```
feat(engine): 优化智能体循环中的助手消息处理逻辑 - 在没有工具调用时才添加助手消息到上下文 - 确保工具调用响应正确添加到消息上下文中 - 修复了消息构建的条件逻辑 fix(cron): 改进定时任务调度的时间解析功能 - 添加正则表达式导入用于时间显示解析 - 实现从显示文本中提取毫秒间隔的功能 - 增强整数转换的安全性，避免类型错误 - 优化定时任务配置的解析逻辑 feat(outlook): 增强Outlook集成的功能和稳定性 - 将默认超时时间从10秒增加到180秒 - 为状态检查函数添加可选的验证参数 - 串行执行邮件概览获取操作而非并行 - 改进连接状态验证逻辑 feat(channel): 添加设备名称作为会话标识的选项 - 为终端WebSocket适配器添加新的配置选项 - 实现基于设备名称生成会话对等ID的功能 - 记录原始对等ID和设备名称的元数据 - 支持从设备名称创建会话对等ID feat(skills): 完善技能学习评估系统和进度跟踪 - 在应用启动时自动调度待评估的技能草稿 - 为技能评估工作创建独立的循环工厂 - 实现异步技能评估任务的取消和清理机制 - 添加技能评估进度报告和状态跟踪功能 - 扩展会话列表API以包含更多详细信息 - 防止对不存在的会话进行操作 - 优化技能草稿提交和评估的业务逻辑 perf(skills): 提升技能评估的并发性能 - 实现并行技能案例评估以提高效率 - 添加最大并行案例数的环境变量控制 - 实现实时评估进度更新和回调机制 - 优化评估过程中的资源管理和同步 refactor(services): 创建隔离的智能体循环实例 - 添加创建独立智能体循环的工厂方法 - 确保新循环继承运行时服务配置 - 支持技能评估等需要隔离环境的场景 ```
This commit is contained in:
@ -2,7 +2,7 @@
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
from typing import Any
|
||||
from typing import Any, Callable
|
||||
|
||||
from beaver.engine.providers import ProviderBundle
|
||||
from beaver.memory.skills import SkillDraftEvalReport, SkillDraftSafetyReport, SkillLearningCandidate, SkillLearningStore
|
||||
@ -174,12 +174,20 @@ class SkillLearningPipelineService:
|
||||
safety = self.get_safety_report(skill_name, draft_id)
|
||||
if safety is not None and (not safety.passed or safety.risk_level == "critical"):
|
||||
raise ValueError("Draft cannot enter review because safety check failed")
|
||||
return self.review_service.submit_for_review(
|
||||
review = self.review_service.submit_for_review(
|
||||
skill_name,
|
||||
draft_id,
|
||||
reviewer_request=notes,
|
||||
requested_by=requested_by,
|
||||
)
|
||||
self._mark_candidate_by_draft(
|
||||
skill_name,
|
||||
draft_id,
|
||||
"review_pending",
|
||||
"review_submitted",
|
||||
last_error=None,
|
||||
)
|
||||
return review
|
||||
|
||||
def approve(
|
||||
self,
|
||||
@ -258,9 +266,13 @@ class SkillLearningPipelineService:
|
||||
draft = self.get_draft(skill_name, draft_id)
|
||||
report = self.safety_checker.check(draft)
|
||||
self.learning_store.write_safety_report(report)
|
||||
status = "safety_failed" if not report.passed or report.risk_level == "critical" else "draft_ready"
|
||||
status = (
|
||||
"safety_failed"
|
||||
if not report.passed or report.risk_level == "critical"
|
||||
else self._candidate_status_for_draft(draft)
|
||||
)
|
||||
current = self._candidate_by_draft(skill_name, draft_id)
|
||||
if current is not None and current.status == "eval_failed" and status == "draft_ready":
|
||||
if current is not None and current.status == "eval_failed" and status != "safety_failed":
|
||||
status = "eval_failed"
|
||||
self._mark_candidate_by_draft(
|
||||
skill_name,
|
||||
@ -287,6 +299,7 @@ class SkillLearningPipelineService:
|
||||
*,
|
||||
provider_bundle: ProviderBundle | None,
|
||||
replay_runner: ReplayRunner | None = None,
|
||||
progress_callback: Callable[[dict[str, Any]], None] | None = None,
|
||||
) -> SkillDraftEvalReport:
|
||||
draft = self.get_draft(skill_name, draft_id)
|
||||
candidate = self.get_candidate(candidate_id)
|
||||
@ -296,13 +309,14 @@ class SkillLearningPipelineService:
|
||||
draft=draft,
|
||||
provider_bundle=provider_bundle,
|
||||
replay_runner=replay_runner,
|
||||
progress_callback=progress_callback,
|
||||
)
|
||||
self.learning_store.write_eval_report(report)
|
||||
if report.status == "skipped_provider_unavailable":
|
||||
status = "draft_ready"
|
||||
status = self._candidate_status_for_draft(draft)
|
||||
error = "eval skipped: provider unavailable"
|
||||
elif report.passed:
|
||||
status = "draft_ready"
|
||||
status = self._candidate_status_for_draft(draft)
|
||||
error = None
|
||||
else:
|
||||
status = "eval_failed"
|
||||
@ -316,11 +330,43 @@ class SkillLearningPipelineService:
|
||||
status,
|
||||
event_type="eval_completed",
|
||||
eval_report_id=report.report_id,
|
||||
eval_progress={
|
||||
"phase": "completed",
|
||||
"completed_arms": len(report.cases) * 2 if report.mode == "replay" else 0,
|
||||
"total_arms": len(report.cases) * 2 if report.mode == "replay" else 0,
|
||||
"completed_cases": len(report.cases),
|
||||
"total_cases": len(report.cases),
|
||||
},
|
||||
last_error=error,
|
||||
payload=report.to_dict(),
|
||||
)
|
||||
return report
|
||||
|
||||
def mark_eval_progress(self, candidate_id: str, progress: dict[str, Any]) -> SkillLearningCandidate:
|
||||
return self._require_updated(
|
||||
self.learning_store.update_learning_candidate(
|
||||
candidate_id,
|
||||
eval_progress=dict(progress),
|
||||
),
|
||||
candidate_id,
|
||||
)
|
||||
|
||||
def mark_eval_failed(self, candidate_id: str, error: str) -> SkillLearningCandidate:
|
||||
candidate = self.get_candidate(candidate_id)
|
||||
progress = dict(candidate.eval_progress)
|
||||
progress["phase"] = "failed"
|
||||
return self._require_updated(
|
||||
self.learning_store.transition_learning_candidate(
|
||||
candidate_id,
|
||||
"eval_failed",
|
||||
eval_progress=progress,
|
||||
event_type="eval_failed",
|
||||
last_error=error,
|
||||
payload={"error": error},
|
||||
),
|
||||
candidate_id,
|
||||
)
|
||||
|
||||
def _validate_publish_gates(self, draft: SkillDraft, *, confirm_high_risk: bool) -> None:
|
||||
reviews = self.reviews_for_draft(draft.skill_name, draft.draft_id)
|
||||
if not any(review.status in {SkillReviewState.IN_REVIEW.value, SkillReviewState.APPROVED.value} for review in reviews):
|
||||
@ -372,6 +418,14 @@ class SkillLearningPipelineService:
|
||||
return candidate
|
||||
return None
|
||||
|
||||
@staticmethod
|
||||
def _candidate_status_for_draft(draft: SkillDraft) -> str:
|
||||
if draft.status == SkillReviewState.APPROVED.value:
|
||||
return "approved"
|
||||
if draft.status == SkillReviewState.IN_REVIEW.value:
|
||||
return "review_pending"
|
||||
return "draft_ready"
|
||||
|
||||
@staticmethod
|
||||
def _require_updated(candidate: SkillLearningCandidate | None, candidate_id: str) -> SkillLearningCandidate:
|
||||
if candidate is None:
|
||||
|
||||
Reference in New Issue
Block a user