From 64cb9986f65ef36c8523a8e47390b67d83d0bcb0 Mon Sep 17 00:00:00 2001 From: Codex Date: Mon, 13 Jul 2026 04:51:58 +0200 Subject: [PATCH] docs: close artificer default model task --- docs/MDTODO/artificer-runtime-capabilities.md | 2 +- .../R1.2_Task_Report.md | 26 +++++++++++++++++++ 2 files changed, 27 insertions(+), 1 deletion(-) create mode 100644 docs/MDTODO/details/artificer-runtime-capabilities/R1.2_Task_Report.md diff --git a/docs/MDTODO/artificer-runtime-capabilities.md b/docs/MDTODO/artificer-runtime-capabilities.md index ec992dae..470b29d1 100644 --- a/docs/MDTODO/artificer-runtime-capabilities.md +++ b/docs/MDTODO/artificer-runtime-capabilities.md @@ -11,7 +11,7 @@ 在 [UniDesk #1846](https://github.com/pikasTech/unidesk/issues/1846) 将 gpt.pika 的 auth.json/config.toml 唯一 sourceRef 改为 /root/.codex/*.pika；plan/sync 默认视图显示精确 sourceRef 与 source/effective profile，config.toml 同步前验证显式 provider/base URL 且不输出值，activation fence blocked 时不提示 --confirm，缺失或不安全配置 fail closed，完成任务后将详细报告写入[任务报告](./details/artificer-runtime-capabilities/R1.1_Task_Report.md)。 -### R1.2 [in_progress] +### R1.2 [completed] 在 [AgentRun #325](https://github.com/pikasTech/agentrun/issues/325) 让 AipodSpec 默认 backendProfile/credential.profile=gpt-pika、默认模型 gpt-5.6-sol、默认 reasoning effort=medium，并支持显式 model/reasoning override；provider/SecretRef 不随模型切换，带点的 gpt.pika 在 AgentRun admission 中 fail closed，完成任务后将详细报告写入[任务报告](./details/artificer-runtime-capabilities/R1.2_Task_Report.md)。 diff --git a/docs/MDTODO/details/artificer-runtime-capabilities/R1.2_Task_Report.md b/docs/MDTODO/details/artificer-runtime-capabilities/R1.2_Task_Report.md new file mode 100644 index 00000000..74d0247e --- /dev/null +++ b/docs/MDTODO/details/artificer-runtime-capabilities/R1.2_Task_Report.md @@ -0,0 +1,26 @@ +# R1.2 任务报告 + +## 结论 + +Artificer 的唯一运行配置 owner 为 AgentRun `config/aipods/artificer.yaml`。默认模型已改为 `gpt-5.6-sol`，默认 reasoning effort 已改为 `medium`；`backendProfile`、`credential.profile` 继续为 `gpt-pika`，provider SecretRef 与任务级显式 model/reasoning override 边界未改变。 + +## 交付 + +- AgentRun issue：[pikasTech/agentrun#325](https://github.com/pikasTech/agentrun/issues/325) +- AgentRun PR：[pikasTech/agentrun#345](https://github.com/pikasTech/agentrun/pull/345) +- 独立复审：YAML → render → durable task payload/retry → runner `turn/start` 链路通过；26 项定向测试、3 个 `bun --check` 与 `git diff --check` 通过。 +- AgentRun 长期 SPEC 已在 `v0.2` 提交 `887acc2`，改为只引用 owning YAML，不再复制模型、reasoning、profile 或 SecretRef 默认值。 +- UniDesk subagent skill 与 MDTODO 已在 `f2a1ac8b`、`2bba0f20` 更新。 + +## 自动发布与真实验收 + +- 自动 PaC source commit：`9cdf80395994` +- GitOps commit：`f9b8366f27b4` +- 运行 digest：`sha256:fd110eb13ce...` +- 状态：`pac-ready-gitops-exact`，Argo `Synced/Healthy`。 +- 新建未显式传 model/reasoning 的 canary task：`qt_8051b6b0676f408b909992c1815f2a04`。 +- durable task input 显示 `model=gpt-5.6-sol`、`modelSource=aipod-spec-default`、`reasoningEffort=medium`、`reasoningEffortSource=aipod-spec-default`。 +- runner `turn/start:completed` 事件显示 `model=gpt-5.6-sol`、`reasoningEffort=medium`、`reasoningProtocolField=effort`。 +- 权威结果：`ARTIFICER_DEFAULT_SOL_MEDIUM_OK`。 + +全程未手工触发 mirror、PipelineRun、Argo sync/refresh 或部署。