ai-eval-scope 0.3.2__tar.gz → 0.3.3__tar.gz

{ai_eval_scope-0.3.2 → ai_eval_scope-0.3.3}/CHANGELOG.md

@@ -1,3 +1,9 @@
+## v0.3.3 (2026-09-15)
+
+### Fix
+
+- **evaluator**: 观测推送队列目录随 run workspace——容器只读 CWD 吞掉整个上传
+
 ## v0.3.2 (2026-09-15)
 
 ### Feat

{ai_eval_scope-0.3.2 → ai_eval_scope-0.3.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: ai-eval-scope
-Version: 0.3.2
+Version: 0.3.3
 Summary: Agent 能力评估系统 — 基于 Agent-Driven 架构的评测框架
 Author: Agent Eval Team
 License-Expression: MIT

{ai_eval_scope-0.3.2 → ai_eval_scope-0.3.3}/agent_eval/__init__.py

@@ -1,3 +1,3 @@
 """Agent 能力评估系统"""
 
-__version__ = "0.3.2"
+__version__ = "0.3.3"

{ai_eval_scope-0.3.2 → ai_eval_scope-0.3.3}/agent_eval/cli/_common.py

@@ -201,15 +201,18 @@ def _flush_observability(
     """
     from agent_eval.observability import ResultSink, load_config
 
-    cfg = load_config(upload_override=upload_override)
-    if not cfg.enabled:
-        return
-
-    rprint("[blue]可观测平台:[/blue] 推送结果中…")
+    # 队列目录必须随 run workspace：config 的回退是 CWD 相对路径（.ingest_queue），
+    # 而容器里 CWD 常是只读的场景包挂载（:ro）——创建即炸会吞掉整个推送，
+    # 且表现为「构建绿但平台无数据」（构建 #10-#18 实录）。先取 workspace 再建配置。
     run_workspace = None
     rw = getattr(result, "run_workspace", None)
     if rw is not None:
         run_workspace = getattr(rw, "root", None) or (rw.path if hasattr(rw, "path") else None)
+    cfg = load_config(workspace=run_workspace, upload_override=upload_override)
+    if not cfg.enabled:
+        return
+
+    rprint("[blue]可观测平台:[/blue] 推送结果中…")
     # 查看页地址确定性拼装（/run/:id，后端按 externalRunId 兼容查询）——Jenkins
     # console 正则提取进构建描述；URL 不依赖推送成败，离线重放成功后同样有效
     run_id = str(getattr(result, "run_id", "") or "")

{ai_eval_scope-0.3.2 → ai_eval_scope-0.3.3}/pyproject.toml

@@ -6,7 +6,7 @@ build-backend = "hatchling.build"
 # distribution 名 ai-eval-scope（agent-eval 在 PyPI 已被占用，arch/17 D-PKG-1）；
 # import 名 agent_eval 与 CLI 命令名 agent-eval 不变
 name = "ai-eval-scope"
-version = "0.3.2"
+version = "0.3.3"
 description = "Agent 能力评估系统 — 基于 Agent-Driven 架构的评测框架"
 readme = "README.md"
 requires-python = ">=3.11"

{ai_eval_scope-0.3.2 → ai_eval_scope-0.3.3}/report/mypy-results.xml

@@ -1,5 +1,5 @@
 <?xml version="1.0" encoding="utf-8"?>
-<testsuite errors="0" failures="0" name="mypy" skips="0" tests="1" time="10.498">
-  <testcase classname="mypy" file="mypy" line="1" name="mypy-py3_11-linux" time="10.498">
+<testsuite errors="0" failures="0" name="mypy" skips="0" tests="1" time="10.274">
+  <testcase classname="mypy" file="mypy" line="1" name="mypy-py3_11-linux" time="10.274">
   </testcase>
 </testsuite>