[feat][evaluation]trae eval - #551

Open

tpfz wants to merge 59 commits into

mainfrom

feat/wzq/trae_eval

Commits on Jun 13, 2026

feat(evaluation): update init-sql DDL and regenerate gorm_gen models for item-centric experiment refactoring

xueyizheng
and
claude
committed
2e867ad
feat(evaluation/idl): add multi-evalset IDL structs for item-centric experiment

xueyizheng
and
claude
committed
70ccc44
feat(evaluation/entity): add item-centric domain entities and upgrade EvaluatorResults

xueyizheng
and
claude
committed
7c3d050
feat(evaluation/domain): add IExptItemRefRepo interface and ExptEvalSetItemCount entity

xueyizheng
and
claude
committed
06e5314
feat(evaluation/infra): implement ExptItemRefRepo and propagate new fields in converters

xueyizheng
and
claude
committed
3fc9a12
feat(evaluation): CreateExperiment multi-evalset path and application convertor

xueyizheng
and
claude
committed
571c747
feat(evaluation): ExptStart multi-set path + storeTurnRunResult EvaluatorResults upgrade

xueyizheng
and
claude
committed
af61dd8
test(evaluation): add unit tests for item-centric experiment refactoring

xueyizheng
and
claude
committed
adb7cbd

Commits on Jun 14, 2026

fix(evaluation): upgrade callEvaluators recordMap key to prevent alias collision

xueyizheng
and
claude
committed
eba6fe3
fix
xueyizheng
committed
fd52d96
feat(eval): AI创建实验接口修改记录

alanyf
authored and
xueyizheng
committed
e71112e
fix
xueyizheng
committed
4ca8dba
feat(eval): aggr 按维度分流 MultiSetConfig 跳过 evaluator 平均分

xueyizheng
and
claude
committed
785a1b4
fix(evaluation): build ConnectorConf from eval_set_configs in new path

alanyf
and
claude
committed
ff64c57
feat(eval): 执行侧注入 ItemConfig + 按 filter 守卫 evaluator 执行 (MVP)

xueyizheng
and
claude
committed
bf4940c
fix(eval): 调度路径注入 IExptItemRefRepo, 修复 exptStartMultiSet nil 报错

xueyizheng
and
claude
committed
f6ee031

Commits on Jun 15, 2026

Commits on Jun 16, 2026

feat(eval): ExptMangerImpl 注入 IExptItemRefRepo 填 EvalSetDetails (PR4)
xueyizheng
committed
bbe520a
feat(eval): list&get experiment
alanyf
committed
db5b73a
feat(eval): experiment list filter
alanyf
committed
4167db0
feat(eval): experiment list filter
alanyf
committed
cde491c
feat(eval): 评估器运行支持动态参数打通
 alanyf
committed
0524346
refactor(evaluation): WS5 实验结果旧消费链路支持 alias 多实例 + CK item_version_id

NoahKex
and
claude
committed
0086649

Commits on Jun 17, 2026

feat(eval): experiment list filter
alanyf
committed
5d4d0d4
feat(eval): 补齐 OpenAPI item-centric 字段 (单实验 Get 回填 + List source_types 筛选)

alanyf
and
claude
committed
18831b1
feat(eval): experiment create api filter
alanyf
committed
24b2792

Commits on Jun 18, 2026

feat(eval): 尝试修复多评测集运行时，实验结果没有第二个评测集数据和实验结果的问题
 alanyf
committed
f1478cd

Commits on Jun 23, 2026

Commits on Jun 24, 2026

feat(eval): add notification type stubs for gen-commercial compatibility

VinCinx
and
claude
committed
235ff80
[feat][evaluation] add SandboxAgent eval target type

tpfz
and
claude
committed
fa53ca5
[feat][evaluation] add openapi for eval_target async_debug and get_record

tpfz
and
claude
committed
c115035
[feat][evaluation] add CustomFieldSchemas to SandboxAgent entity

tpfz
and
claude
committed
fa0fff5
[feat][evaluation] support async report for SandboxAgent eval target

tpfz
and
claude
committed
6ea6728
[fix][evaluation] align async_debug eval_target openapi route path

tpfz
and
claude
committed
6ce3780
[feat][evaluation] support SandboxAgent in AsyncDebugEvalTargetOApi

tpfz
and
claude
committed
6c79e81
[feat][evaluation] add ISandboxSchedulerAdapter for SandboxAgent target
tpfz
committed
8ab5607
chore(evaluation): align ISandboxSchedulerAdapter with new sandbox_scheduler IDL
tpfz
committed
4397abb
chore(evaluation): drop Image from SandboxRunRequest
tpfz
committed
07c05e4
[feat][evaluation] manage SandboxAgent sandbox lifecycle on submit and per-row completion

tpfz
and
claude
committed
49ad7be
[feat][evaluation] support SandboxAgent in SubmitExperimentOApi

tpfz
and
claude
committed
bd77062
[feat][evaluation] manage SandboxAgent sandbox lifecycle on AsyncDebug
tpfz
committed
40eacfc
[feat][evaluation] bump SandboxAgent debug concurrency to 50
tpfz
committed
394e837

Commits on Jul 10, 2026

[feat][evaluation] destroy SandboxAgent sandbox on item zombie timeout and manual cancel

tpfz
and
claude
committed
acc1861
[feat][evaluation] support runtime param for SandboxAgent eval target

tpfz
and
claude
committed
bcdc999
test(evaluation): 补 SandboxAgent eval target 单测与 mock
tpfz
committed
41f9065
feat(evaluation): 提交实验时 item_concur_num 未传或非正数兜底为 5
tpfz
committed
d45e5d0
feat(evaluation): 透传 log_id 与评测集/item 元数据到评测对象 AsyncExecute

tpfz
and
claude
committed
c56acdf
refactor(evaluation): EvalSetItemMeta 的 id 字段改为 string 避免精度丢失

tpfz
and
claude
committed
19cd4d6
feat(evaluation): SandboxAgent 新增 image 字段透传沙箱镜像

tpfz
and
claude
committed
cfca30a
fix(evaluation): SandboxAgent 提交实验校验放行, 避免误报 invalid target connector

tpfz
and
claude
committed
2dd2fcd
fix(evaluation): 适配 trae_eval 分支已有接口, 修复 cherry-pick 编译

tpfz
and
claude
committed
87b2bb2
fix(evaluation): SandboxAgent 实验 Retry 前重新初始化沙箱任务

tpfz
and
claude
committed
1b7d8be