0
代码
代码
Issues
Pull Requests
讨论
Wiki
分析
0
OopenJiuwen-botfix(evolving):Evaluation of Pipeline Module Refactoring and UT Supplement
edea53e8创建于 3 天前历史提交
文件最后提交记录最后更新时间
agent_rl
fix: fix typing error Co-authored-by: iamcandiceguo<guoqian14@huawei.com> # message auto-generated for no-merge-commit merge: !1503 fix: fix typing error From: @iamcandiceguo Reviewed-by: @deyang, @wangxiaolong100 See merge request: openJiuwen/agent-core!15037 天前
checkpointing
fix(evolution): harden skill evolution approval persistence - distinguish missing skill definitions, generation failures, and persistence failures - make evolution record persistence atomic per record with rollback - support selective approval counts and retry-safe pending payloads Refs: #975 Co-authored-by: zhangzhongteng<zhangzhongteng1@huawei.com> # message auto-generated for no-merge-commit merge: !1524 fix(skill-evolution): harden skill evolution approval persistence From: @z2tng Reviewed-by: @deyang, @iamcandiceguo See merge request: openJiuwen/agent-core!15244 天前
dataset
feat(agent_evolving): Add case data structure and loader - Add Case class for training/evaluation samples with inputs, label, tools - Add EvaluatedCase class with metric storage, automatic score clamping - Add CaseLoader with iteration, split, and shuffle support Refs: #327 Co-authored-by: zhangzhongteng<zhangzhongteng1@huawei.com> # message auto-generated for no-merge-commit merge: !393 feat(agent_evolving): Add case data structure and loader for agent evolving From: @z2tng Reviewed-by: @deyang, @yangzequ See merge request: openJiuwen/agent-core!3933 个月前
evaluator
fix(evolving):Evaluation of Pipeline Module Refactoring and UT Supplement Refs: #879 Co-authored-by: chenyan<chenyan264@huawei.com> # message auto-generated for no-merge-commit merge: !1508 fix(evolving):Evaluation of Pipeline Module Refactoring and UT Supplement From: @dev_chenyan Reviewed-by: @deyang, @iamcandiceguo See merge request: openJiuwen/agent-core!15083 天前
experience
fix(evolution): harden skill evolution approval persistence - distinguish missing skill definitions, generation failures, and persistence failures - make evolution record persistence atomic per record with rollback - support selective approval counts and retry-safe pending payloads Refs: #975 Co-authored-by: zhangzhongteng<zhangzhongteng1@huawei.com> # message auto-generated for no-merge-commit merge: !1524 fix(skill-evolution): harden skill evolution approval persistence From: @z2tng Reviewed-by: @deyang, @iamcandiceguo See merge request: openJiuwen/agent-core!15244 天前
optimizer
fix(evolution): harden skill evolution approval persistence - distinguish missing skill definitions, generation failures, and persistence failures - make evolution record persistence atomic per record with rollback - support selective approval counts and retry-safe pending payloads Refs: #975 Co-authored-by: zhangzhongteng<zhangzhongteng1@huawei.com> # message auto-generated for no-merge-commit merge: !1524 fix(skill-evolution): harden skill evolution approval persistence From: @z2tng Reviewed-by: @deyang, @iamcandiceguo See merge request: openJiuwen/agent-core!15244 天前
sharing
feat(evolving): agent evolving sharing module refs: #880 Co-authored-by: rolandyang<yangdongdong34@huawei.com> # message auto-generated for no-merge-commit merge: !1459 feat(evolving): agent evolving sharing module From: @rolandyang Reviewed-by: @deyang, @iamcandiceguo See merge request: openJiuwen/agent-core!145910 天前
signal
feat(evolving): agent evolving sharing module refs: #880 Co-authored-by: rolandyang<yangdongdong34@huawei.com> # message auto-generated for no-merge-commit merge: !1459 feat(evolving): agent evolving sharing module From: @rolandyang Reviewed-by: @deyang, @iamcandiceguo See merge request: openJiuwen/agent-core!145910 天前
trainer
refactor(agent_evolving): make optimizer base async and integrate EvolutionSignal - BaseOptimizer.backward() changed to async with _bad_signals filter - InstructionOptimizer fully async: _backward() precomputes, _step() sync - Updater protocol and Trainer updated for async calls Refs: #713 Co-authored-by: zhangzhongteng<zhangzhongteng1@huawei.com> # message auto-generated for no-merge-commit merge: !1083 refactor(agent_evolving): make optimizer base async and integrate EvolutionSignal From: @z2tng Reviewed-by: @deyang, @SnapeK See merge request: openJiuwen/agent-core!10831 个月前
trajectory
feat(evolution): add runtime team trajectory registry Refs: #915 Co-authored-by: zhangzhongteng<zhangzhongteng1@huawei.com> # message auto-generated for no-merge-commit merge: !1374 docs(agent_evolving): document runtime trajectory registry From: @z2tng Reviewed-by: @deyang, @iamcandiceguo See merge request: openJiuwen/agent-core!137417 天前
updater
fix(agent-evolving): modularize and tighten skill evolution pipelines Refactor agent evolution into explicit checkpointing, experience, signal, trajectory, and updater pipelines. - Add shared approval events and approval runtime - Extract experience scoring, tracking, lifecycle, and manager modules - Extract checkpointing archive, record, and projection helpers - Add team skill experience generation and signal detection - Accumulate session trajectories for evolution signal handling - Improve evolution experience index guidance in projected SKILL.md - Update API docs, examples, and unit coverage for the new layout BREAKING CHANGE: remove the legacy skill rewrite API and SkillRewriter exports. Refs: #890 Co-authored-by: zhangzhongteng<zhangzhongteng1@huawei.com> # message auto-generated for no-merge-commit merge: !1337 fix(agent-evolving): modularize and tighten skill evolution pipelines From: @z2tng Reviewed-by: @deyang, @SnapeK See merge request: openJiuwen/agent-core!133721 天前
__init__.py
feat(agent_rl): Move to evolving dir and supports online module Move agentrl to agent_evolving directory and suport online/offline agent rl. Refs: #727 Co-authored-by: rolandyang<yangdongdong34@huawei.com> # message auto-generated for no-merge-commit merge: !1065 feat(agent_rl): Move to evolving dir and supports online module From: @rolandyang Reviewed-by: @deyang, @iamcandiceguo See merge request: openJiuwen/agent-core!10651 个月前
constant.py
feat(agent_evolving): Add hyperparameter management and utility modules - Add TuneConstant class with default values and valid ranges for iterations, parallel count - Add TuneUtils utility class for parameter validation, serialization, LLM response parsing - Export all public APIs uniformly Refs: #330 Co-authored-by: zhangzhongteng<zhangzhongteng1@huawei.com> # message auto-generated for no-merge-commit merge: !402 feat(agent_evolving): Add hyperparameter management and utility modules From: @z2tng Reviewed-by: @deyang, @yangzequ See merge request: openJiuwen/agent-core!4023 个月前
protocols.py
fix(agent-evolving): modularize and tighten skill evolution pipelines Refactor agent evolution into explicit checkpointing, experience, signal, trajectory, and updater pipelines. - Add shared approval events and approval runtime - Extract experience scoring, tracking, lifecycle, and manager modules - Extract checkpointing archive, record, and projection helpers - Add team skill experience generation and signal detection - Accumulate session trajectories for evolution signal handling - Improve evolution experience index guidance in projected SKILL.md - Update API docs, examples, and unit coverage for the new layout BREAKING CHANGE: remove the legacy skill rewrite API and SkillRewriter exports. Refs: #890 Co-authored-by: zhangzhongteng<zhangzhongteng1@huawei.com> # message auto-generated for no-merge-commit merge: !1337 fix(agent-evolving): modularize and tighten skill evolution pipelines From: @z2tng Reviewed-by: @deyang, @SnapeK See merge request: openJiuwen/agent-core!133721 天前
types.py
fix(agent-evolving): modularize and tighten skill evolution pipelines Refactor agent evolution into explicit checkpointing, experience, signal, trajectory, and updater pipelines. - Add shared approval events and approval runtime - Extract experience scoring, tracking, lifecycle, and manager modules - Extract checkpointing archive, record, and projection helpers - Add team skill experience generation and signal detection - Accumulate session trajectories for evolution signal handling - Improve evolution experience index guidance in projected SKILL.md - Update API docs, examples, and unit coverage for the new layout BREAKING CHANGE: remove the legacy skill rewrite API and SkillRewriter exports. Refs: #890 Co-authored-by: zhangzhongteng<zhangzhongteng1@huawei.com> # message auto-generated for no-merge-commit merge: !1337 fix(agent-evolving): modularize and tighten skill evolution pipelines From: @z2tng Reviewed-by: @deyang, @SnapeK See merge request: openJiuwen/agent-core!133721 天前
update_execution.py
fix(agent-evolving): modularize and tighten skill evolution pipelines Refactor agent evolution into explicit checkpointing, experience, signal, trajectory, and updater pipelines. - Add shared approval events and approval runtime - Extract experience scoring, tracking, lifecycle, and manager modules - Extract checkpointing archive, record, and projection helpers - Add team skill experience generation and signal detection - Accumulate session trajectories for evolution signal handling - Improve evolution experience index guidance in projected SKILL.md - Update API docs, examples, and unit coverage for the new layout BREAKING CHANGE: remove the legacy skill rewrite API and SkillRewriter exports. Refs: #890 Co-authored-by: zhangzhongteng<zhangzhongteng1@huawei.com> # message auto-generated for no-merge-commit merge: !1337 fix(agent-evolving): modularize and tighten skill evolution pipelines From: @z2tng Reviewed-by: @deyang, @SnapeK See merge request: openJiuwen/agent-core!133721 天前
utils.py
fix(team-skill): trigger creation after team completion Register team completion callbacks for both team-skill evolution and creation rails, and only enqueue creation prompts after the completed session meets the spawn-member threshold. Preserve leader trajectories by recording member roles, merge repeated member trajectory snapshots, and treat both team-skill and swarm-skill frontmatter kinds as team skills for detection and filtering. Refs: #907 Co-authored-by: zhangzhongteng<zhangzhongteng1@huawei.com> # message auto-generated for no-merge-commit merge: !1359 fix(team-skill): trigger creation after team completion From: @z2tng Reviewed-by: @deyang, @iamcandiceguo See merge request: openJiuwen/agent-core!135918 天前