundle(*responses: str, route_action: str = "new_task") -> ProviderBundle:
|
||||
return ProviderBundle(
|
||||
main_runtime=SimpleNamespace(model="stub-model", provider_name="stub"),
|
||||
main_provider=StubProvider(
|
||||
@ -81,6 +100,8 @@ def _bundle(*responses: str) -> ProviderBundle:
|
||||
for response in responses
|
||||
]
|
||||
),
|
||||
auxiliary_runtime=SimpleNamespace(model="stub-model", provider_name="stub"),
|
||||
auxiliary_provider=StubProvider([_route_response(route_action)]),
|
||||
)
|
||||
|
||||
|
||||
@ -110,6 +131,25 @@ def _provider_bundle(provider: StubProvider) -> ProviderBundle:
|
||||
return ProviderBundle(
|
||||
main_runtime=SimpleNamespace(model="stub-model", provider_name="stub"),
|
||||
main_provider=provider,
|
||||
auxiliary_runtime=SimpleNamespace(model="stub-model", provider_name="stub"),
|
||||
auxiliary_provider=StubProvider([_route_response("new_task")]),
|
||||
)
|
||||
|
||||
|
||||
def _main_only_bundle(*responses: str) -> ProviderBundle:
|
||||
return ProviderBundle(
|
||||
main_runtime=SimpleNamespace(model="stub-model", provider_name="stub"),
|
||||
main_provider=StubProvider(
|
||||
[
|
||||
LLMResponse(
|
||||
content=response,
|
||||
finish_reason="stop",
|
||||
provider_name="stub",
|
||||
model="stub-model",
|
||||
)
|
||||
for response in responses
|
||||
]
|
||||
),
|
||||
)
|
||||
|
||||
|
||||
@ -126,7 +166,7 @@ def test_simple_question_does_not_create_task(tmp_path: Path) -> None:
|
||||
service.process_direct(
|
||||
"hello?",
|
||||
session_id="web:simple",
|
||||
provider_bundle=_bundle("hi"),
|
||||
provider_bundle=_bundle("hi", route_action="simple_chat"),
|
||||
)
|
||||
)
|
||||
loaded = service.create_loop().boot()
|
||||
@ -165,8 +205,89 @@ def test_complex_request_creates_task_and_records_validation(tmp_path: Path) ->
|
||||
assert any(event.event_type == "task_validation_snapshotted" for event in events)
|
||||
assert run_record.task_id == result.task_id
|
||||
assert run_record.validation_result["accepted"] is True
|
||||
assert skill_effects.event_payload["learning_candidate_enabled"] is False
|
||||
assert skill_effects.event_payload["candidate_generation_allowed"] is False
|
||||
assert skill_effects.event_payload["learning_candidates"] == []
|
||||
assert task.metadata["short_title"] == "Test task"
|
||||
|
||||
|
||||
def test_task_mode_uses_task_aware_skill_selection_context(tmp_path: Path) -> None:
|
||||
skill_assembler = RecordingSkillAssembler()
|
||||
service = AgentService(
|
||||
loader=EngineLoader(
|
||||
workspace=tmp_path,
|
||||
task_execution_planner=_single_planner(),
|
||||
validation_service=StubValidationService(
|
||||
[ValidationResult(passed=True, score=1.0, validator="test")]
|
||||
),
|
||||
skill_assembler=skill_assembler,
|
||||
)
|
||||
)
|
||||
|
||||
result = asyncio.run(
|
||||
service.process_direct(
|
||||
"继续按刚才的方案改",
|
||||
session_id="web:task-skill-query",
|
||||
provider_bundle=_bundle("done", route_action="new_task"),
|
||||
)
|
||||
)
|
||||
|
||||
assert result.task_id
|
||||
assert skill_assembler.task_descriptions
|
||||
query = skill_assembler.task_descriptions[0]
|
||||
assert "Task goal:" in query
|
||||
assert "Current user request:" in query
|
||||
assert "Previously activated skills:" in query
|
||||
assert "If no published skill matches, return []" in query
|
||||
|
||||
|
||||
def test_active_task_continues_until_llm_closes_it(tmp_path: Path) -> None:
|
||||
service = AgentService(
|
||||
loader=EngineLoader(
|
||||
workspace=tmp_path,
|
||||
task_execution_planner=_single_planner(),
|
||||
validation_service=StubValidationService(
|
||||
[
|
||||
ValidationResult(passed=True, score=0.9, validator="test"),
|
||||
ValidationResult(passed=True, score=0.9, validator="test"),
|
||||
]
|
||||
),
|
||||
)
|
||||
)
|
||||
|
||||
first = asyncio.run(
|
||||
service.process_direct(
|
||||
"implement the search workflow",
|
||||
session_id="web:continue",
|
||||
provider_bundle=_bundle("first done", route_action="new_task"),
|
||||
)
|
||||
)
|
||||
second = asyncio.run(
|
||||
service.process_direct(
|
||||
"also add tests for it",
|
||||
session_id="web:continue",
|
||||
provider_bundle=_bundle("tests added", route_action="continue_task"),
|
||||
)
|
||||
)
|
||||
loaded = service.create_loop().boot()
|
||||
task = loaded.task_service.get_task(first.task_id)
|
||||
|
||||
assert task is not None
|
||||
assert second.task_id == first.task_id
|
||||
assert len(task.run_ids) == 2
|
||||
|
||||
closed = asyncio.run(
|
||||
service.process_direct(
|
||||
"这个任务结束了",
|
||||
session_id="web:continue",
|
||||
provider_bundle=_bundle("好的，已结束。", route_action="close_task"),
|
||||
)
|
||||
)
|
||||
task = loaded.task_service.get_task(first.task_id)
|
||||
|
||||
assert closed.task_id is None
|
||||
assert task is not None
|
||||
assert task.status == "closed"
|
||||
assert loaded.task_service.active_task_view("web:continue") is None
|
||||
|
||||
|
||||
def test_validation_failure_retries_once(tmp_path: Path) -> None:
|
||||
@ -229,11 +350,11 @@ def test_feedback_closes_or_abandons_internal_task(tmp_path: Path) -> None:
|
||||
loaded = service.create_loop().boot()
|
||||
learning_calls = []
|
||||
|
||||
def build_learning_candidates() -> list[FakeLearningCandidate]:
|
||||
learning_calls.append("called")
|
||||
def build_learning_candidates_for_task(task_id: str, *, trigger_run_id: str) -> list[FakeLearningCandidate]:
|
||||
learning_calls.append((task_id, trigger_run_id))
|
||||
return [FakeLearningCandidate()]
|
||||
|
||||
loaded.skill_learning_service.build_learning_candidates = build_learning_candidates
|
||||
loaded.skill_learning_service.build_learning_candidates_for_task = build_learning_candidates_for_task
|
||||
|
||||
feedback = asyncio.run(
|
||||
service.submit_feedback(
|
||||
@ -247,7 +368,7 @@ def test_feedback_closes_or_abandons_internal_task(tmp_path: Path) -> None:
|
||||
assert feedback["learning_candidates"] == [
|
||||
{"candidate_id": "candidate-1", "kind": "new_skill", "status": "open"}
|
||||
]
|
||||
assert learning_calls == ["called"]
|
||||
assert learning_calls == [(result.task_id, result.run_id)]
|
||||
|
||||
service2 = AgentService(
|
||||
loader=EngineLoader(
|
||||
@ -279,6 +400,14 @@ def test_feedback_closes_or_abandons_internal_task(tmp_path: Path) -> None:
|
||||
|
||||
assert abandon_feedback["task_status"] == "abandoned"
|
||||
assert abandon_feedback["learning_candidates"] == []
|
||||
loaded2 = service2.create_loop().boot()
|
||||
failure_events = [
|
||||
event
|
||||
for event in loaded2.session_manager.get_run_event_records(abandoned.session_id, abandoned.run_id)
|
||||
if event.event_type == "task_failure_evidence_recorded"
|
||||
]
|
||||
assert len(failure_events) == 1
|
||||
assert loaded2.memory_service.get_store().memory_entries == []
|
||||
|
||||
|
||||
def test_feedback_is_idempotent_and_projected_to_assistant_message(tmp_path: Path) -> None:
|
||||
@ -466,7 +595,7 @@ def test_task_mode_team_retry_hides_first_synthesis_run(tmp_path: Path) -> None:
|
||||
events = loaded.session_manager.get_run_event_records(record.session_id, run_id)
|
||||
skill_effects = [event for event in events if event.event_type == "skill_effects_snapshotted"]
|
||||
assert skill_effects
|
||||
assert skill_effects[-1].event_payload["learning_candidate_enabled"] is False
|
||||
assert skill_effects[-1].event_payload["candidate_generation_allowed"] is False
|
||||
|
||||
|
||||
def test_context_builder_strips_ui_projection_fields_from_provider_history() -> None:
|
||||
@ -490,17 +619,43 @@ def test_context_builder_strips_ui_projection_fields_from_provider_history() ->
|
||||
assert assistant == {"role": "assistant", "content": "done"}
|
||||
|
||||
|
||||
def test_context_builder_normalizes_persisted_tool_arguments() -> None:
|
||||
result = ContextBuilder().build_messages(
|
||||
ContextBuildInput(
|
||||
history=[
|
||||
{
|
||||
"role": "assistant",
|
||||
"content": None,
|
||||
"tool_calls": [
|
||||
{
|
||||
"id": "call-1",
|
||||
"type": "function",
|
||||
"function": {
|
||||
"name": "cron",
|
||||
"arguments": {"action": "add", "mode": "notification"},
|
||||
},
|
||||
}
|
||||
],
|
||||
}
|
||||
],
|
||||
)
|
||||
)
|
||||
|
||||
tool_call = result.messages[-1]["tool_calls"][0]
|
||||
assert tool_call["function"]["arguments"] == '{"action": "add", "mode": "notification"}'
|
||||
|
||||
|
||||
def test_llm_validator_parse_failure_is_not_accepted(tmp_path: Path) -> None:
|
||||
task_service = TaskService(tmp_path / "tasks")
|
||||
task = task_service.create_task(session_id="web:validator", description="implement validator handling")
|
||||
validation = asyncio.run(
|
||||
ValidationService().validate_task_result(
|
||||
task=task,
|
||||
user_message="implement validator handling",
|
||||
final_output="done",
|
||||
provider_bundle=_bundle("not json"),
|
||||
task=task,
|
||||
user_message="implement validator handling",
|
||||
final_output="done",
|
||||
provider_bundle=_main_only_bundle("not json"),
|
||||
)
|
||||
)
|
||||
)
|
||||
|
||||
assert validation.accepted is False
|
||||
assert validation.validator == "llm_error"
|
||||
|
||||
Reference in New Issue
Block a user