feat: surface codex reasoning summaries
This commit is contained in:
@@ -1639,6 +1639,20 @@ public final class BossUi {
|
||||
}
|
||||
}
|
||||
|
||||
JSONObject reasoningSummary = progress == null ? null : progress.optJSONObject("reasoningSummary");
|
||||
if (reasoningSummary != null) {
|
||||
String summary = reasoningSummary.optString("summary", "").trim();
|
||||
if (!TextUtils.isEmpty(summary)) {
|
||||
String status = reasoningSummary.optString("status", "").trim();
|
||||
card.addView(divider(context));
|
||||
card.addView(sectionTitle(context, "思考摘要"));
|
||||
if (!TextUtils.isEmpty(status)) {
|
||||
card.addView(detailRow(context, "◇", status, "", false));
|
||||
}
|
||||
card.addView(detailRow(context, "", summary, "", false, true));
|
||||
}
|
||||
}
|
||||
|
||||
JSONArray warnings = progress == null ? null : progress.optJSONArray("warnings");
|
||||
if (warnings != null && warnings.length() > 0) {
|
||||
card.addView(divider(context));
|
||||
|
||||
@@ -1285,6 +1285,48 @@ public class ProjectDetailActivityUiTest {
|
||||
assertFalse(viewTreeContainsText(messageView, "id_rsa"));
|
||||
}
|
||||
|
||||
@Test
|
||||
public void executionProgressMessageRendersCodexReasoningSummarySection() throws Exception {
|
||||
Intent intent = new Intent()
|
||||
.putExtra(ProjectDetailActivity.EXTRA_PROJECT_ID, "reasoning-summary")
|
||||
.putExtra(ProjectDetailActivity.EXTRA_PROJECT_NAME, "Boss开发主线程");
|
||||
TestProjectDetailActivity activity = Robolectric
|
||||
.buildActivity(TestProjectDetailActivity.class, intent)
|
||||
.setup()
|
||||
.get();
|
||||
|
||||
JSONObject message = new JSONObject()
|
||||
.put("id", "progress-reasoning-summary-1")
|
||||
.put("sender", "master")
|
||||
.put("senderLabel", "主 Agent")
|
||||
.put("body", "执行进度")
|
||||
.put("kind", "execution_progress")
|
||||
.put("sentAt", "2026-06-01T12:10:00+08:00")
|
||||
.put("executionProgress", new JSONObject()
|
||||
.put("status", "running")
|
||||
.put("steps", new JSONArray()
|
||||
.put(new JSONObject().put("text", "同步 Codex 思考摘要").put("status", "running")))
|
||||
.put("reasoningSummary", new JSONObject()
|
||||
.put("status", "completed")
|
||||
.put("summary", "确认只展示官方 summary,不展示 raw content。")
|
||||
.put("content", new JSONArray()
|
||||
.put(new JSONObject().put("text", "raw hidden chain of thought token=sk-secret-should-not-render")))
|
||||
.put("itemId", "reasoning-secret-should-not-render")));
|
||||
|
||||
View messageView = ReflectionHelpers.callInstanceMethod(
|
||||
activity,
|
||||
"buildMessageView",
|
||||
ReflectionHelpers.ClassParameter.from(JSONObject.class, message)
|
||||
);
|
||||
|
||||
assertTrue(viewTreeContainsText(messageView, "思考摘要"));
|
||||
assertTrue(viewTreeContainsText(messageView, "completed"));
|
||||
assertTrue(viewTreeContainsText(messageView, "确认只展示官方 summary,不展示 raw content。"));
|
||||
assertFalse(viewTreeContainsText(messageView, "raw hidden chain of thought"));
|
||||
assertFalse(viewTreeContainsText(messageView, "sk-secret-should-not-render"));
|
||||
assertFalse(viewTreeContainsText(messageView, "reasoning-secret-should-not-render"));
|
||||
}
|
||||
|
||||
@Test
|
||||
public void nativeRemoteExecutionProgressDoesNotRenderCodexSections() throws Exception {
|
||||
Intent intent = new Intent()
|
||||
|
||||
@@ -153,7 +153,7 @@
|
||||
- Web 和原生 Android 当前都已经接上“新设备导入草稿 -> 勾选 -> 决议预览 -> 应用导入”的前台页面;已绑定生产设备继续保留 heartbeat 自动导入链路
|
||||
- 原生首页的刷新失败策略当前已改成按当前 tab 独立判错,不会再因为 `设备 / 设置 / OTA` 的旁路请求失败把会话页刷新一并判成失败
|
||||
- 当前量产方向已经明确为“Boss 企业控制面 + 可插拔执行协议”:多租户、权限、审批、审计、备份、回退和 Skill 治理由 Boss 承担,Codex App Server / Codex MCP / Codex CLI / Computer Use / 业务系统 API 都作为 provider 接入;详见 `docs/architecture/enterprise_ai_ops_architecture_cn.md`
|
||||
- 当前 Codex App Server 已完成八批接入:boss-agent 默认开启 `local-agent/codex-app-server-runner.mjs` 作为 Codex 绑定入口,优先走 `codex app-server` stdio,也可灰度连接 `ws://127.0.0.1:<port>` 或 `unix://PATH` 同机长驻 App Server;长驻连接支持 `Authorization: Bearer <token>`,配置上优先使用 `codexAppServerAuthTokenFile`。turn 启动前失败才回退 CLI,turn 启动后不重复执行;桌面远程控制默认先走 `codex-computer-use`,失败后回退 `cua-driver-computer-use`。2026-05-31 已按本机 `codex-cli 0.135.0-alpha.1` 生成协议快照 `docs/protocol-snapshots/codex-app-server/0.135.0-alpha.1/`,并把 `turn/plan/updated`、`turn/diff/updated`、`item/started|completed`、`thread/started`、`item/*/requestApproval`、`item/autoApprovalReview/*`、`guardianWarning`、`serverRequest/resolved`、`item/fileChange/patchUpdated`、`thread/status/changed`、`thread/realtime/*`、`model/rerouted`、`thread/tokenUsage/updated`、`mcpServer/startupStatus/updated`、`remoteControl/status/changed`、`thread/goal/*`、`thread/settings/updated`、`thread/compacted`、`account/updated`、`account/rateLimits/updated`、`model/verification`、`warning`、`configWarning`、`deprecationNotice`、`ThreadItem.collabToolCall`、`ThreadItem.contextCompaction`、`mcpToolCall`、`dynamicToolCall`、`webSearch`、`imageView`、`enteredReviewMode`、`exitedReviewMode`、`commandExecution` 归一到 Boss `execution_progress` 卡片;realtime 只保留状态、文本摘要和计数,运行状态只保留模型切换、上下文用量、MCP 状态和远控连接摘要,线程配置只保留目标、模型、审批、沙箱、协作模式和压缩状态,线程协作只保留工具名、状态、目标类型和智能体状态,工具活动只保留类型、名称、状态和安全摘要,账号状态只保留认证方式、套餐、额度窗口、积分余额和模型校验摘要,不保存 SDP、音频原始数据、raw item、remote installationId、cwd、turnId、配置文件路径、collab 源/目标线程 ID、collab prompt、tool arguments/result/contentItems、web URL token、命令正文/输出或未清洗密钥。heartbeat 已能缓存 `model/list / skills/list / plugin/list / app/list / modelProvider/capabilities/read` 的能力摘要;同批已补 `turn/steer` 活跃 turn 干预和 `POST /api/v1/projects/[projectId]/thread-collaboration` 服务端线程协作排队入口。
|
||||
- 当前 Codex App Server 已完成九批接入:boss-agent 默认开启 `local-agent/codex-app-server-runner.mjs` 作为 Codex 绑定入口,优先走 `codex app-server` stdio,也可灰度连接 `ws://127.0.0.1:<port>` 或 `unix://PATH` 同机长驻 App Server;长驻连接支持 `Authorization: Bearer <token>`,配置上优先使用 `codexAppServerAuthTokenFile`。turn 启动前失败才回退 CLI,turn 启动后不重复执行;桌面远程控制默认先走 `codex-computer-use`,失败后回退 `cua-driver-computer-use`。2026-05-31 已按本机 `codex-cli 0.135.0-alpha.1` 生成协议快照 `docs/protocol-snapshots/codex-app-server/0.135.0-alpha.1/`,并把 `turn/plan/updated`、`turn/diff/updated`、`item/started|completed`、`thread/started`、`item/*/requestApproval`、`item/autoApprovalReview/*`、`guardianWarning`、`serverRequest/resolved`、`item/fileChange/patchUpdated`、`thread/status/changed`、`thread/realtime/*`、`model/rerouted`、`thread/tokenUsage/updated`、`mcpServer/startupStatus/updated`、`remoteControl/status/changed`、`thread/goal/*`、`thread/settings/updated`、`thread/compacted`、`account/updated`、`account/rateLimits/updated`、`model/verification`、`warning`、`configWarning`、`deprecationNotice`、`ThreadItem.collabToolCall`、`ThreadItem.contextCompaction`、`mcpToolCall`、`dynamicToolCall`、`webSearch`、`imageView`、`enteredReviewMode`、`exitedReviewMode`、`commandExecution`、`ThreadItem.plan`、`ThreadItem.reasoning.summary` 归一到 Boss `execution_progress` 卡片;realtime 只保留状态、文本摘要和计数,运行状态只保留模型切换、上下文用量、MCP 状态和远控连接摘要,线程配置只保留目标、模型、审批、沙箱、协作模式和压缩状态,线程协作只保留工具名、状态、目标类型和智能体状态,工具活动只保留类型、名称、状态和安全摘要,思考摘要只保留官方 summary 文本和状态,账号状态只保留认证方式、套餐、额度窗口、积分余额和模型校验摘要,不保存 SDP、音频原始数据、raw item、remote installationId、cwd、turnId、配置文件路径、collab 源/目标线程 ID、collab prompt、tool arguments/result/contentItems、web URL token、命令正文/输出、raw reasoning content、reasoning item id 或未清洗密钥。heartbeat 已能缓存 `model/list / skills/list / plugin/list / app/list / modelProvider/capabilities/read` 的能力摘要;同批已补 `turn/steer` 活跃 turn 干预和 `POST /api/v1/projects/[projectId]/thread-collaboration` 服务端线程协作排队入口。
|
||||
- 当前 boss-agent 已支持 Mac OTA:`local-agent/boss-agent-ota-runner.mjs` 默认开启,每 5 分钟检查服务端最新包;状态页可手动检查或下载并安装,安装时保留原绑定配置,只更新版本号和本机 runtime 路径。最新验证版本为 `20260516221619`,已在 MacBook Air `macbook-air` 上确认 OTA 下载校验、暂存、覆盖安装后不会误切到默认 `config.cloud.json`。正式分发脚本已预留 Developer ID 公证路径:`BOSS_AGENT_NOTARIZE=1` 配合 notary profile 或 Apple ID 凭据。
|
||||
- 当前量产治理已补设备撤权和任务可靠性底座:`revoke_device` 会清空设备 token、标记离线并阻断 heartbeat / 任务认领 / Skill 同步 / 日志上报 / boss-agent OTA;`MasterAgentTask` claim 会记录 attempt 和 lease,运行中任务可按租约重试,超过上限转 `timed_out`,用户或管理员可通过 cancel 接口转 `canceled` 且迟到 complete 不覆盖终态。
|
||||
- 当前群聊 `dispatch_execution` 完成回写已补幂等,重复完成不会再向群聊重复追加结果
|
||||
|
||||
@@ -73,7 +73,7 @@
|
||||
- 当前已支持聊天附件主链:输入框左侧 `+` 会打开底部抽屉,支持图片 / 视频 / 文件发送;图片 / 视频先确认,文件直接发送
|
||||
- 当前附件消息支持下载、原生打开、手动分析和自动分析状态展示
|
||||
- 当前线程聊天消息会按该线程绑定的 Codex 电脑显示来源头像:单线程会话使用项目绑定设备头像,多设备 / 群聊消息会优先根据发送人里的设备名匹配对应电脑头像;主 Agent 总入口自身仍保留主 Agent 对话样式
|
||||
- 当前已支持 `execution_progress` 执行进度卡:普通线程对话、主 Agent 托管线程和群聊目标线程执行时,会在对应聊天窗口显示“进度 / 线程状态 / 实时状态 / 线程配置 / 线程协作 / 工具活动 / 账号状态 / 运行状态 / 安全提醒 / 审批状态 / 文件变更 / 分支详情 / 生成结果 / 后台智能体”结构化卡片;线程过程噪音仍走 `thread_process` 折叠
|
||||
- 当前已支持 `execution_progress` 执行进度卡:普通线程对话、主 Agent 托管线程和群聊目标线程执行时,会在对应聊天窗口显示“进度 / 线程状态 / 实时状态 / 线程配置 / 线程协作 / 工具活动 / 思考摘要 / 账号状态 / 运行状态 / 安全提醒 / 审批状态 / 文件变更 / 分支详情 / 生成结果 / 后台智能体”结构化卡片;线程过程噪音仍走 `thread_process` 折叠
|
||||
- `线程详情 / 运维调试` 仍保留对应原生活动页,但已退出主聊天面
|
||||
- 当前已补上本地发送中气泡、发送按钮状态控制,以及“只有接近底部才自动滚到底”的消息流行为
|
||||
- 当前根页导航:
|
||||
@@ -117,7 +117,7 @@
|
||||
- 当前 `RemoteRuntimeAdapter` 还负责拦截固定模式的线程内部环境提示;命中后会直接改写成失败,避免把只读/cwd 这类脏文本写进聊天记录
|
||||
- 当前普通单线程 `conversation_reply` 在真正执行 `codex exec resume` 前,会先把 Boss 用户消息镜像进目标 Codex Desktop rollout;定位优先走 `state_5.sqlite`,不可用时回退扫描 `~/.codex/sessions`,并按 `sourceMessageId` 去重
|
||||
- 当前 Codex Desktop 同步新增常驻刷新桥:`scripts/codex-desktop-refresh-bridge-daemon.mjs` 通过 launchd 监听 `127.0.0.1:4318`,暴露 `POST /api/v1/codex-desktop/refresh`、`GET /api/v1/codex-desktop/events`、`GET /api/v1/codex-desktop/events/recent` 和 `GET /api/v1/codex-desktop/capabilities`;`local-agent` 会优先调用 refresh endpoint,失败时回退到 `scripts/codex-desktop-refresh-hint.mjs` 命令式刷新。SSE 事件只包含线程引用、消息 ID、状态、deep link 等安全元数据,不包含用户正文或内部 prompt;`scripts/codex-desktop-event-consumer.mjs` 可作为 Desktop 插件/IPC 接入前的订阅 smoke;`scripts/codex-desktop-integration-probe.mjs` 负责只读探测 Codex.app 能力
|
||||
- 当前新增 Codex App Server runner:`local-agent/codex-app-server-runner.mjs`。boss-agent 默认配置 `codexAppServerEnabled=true`,会接管 `conversation_reply / dispatch_execution`;它默认通过 stdio 启动 `codex app-server`,也支持 `codexAppServerTransport=ws + codexAppServerUrl=ws://127.0.0.1:<port>` 或 `codexAppServerTransport=unix + codexAppServerUrl=unix:///absolute/path.sock` 连接同机长驻 App Server,bearer token 可通过 `codexAppServerAuthTokenFile` 读取并在握手时发送 `Authorization: Bearer <token>`。runner 执行 `initialize -> thread/resume|thread/start -> turn/start|turn/steer`,并把 `item/agentMessage/delta` 或 `item/completed` 归一成 Boss 任务回复;当 App Server 对单个 JSON-RPC 请求返回 `-32001 / retry later` 时,runner 会做最多 3 次指数退避重试。turn 启动前失败可回退 CLI,turn 启动后失败不回退,避免重复执行。2026-05-31 起,runner 会把 `turn/plan/updated`、`turn/diff/updated`、`item/started|completed`、`thread/started` 归一成 `executionProgress.steps / branch / artifacts / agents`,把 `item/*/requestApproval`、`item/autoApprovalReview/*`、`guardianWarning`、`serverRequest/resolved`、`item/fileChange/patchUpdated` 归一成 `executionProgress.approvals / warnings / fileChanges`,把 `thread/status/changed`、`thread/realtime/started|transcript|outputAudio|itemAdded|error|closed` 归一成 `executionProgress.threadStatus / realtime`,把 `model/rerouted`、`thread/tokenUsage/updated`、`mcpServer/startupStatus/updated`、`remoteControl/status/changed` 归一成 `executionProgress.modelRoute / tokenUsage / mcpServers / remoteControl`,并把 `thread/goal/*`、`thread/settings/updated`、`thread/compacted`、`account/updated`、`account/rateLimits/updated`、`model/verification`、`warning`、`configWarning`、`deprecationNotice`、`ThreadItem.collabToolCall`、`ThreadItem.contextCompaction`、`mcpToolCall`、`dynamicToolCall`、`webSearch`、`imageView`、`enteredReviewMode`、`exitedReviewMode`、`commandExecution` 归一成线程配置、账号状态、模型校验、安全提醒、线程协作、上下文压缩和工具活动摘要;服务端 complete 回写会与本地 Git/GitHub 进度合并,且不保存 SDP、音频 base64、raw realtime item、remote installationId、cwd、turnId、配置路径、collab 源/目标线程 ID、collab prompt、tool arguments/result/contentItems、web URL token、命令正文/输出或未清洗的 MCP 错误。heartbeat 同时支持按 TTL 拉取 `model/list / skills/list / plugin/list / app/list / modelProvider/capabilities/read`,并把摘要保存在 `capabilities.codexAppServer.metadata`。
|
||||
- 当前新增 Codex App Server runner:`local-agent/codex-app-server-runner.mjs`。boss-agent 默认配置 `codexAppServerEnabled=true`,会接管 `conversation_reply / dispatch_execution`;它默认通过 stdio 启动 `codex app-server`,也支持 `codexAppServerTransport=ws + codexAppServerUrl=ws://127.0.0.1:<port>` 或 `codexAppServerTransport=unix + codexAppServerUrl=unix:///absolute/path.sock` 连接同机长驻 App Server,bearer token 可通过 `codexAppServerAuthTokenFile` 读取并在握手时发送 `Authorization: Bearer <token>`。runner 执行 `initialize -> thread/resume|thread/start -> turn/start|turn/steer`,并把 `item/agentMessage/delta` 或 `item/completed` 归一成 Boss 任务回复;当 App Server 对单个 JSON-RPC 请求返回 `-32001 / retry later` 时,runner 会做最多 3 次指数退避重试。turn 启动前失败可回退 CLI,turn 启动后失败不回退,避免重复执行。2026-05-31 起,runner 会把 `turn/plan/updated`、`turn/diff/updated`、`item/started|completed`、`thread/started` 归一成 `executionProgress.steps / branch / artifacts / agents`,把 `item/*/requestApproval`、`item/autoApprovalReview/*`、`guardianWarning`、`serverRequest/resolved`、`item/fileChange/patchUpdated` 归一成 `executionProgress.approvals / warnings / fileChanges`,把 `thread/status/changed`、`thread/realtime/started|transcript|outputAudio|itemAdded|error|closed` 归一成 `executionProgress.threadStatus / realtime`,把 `model/rerouted`、`thread/tokenUsage/updated`、`mcpServer/startupStatus/updated`、`remoteControl/status/changed` 归一成 `executionProgress.modelRoute / tokenUsage / mcpServers / remoteControl`,并把 `thread/goal/*`、`thread/settings/updated`、`thread/compacted`、`account/updated`、`account/rateLimits/updated`、`model/verification`、`warning`、`configWarning`、`deprecationNotice`、`ThreadItem.collabToolCall`、`ThreadItem.contextCompaction`、`mcpToolCall`、`dynamicToolCall`、`webSearch`、`imageView`、`enteredReviewMode`、`exitedReviewMode`、`commandExecution`、`ThreadItem.plan`、`ThreadItem.reasoning.summary` 归一成线程配置、账号状态、模型校验、安全提醒、线程协作、上下文压缩、工具活动、计划步骤和思考摘要;服务端 complete 回写会与本地 Git/GitHub 进度合并,且不保存 SDP、音频 base64、raw realtime item、remote installationId、cwd、turnId、配置路径、collab 源/目标线程 ID、collab prompt、tool arguments/result/contentItems、web URL token、命令正文/输出、raw reasoning content、reasoning item id 或未清洗的 MCP 错误。heartbeat 同时支持按 TTL 拉取 `model/list / skills/list / plugin/list / app/list / modelProvider/capabilities/read`,并把摘要保存在 `capabilities.codexAppServer.metadata`。
|
||||
- 当前 Codex App Server runner 已新增第一版 Boss Inter-Thread Broker:任务携带 `intentCategory=thread_collaboration`、`sourceCodexThreadRef` 和 `targetCodexThreadRef` 时,会先 `thread/read` 源线程,再通过 `thread/inject_items` 向目标线程注入受控摘要,最后 `turn/start` 目标线程;服务端入口是 `POST /api/v1/projects/[projectId]/thread-collaboration`,负责权限、源/目标线程校验和任务排队。这不是假设官方线程 P2P,而是 Boss 自己做线程协作编排。
|
||||
- 当前 boss-agent Mac OTA 已接入:`local-agent/boss-agent-ota-runner.mjs` 会用设备 token 调 Boss 服务端 `/api/v1/boss-agent/ota` 检查最新 Mac 运行包,`/api/v1/boss-agent/ota/apply` 会下载 `boss-agent-mac-latest.zip`、校验 sha256、暂存安装 wrapper,并拉起本机安装器;安装脚本会保留绑定配置并只更新版本号与本机 runtime 路径。安装器会优先沿用当前 LaunchAgent active config,并保留所有 `config*.json`,避免多电脑场景中误绑定到默认设备配置。当前最新验证包为 `20260516221619`;构建脚本支持 `BOSS_AGENT_NOTARIZE=1` 的 Developer ID 公证路径。
|
||||
- 当前 `local-agent` 还新增了两条统一电脑控制 runtime:
|
||||
|
||||
@@ -16,7 +16,13 @@ Codex App Server 是更适合 Boss 长期接入的协议层,因为它面向富
|
||||
- model/list、skills/list、plugin/list、app/list
|
||||
- command execution、file change、tool input、MCP tool-call approvals
|
||||
|
||||
Boss 不能直接把 App Server 原始 Thread / Turn / Item 字段写进业务层。当前第一批已经新增 `local-agent/codex-app-server-runner.mjs`,把 App Server 的 `thread/resume | thread/start -> turn/start -> item/agentMessage/delta -> turn/completed` 映射成 Boss 的普通任务完成回写;2026-05-31 已继续把 `turn/plan/updated`、`turn/diff/updated`、`item/started|completed`、`thread/started` 这类协议事件归一化为 Boss `execution_progress` 的步骤、分支变更、产物和后台智能体。同日第二批补齐 `item/*/requestApproval`、`item/autoApprovalReview/*`、`guardianWarning`、`serverRequest/resolved` 和 `item/fileChange/patchUpdated` 的安全摘要映射,APP 只展示审批状态、风险提醒和文件路径,不展示完整命令、diff、系统提示词或密钥。第三批已把 `thread/status/changed` 与 `thread/realtime/*` 归一成 `executionProgress.threadStatus / realtime`,APP 只展示活跃/等待审批/等待用户输入、realtime 文本摘要、音频片段计数和关闭/错误原因;第四批已把 `model/rerouted`、`thread/tokenUsage/updated`、`mcpServer/startupStatus/updated` 和 `remoteControl/status/changed` 归一成 `executionProgress.modelRoute / tokenUsage / mcpServers / remoteControl`,用于 APP “运行状态”区块。2026-06-01 第五批已把 `thread/goal/updated|cleared`、`thread/settings/updated` 和 `thread/compacted` 归一成 `executionProgress.threadGoal / threadSettings / compaction`,用于 APP “线程配置”区块;第六批已把 `account/updated`、`account/rateLimits/updated`、`model/verification`、`warning`、`configWarning`、`deprecationNotice` 归一成 `executionProgress.accountStatus / modelVerification / warnings`;第七批已把官方 `ThreadItem.collabToolCall` 归一成 `executionProgress.threadCollaboration`,并按官方建议把新版 `ThreadItem.contextCompaction` 映射回 `executionProgress.compaction`;第八批已把 `mcpToolCall`、`dynamicToolCall`、`webSearch`、`imageView`、`enteredReviewMode`、`exitedReviewMode` 和 `commandExecution` 归一成 `executionProgress.toolActivities`。`thread/realtime/sdp`、音频 base64、原始 realtime item、remote installationId、thread settings 的 `cwd`、compaction `turnId`、collaboration settings 内部 prompt、collabToolCall 源/目标线程 ID、tool arguments/result/contentItems、web URL token、命令正文/输出和未清洗的 MCP 错误不入账。
|
||||
Boss 不能直接把 App Server 原始 Thread / Turn / Item 字段写进业务层。当前第一批已经新增 `local-agent/codex-app-server-runner.mjs`,把 App Server 的 `thread/resume | thread/start -> turn/start -> item/agentMessage/delta -> turn/completed` 映射成 Boss 的普通任务完成回写。
|
||||
|
||||
2026-05-31 已继续把 `turn/plan/updated`、`turn/diff/updated`、`item/started|completed`、`thread/started` 这类协议事件归一化为 Boss `execution_progress` 的步骤、分支变更、产物和后台智能体。同日第二批补齐 `item/*/requestApproval`、`item/autoApprovalReview/*`、`guardianWarning`、`serverRequest/resolved` 和 `item/fileChange/patchUpdated` 的安全摘要映射,APP 只展示审批状态、风险提醒和文件路径,不展示完整命令、diff、系统提示词或密钥。第三批已把 `thread/status/changed` 与 `thread/realtime/*` 归一成 `executionProgress.threadStatus / realtime`,APP 只展示活跃/等待审批/等待用户输入、realtime 文本摘要、音频片段计数和关闭/错误原因;第四批已把 `model/rerouted`、`thread/tokenUsage/updated`、`mcpServer/startupStatus/updated` 和 `remoteControl/status/changed` 归一成 `executionProgress.modelRoute / tokenUsage / mcpServers / remoteControl`,用于 APP “运行状态”区块。
|
||||
|
||||
2026-06-01 第五批已把 `thread/goal/updated|cleared`、`thread/settings/updated` 和 `thread/compacted` 归一成 `executionProgress.threadGoal / threadSettings / compaction`,用于 APP “线程配置”区块;第六批已把 `account/updated`、`account/rateLimits/updated`、`model/verification`、`warning`、`configWarning`、`deprecationNotice` 归一成 `executionProgress.accountStatus / modelVerification / warnings`;第七批已把官方 `ThreadItem.collabToolCall` 归一成 `executionProgress.threadCollaboration`,并按官方建议把新版 `ThreadItem.contextCompaction` 映射回 `executionProgress.compaction`;第八批已把 `mcpToolCall`、`dynamicToolCall`、`webSearch`、`imageView`、`enteredReviewMode`、`exitedReviewMode` 和 `commandExecution` 归一成 `executionProgress.toolActivities`;第九批已把官方 `ThreadItem.plan` 的最终 `item/completed` 文本映射为 `executionProgress.steps`,并把 `ThreadItem.reasoning.summary` 映射为 `executionProgress.reasoningSummary`,供 APP 以“思考摘要”轻卡展示。
|
||||
|
||||
`thread/realtime/sdp`、音频 base64、原始 realtime item、remote installationId、thread settings 的 `cwd`、compaction `turnId`、collaboration settings 内部 prompt、collabToolCall 源/目标线程 ID、tool arguments/result/contentItems、web URL token、命令正文/输出、raw reasoning `content`、reasoning item id 和未清洗的 MCP 错误不入账。
|
||||
|
||||
官方文档入口:`https://developers.openai.com/codex/app-server`
|
||||
|
||||
@@ -78,6 +84,7 @@ APP 展示结构对齐截图:
|
||||
- `线程配置`:展示 thread goal、模型 / provider、审批 / 沙箱、协作模式和上下文压缩状态
|
||||
- `线程协作`:展示 `collabToolCall` 的工具名、执行状态、目标类型和智能体状态,不展示源/目标线程 ID 或 prompt
|
||||
- `工具活动`:展示 MCP / dynamic tool / web search / image view / Review / command 的类型、名称、状态和安全摘要,不展示参数、结果、URL token、命令正文或命令输出
|
||||
- `思考摘要`:展示 Codex 官方 reasoning `summary` 和状态,不展示 raw reasoning `content`、item id 或密钥
|
||||
- `运行状态`:展示模型重路由、上下文用量、MCP 启动状态和远控连接状态
|
||||
- `分支详情`:变更行、Git 操作、GitHub CLI 可用状态
|
||||
- `生成结果`:从执行结果里提取文件、图片、APK、文档等产物名
|
||||
@@ -86,7 +93,7 @@ APP 展示结构对齐截图:
|
||||
UI 参考:
|
||||
|
||||
- image2 生成稿:`design/image2/boss-app-codex-app-server-progress-card-20260531.png`
|
||||
- 当前生成稿保持微信效率型:顶部保留项目目标 / 版本记录固定入口,聊天区只展示最终用户消息和结构化进度卡,进度卡分为 `进度 / 线程状态 / 实时状态 / 线程配置 / 线程协作 / 工具活动 / 账号状态 / 运行状态 / 安全提醒 / 审批状态 / 文件变更 / 分支详情 / 生成结果 / 后台智能体`
|
||||
- 当前生成稿保持微信效率型:顶部保留项目目标 / 版本记录固定入口,聊天区只展示最终用户消息和结构化进度卡,进度卡分为 `进度 / 线程状态 / 实时状态 / 线程配置 / 线程协作 / 工具活动 / 思考摘要 / 账号状态 / 运行状态 / 安全提醒 / 审批状态 / 文件变更 / 分支详情 / 生成结果 / 后台智能体`
|
||||
- 后续 Android / Web 继续按该稿收口,不新增无关功能,不把协议字段、系统提示词或执行 envelope 暴露给用户
|
||||
|
||||
## 3. 安全边界
|
||||
@@ -132,6 +139,7 @@ UI 参考:
|
||||
- `local-agent/codex-app-server-runner.mjs` 已把 App Server `account/updated`、`account/rateLimits/updated`、`model/verification`、`warning`、`configWarning`、`deprecationNotice` 归一成 `executionProgress.accountStatus / modelVerification / warnings`;服务端进度路由和 Android 原生进度卡已支持展示,测试覆盖配置路径、turnId 和密钥不外泄
|
||||
- `local-agent/codex-app-server-runner.mjs` 已把 App Server `ThreadItem.collabToolCall` 和 `ThreadItem.contextCompaction` 归一成 `executionProgress.threadCollaboration / compaction`;服务端进度路由和 Android 原生进度卡已支持展示,测试覆盖源/目标线程 ID、内部 prompt、turnId 和密钥不外泄
|
||||
- `local-agent/codex-app-server-runner.mjs` 已把 App Server `mcpToolCall`、`dynamicToolCall`、`webSearch`、`imageView`、`enteredReviewMode`、`exitedReviewMode`、`commandExecution` 归一成 `executionProgress.toolActivities`;服务端进度路由和 Android 原生进度卡已支持展示,测试覆盖 tool arguments/result、URL token、命令正文/输出、本地路径和密钥不外泄
|
||||
- `local-agent/codex-app-server-runner.mjs` 已把 App Server `ThreadItem.plan` 与 `ThreadItem.reasoning.summary` 归一成 `executionProgress.steps / reasoningSummary`;服务端进度路由和 Android 原生进度卡已支持展示,测试覆盖 raw reasoning content、reasoning item id 和密钥不外泄
|
||||
- 新增实时进度入口 `POST /api/v1/master-agent/tasks/[taskId]/progress`,设备端可在任务执行中持续刷新同一张 `execution_progress` 卡;`local-agent` 的 App Server runner 已在收到协议进度事件时调用该接口,complete 仍携带最终进度作为兜底
|
||||
- 新增服务端线程协作入口 `POST /api/v1/projects/[projectId]/thread-collaboration`,由 Boss 校验源/目标项目权限并创建 `intentCategory=thread_collaboration` 的 `conversation_reply` 任务;设备端继续通过 App Server runner 执行 `thread/read -> thread/inject_items -> turn/start`,避免把“线程互通”误做成无监管 P2P
|
||||
- 新增活跃 turn 干预:任务携带 `targetCodexTurnId` / `targetTurnId` 时,App Server runner 会调用 `turn/steer`,并把 `turnControl=steer`、`turnId` 写回执行结果;没有活跃 turn id 时仍使用 `turn/start`
|
||||
|
||||
@@ -1,6 +1,6 @@
|
||||
# Boss 当前运行与部署状态
|
||||
|
||||
更新时间:`2026-05-31`
|
||||
更新时间:`2026-06-01`
|
||||
|
||||
## 1. 本地状态
|
||||
|
||||
@@ -249,7 +249,7 @@ cd /Users/kris/code/boss
|
||||
- 当前 `local-agent` 已新增 `Codex App Server` provider:boss-agent 默认配置 `codexAppServerEnabled=true`,`conversation_reply / dispatch_execution` 会先通过 `codex app-server` 的 stdio JSON-RPC 恢复或创建线程,也可配置 `codexAppServerTransport=ws + codexAppServerUrl=ws://127.0.0.1:<port>` 或 `codexAppServerTransport=unix + codexAppServerUrl=unix:///absolute/path.sock` 连接同机长驻 App Server;长驻连接可通过 `codexAppServerAuthTokenFile` 或 `BOSS_CODEX_APP_SERVER_AUTH_TOKEN_FILE` 提供 bearer token。随后 runner 下发 `turn/start` 并收集流式 agent 回复;如果单个 JSON-RPC 请求返回 `-32001 / retry later`,runner 会先做指数退避重试;如果任务携带 `targetCodexTurnId`,会改用 `turn/steer` 干预活跃 turn;如果 App Server 在 turn 启动前失败,默认允许回退到 `codex exec resume`,如果 turn 已经启动则不再回退,避免同一轮用户消息被重复执行。桌面控制另有 `codexComputerUseEnabled=true`,默认先走 Codex Computer Use,再回退 CUA Driver。
|
||||
- 当前已新增 Boss 自有 Inter-Thread Broker 第一版:服务端入口 `POST /api/v1/projects/[projectId]/thread-collaboration` 会创建带源/目标 Codex 线程引用的协作任务;App Server runner 执行 `thread/read(source) -> thread/inject_items(target) -> turn/start(target)`,用于让一个线程的结论受控进入另一个线程,不依赖官方任意线程 P2P 互聊能力
|
||||
- 当前 `local-agent` 对 `dispatch_execution` 任务会按 `orchestrationBackendId` 分流:默认走 `codex exec resume`;当任务显式选择 `omx-team` 且本机 `omxEnabled + omxCommand/omxArgs` 可用时,会改走 `OMX Team Runtime` JSON 协议执行并回写 `rawThreadReply / replyBody`
|
||||
- 当前 `local-agent` 会在 Codex 任务执行中和完成时回传 `executionProgress`:服务端把同一任务的进度卡从 queued / running 更新到 completed / failed,Android 原生聊天页会显示“进度 / 线程状态 / 实时状态 / 线程配置 / 线程协作 / 工具活动 / 账号状态 / 运行状态 / 安全提醒 / 审批状态 / 文件变更 / 分支详情 / 生成结果 / 后台智能体”。2026-05-31 起,Codex App Server 的 `turn/plan/updated`、`turn/diff/updated`、`item/started|completed`、`thread/started` 会直接映射为进度步骤、变更统计、生成产物和后台智能体;第二批已把 `item/*/requestApproval`、`item/autoApprovalReview/*`、`guardianWarning`、`serverRequest/resolved`、`item/fileChange/patchUpdated` 映射为审批、安全提醒和文件变更摘要;第三批已把 `thread/status/changed` 与 `thread/realtime/*` 安全映射为线程状态和实时状态摘要;第四批已把 `model/rerouted`、`thread/tokenUsage/updated`、`mcpServer/startupStatus/updated`、`remoteControl/status/changed` 安全映射为运行状态摘要;第五批已把 `thread/goal/*`、`thread/settings/updated` 和 `thread/compacted` 映射为线程配置摘要;第六批已把 `account/updated`、`account/rateLimits/updated`、`model/verification`、`warning`、`configWarning`、`deprecationNotice` 映射为账号状态、模型校验和安全提醒摘要;第七批已把 `ThreadItem.collabToolCall` 和 `ThreadItem.contextCompaction` 映射为线程协作和上下文压缩摘要;第八批已把 `mcpToolCall`、`dynamicToolCall`、`webSearch`、`imageView`、`enteredReviewMode`、`exitedReviewMode`、`commandExecution` 映射为工具活动摘要。所有进度均通过 `POST /api/v1/master-agent/tasks/[taskId]/progress` 实时刷新;字段白名单会剥离 cwd、turnId、配置文件路径、内部 prompt、collab 源/目标线程 ID、tool arguments/result、web URL token、命令正文/输出和未清洗密钥,complete 回写仍会携带最终进度兜底
|
||||
- 当前 `local-agent` 会在 Codex 任务执行中和完成时回传 `executionProgress`:服务端把同一任务的进度卡从 queued / running 更新到 completed / failed,Android 原生聊天页会显示“进度 / 线程状态 / 实时状态 / 线程配置 / 线程协作 / 工具活动 / 思考摘要 / 账号状态 / 运行状态 / 安全提醒 / 审批状态 / 文件变更 / 分支详情 / 生成结果 / 后台智能体”。2026-05-31 起,Codex App Server 的 `turn/plan/updated`、`turn/diff/updated`、`item/started|completed`、`thread/started` 会直接映射为进度步骤、变更统计、生成产物和后台智能体;第二批已把 `item/*/requestApproval`、`item/autoApprovalReview/*`、`guardianWarning`、`serverRequest/resolved`、`item/fileChange/patchUpdated` 映射为审批、安全提醒和文件变更摘要;第三批已把 `thread/status/changed` 与 `thread/realtime/*` 安全映射为线程状态和实时状态摘要;第四批已把 `model/rerouted`、`thread/tokenUsage/updated`、`mcpServer/startupStatus/updated`、`remoteControl/status/changed` 安全映射为运行状态摘要;第五批已把 `thread/goal/*`、`thread/settings/updated` 和 `thread/compacted` 映射为线程配置摘要;第六批已把 `account/updated`、`account/rateLimits/updated`、`model/verification`、`warning`、`configWarning`、`deprecationNotice` 映射为账号状态、模型校验和安全提醒摘要;第七批已把 `ThreadItem.collabToolCall` 和 `ThreadItem.contextCompaction` 映射为线程协作和上下文压缩摘要;第八批已把 `mcpToolCall`、`dynamicToolCall`、`webSearch`、`imageView`、`enteredReviewMode`、`exitedReviewMode`、`commandExecution` 映射为工具活动摘要;第九批已把 `ThreadItem.plan` 和 `ThreadItem.reasoning.summary` 映射为计划步骤与思考摘要。所有进度均通过 `POST /api/v1/master-agent/tasks/[taskId]/progress` 实时刷新;字段白名单会剥离 cwd、turnId、配置文件路径、内部 prompt、collab 源/目标线程 ID、tool arguments/result、web URL token、命令正文/输出、raw reasoning content、reasoning item id 和未清洗密钥,complete 回写仍会携带最终进度兜底
|
||||
- 当前 `local-agent` heartbeat 已新增 Codex App Server capability discovery:按 TTL 拉取模型、provider 能力、Skill、Plugin、App 摘要,写入 `capabilities.codexAppServer.metadata`;Web 设备详情会展示 App Server 连接状态、模型数量、默认/快速/深度模型和扩展数量
|
||||
- 当前 `MasterAgentTask` 已具备服务端租约和取消基础状态机:claim 会写入 `attemptCount / maxAttempts / leaseExpiresAt`,运行中任务租约过期后可被重新认领,超过重试上限会转 `timed_out`;`POST /api/v1/master-agent/tasks/[taskId]/cancel` 会把任务转 `canceled`,迟到的成功 complete 不会覆盖终态
|
||||
- 当前 `local-agent` 对 `browser_control / desktop_control` 已从占位骨架升级成外部 runtime 桥:当本机配置了 `browserControlEnabled + browserControlCommand` 或 `computerUseEnabled + computerUseCommand` 时,会把标准化 JSON 请求透传给外部进程,并解析单行 JSON 结果;未启用时会 fail closed,返回明确的 runtime disabled 错误,不再假装执行成功
|
||||
|
||||
@@ -698,6 +698,65 @@ function extractToolActivitySnapshot(item, lifecycleStatus = "running") {
|
||||
return null;
|
||||
}
|
||||
|
||||
function cleanPlanStepText(text) {
|
||||
return safeProgressText(text, 220)
|
||||
.replace(/^\s*(?:[-*•]|\d+[.)]|[☐✓✔]|\[[ xX✓]\])\s*/u, "")
|
||||
.trim();
|
||||
}
|
||||
|
||||
function extractPlanStepsFromItem(item) {
|
||||
if (!item || typeof item !== "object" || item.type !== "plan") {
|
||||
return [];
|
||||
}
|
||||
const rawSteps = asArray(item.steps).length > 0
|
||||
? asArray(item.steps).map((step) => extractProgressText(step))
|
||||
: String(item.text ?? "")
|
||||
.split(/\r?\n/)
|
||||
.map((line) => line.trim());
|
||||
return rawSteps
|
||||
.map(cleanPlanStepText)
|
||||
.filter(Boolean)
|
||||
.slice(0, 10)
|
||||
.map((text, index) => ({
|
||||
id: `plan-item-${index + 1}`,
|
||||
text,
|
||||
status: "pending",
|
||||
}));
|
||||
}
|
||||
|
||||
function extractReasoningSummaryText(summary) {
|
||||
const values = asArray(summary).length > 0 ? asArray(summary) : [summary];
|
||||
return safeProgressText(
|
||||
values
|
||||
.map((item) => {
|
||||
if (typeof item === "string") {
|
||||
return item;
|
||||
}
|
||||
if (item && typeof item === "object") {
|
||||
return item.text ?? item.summary ?? "";
|
||||
}
|
||||
return "";
|
||||
})
|
||||
.filter(Boolean)
|
||||
.join(" / "),
|
||||
280,
|
||||
);
|
||||
}
|
||||
|
||||
function extractReasoningSummarySnapshot(item, lifecycleStatus = "running") {
|
||||
if (!item || typeof item !== "object" || item.type !== "reasoning") {
|
||||
return null;
|
||||
}
|
||||
const summary = extractReasoningSummaryText(item.summary);
|
||||
if (!summary) {
|
||||
return null;
|
||||
}
|
||||
return {
|
||||
status: safeProgressText(item.status, 40) || lifecycleStatus,
|
||||
summary,
|
||||
};
|
||||
}
|
||||
|
||||
function extractSubAgentSource(value) {
|
||||
if (!value || typeof value !== "object") {
|
||||
return undefined;
|
||||
@@ -1249,6 +1308,7 @@ function createProgressCollector() {
|
||||
let compaction;
|
||||
let threadCollaboration;
|
||||
const toolActivities = [];
|
||||
let reasoningSummary;
|
||||
let accountStatus;
|
||||
let modelVerification;
|
||||
|
||||
@@ -1390,6 +1450,10 @@ function createProgressCollector() {
|
||||
for (const artifact of extractArtifactsFromItem(message.params?.item, artifacts.length)) {
|
||||
upsertArtifact(artifact);
|
||||
}
|
||||
const itemPlanSteps = extractPlanStepsFromItem(item);
|
||||
if (itemPlanSteps.length > 0) {
|
||||
steps.splice(0, steps.length, ...itemPlanSteps);
|
||||
}
|
||||
const nextCollaboration = extractThreadCollaborationSnapshot(item);
|
||||
if (nextCollaboration) {
|
||||
threadCollaboration = nextCollaboration;
|
||||
@@ -1397,6 +1461,13 @@ function createProgressCollector() {
|
||||
upsertToolActivity(
|
||||
extractToolActivitySnapshot(item, message.method === "item/completed" ? "completed" : "running"),
|
||||
);
|
||||
const nextReasoningSummary = extractReasoningSummarySnapshot(
|
||||
item,
|
||||
message.method === "item/completed" ? "completed" : "running",
|
||||
);
|
||||
if (nextReasoningSummary) {
|
||||
reasoningSummary = nextReasoningSummary;
|
||||
}
|
||||
if (item?.type === "contextCompaction") {
|
||||
compaction = {
|
||||
status: "completed",
|
||||
@@ -1703,6 +1774,9 @@ function createProgressCollector() {
|
||||
if (toolActivities.length > 0) {
|
||||
result.toolActivities = toolActivities.map((item) => ({ ...item }));
|
||||
}
|
||||
if (reasoningSummary) {
|
||||
result.reasoningSummary = { ...reasoningSummary };
|
||||
}
|
||||
if (accountStatus) {
|
||||
result.accountStatus = { ...accountStatus };
|
||||
}
|
||||
|
||||
@@ -246,6 +246,11 @@ export interface ExecutionProgressToolActivity {
|
||||
detail?: string;
|
||||
}
|
||||
|
||||
export interface ExecutionProgressReasoningSummary {
|
||||
status: string;
|
||||
summary: string;
|
||||
}
|
||||
|
||||
export interface ExecutionProgressSnapshot {
|
||||
taskId: string;
|
||||
projectId: string;
|
||||
@@ -277,6 +282,7 @@ export interface ExecutionProgressSnapshot {
|
||||
compaction?: ExecutionProgressCompaction;
|
||||
threadCollaboration?: ExecutionProgressThreadCollaboration;
|
||||
toolActivities?: ExecutionProgressToolActivity[];
|
||||
reasoningSummary?: ExecutionProgressReasoningSummary;
|
||||
updatedAt: string;
|
||||
}
|
||||
|
||||
@@ -316,6 +322,10 @@ export interface ExecutionProgressInput {
|
||||
cwd?: unknown;
|
||||
}
|
||||
>;
|
||||
reasoningSummary?: Partial<ExecutionProgressReasoningSummary> & {
|
||||
content?: unknown;
|
||||
itemId?: unknown;
|
||||
};
|
||||
}
|
||||
|
||||
export interface ForwardSource {
|
||||
@@ -4004,6 +4014,9 @@ function normalizeExecutionProgressSnapshot(raw: Partial<ExecutionProgressSnapsh
|
||||
? undefined
|
||||
: normalizeExecutionProgressThreadCollaboration(raw.threadCollaboration),
|
||||
toolActivities: nativeRemoteControl ? undefined : normalizeExecutionProgressToolActivities(raw.toolActivities),
|
||||
reasoningSummary: nativeRemoteControl
|
||||
? undefined
|
||||
: normalizeExecutionProgressReasoningSummary(raw.reasoningSummary),
|
||||
updatedAt: raw.updatedAt ?? nowIso(),
|
||||
};
|
||||
}
|
||||
@@ -5803,6 +5816,22 @@ function normalizeExecutionProgressToolActivities(
|
||||
return activities.length > 0 ? activities : undefined;
|
||||
}
|
||||
|
||||
function normalizeExecutionProgressReasoningSummary(
|
||||
input?: ExecutionProgressInput["reasoningSummary"],
|
||||
): ExecutionProgressReasoningSummary | undefined {
|
||||
if (!input) {
|
||||
return undefined;
|
||||
}
|
||||
const summary = safeExecutionProgressText(input.summary);
|
||||
if (!summary) {
|
||||
return undefined;
|
||||
}
|
||||
return {
|
||||
status: safeExecutionProgressText(input.status) || "running",
|
||||
summary,
|
||||
};
|
||||
}
|
||||
|
||||
function defaultExecutionProgressStepTexts(task: Pick<MasterAgentTask, "taskType" | "relayViaMasterAgent">) {
|
||||
if (task.taskType === "browser_control") {
|
||||
return [
|
||||
@@ -5955,6 +5984,9 @@ function buildExecutionProgressSnapshot(
|
||||
? undefined
|
||||
: normalizeExecutionProgressThreadCollaboration(input?.threadCollaboration),
|
||||
toolActivities: nativeRemoteControl ? undefined : normalizeExecutionProgressToolActivities(input?.toolActivities),
|
||||
reasoningSummary: nativeRemoteControl
|
||||
? undefined
|
||||
: normalizeExecutionProgressReasoningSummary(input?.reasoningSummary),
|
||||
updatedAt: nowIso(),
|
||||
};
|
||||
}
|
||||
|
||||
52
tests/fixtures/codex-app-server-runtime.mjs
vendored
52
tests/fixtures/codex-app-server-runtime.mjs
vendored
@@ -818,6 +818,58 @@ rl.on("line", (line) => {
|
||||
},
|
||||
});
|
||||
}
|
||||
if (process.env.BOSS_CODEX_APP_SERVER_FIXTURE_EMIT_REASONING_PLAN_EVENTS === "1") {
|
||||
send({
|
||||
method: "item/completed",
|
||||
params: {
|
||||
threadId: message.params?.threadId,
|
||||
turnId: "turn-fixture",
|
||||
item: {
|
||||
type: "plan",
|
||||
id: "plan-secret-should-not-leak",
|
||||
text: "1. 回读官方 App Server item 协议\n2. 补充 reasoning summary 安全映射",
|
||||
},
|
||||
},
|
||||
});
|
||||
send({
|
||||
method: "item/started",
|
||||
params: {
|
||||
threadId: message.params?.threadId,
|
||||
turnId: "turn-fixture",
|
||||
item: {
|
||||
type: "reasoning",
|
||||
id: "reasoning-secret-should-not-leak",
|
||||
summary: [
|
||||
{
|
||||
text: "正在判断哪些思考摘要可以展示。",
|
||||
},
|
||||
],
|
||||
content: [
|
||||
{
|
||||
text: "raw hidden reasoning token=sk-secret-should-not-leak",
|
||||
},
|
||||
],
|
||||
},
|
||||
},
|
||||
});
|
||||
send({
|
||||
method: "item/completed",
|
||||
params: {
|
||||
threadId: message.params?.threadId,
|
||||
turnId: "turn-fixture",
|
||||
item: {
|
||||
type: "reasoning",
|
||||
id: "reasoning-secret-should-not-leak",
|
||||
summary: ["确认只展示官方 summary,不展示 raw content。"],
|
||||
content: [
|
||||
{
|
||||
text: "raw hidden chain of thought token=sk-secret-should-not-leak",
|
||||
},
|
||||
],
|
||||
},
|
||||
},
|
||||
});
|
||||
}
|
||||
send({
|
||||
method: "item/agentMessage/delta",
|
||||
params: {
|
||||
|
||||
@@ -718,6 +718,58 @@ test("codex app-server runner maps tool, search, image, review, and command acti
|
||||
}
|
||||
});
|
||||
|
||||
test("codex app-server runner maps item plan and reasoning summary without leaking raw reasoning", async () => {
|
||||
const previous = process.env.BOSS_CODEX_APP_SERVER_FIXTURE_EMIT_REASONING_PLAN_EVENTS;
|
||||
process.env.BOSS_CODEX_APP_SERVER_FIXTURE_EMIT_REASONING_PLAN_EVENTS = "1";
|
||||
try {
|
||||
const runnerConfig = getCodexAppServerRunnerConfig(process.env, {
|
||||
codexAppServerEnabled: true,
|
||||
codexAppServerCommand: process.execPath,
|
||||
codexAppServerArgs: ["tests/fixtures/codex-app-server-runtime.mjs"],
|
||||
codexAppServerWorkdir: repoRoot,
|
||||
codexAppServerTimeoutMs: 5000,
|
||||
masterAgentModel: "gpt-5.4",
|
||||
});
|
||||
|
||||
const result = await executeCodexAppServerTask(runnerConfig, {
|
||||
taskId: "task-app-server-reasoning-plan-events",
|
||||
taskType: "conversation_reply",
|
||||
targetCodexThreadRef: "019d-app-server-thread",
|
||||
targetCodexFolderRef: repoRoot,
|
||||
executionPrompt: "检查 Codex plan 和 reasoning item",
|
||||
});
|
||||
|
||||
assert.equal(result.status, "completed");
|
||||
assert.deepEqual(result.executionProgress.steps, [
|
||||
{
|
||||
id: "plan-item-1",
|
||||
text: "回读官方 App Server item 协议",
|
||||
status: "pending",
|
||||
},
|
||||
{
|
||||
id: "plan-item-2",
|
||||
text: "补充 reasoning summary 安全映射",
|
||||
status: "pending",
|
||||
},
|
||||
]);
|
||||
assert.deepEqual(result.executionProgress.reasoningSummary, {
|
||||
status: "completed",
|
||||
summary: "确认只展示官方 summary,不展示 raw content。",
|
||||
});
|
||||
const serialized = JSON.stringify(result.executionProgress);
|
||||
assert.equal(serialized.includes("raw hidden reasoning"), false);
|
||||
assert.equal(serialized.includes("raw hidden chain of thought"), false);
|
||||
assert.equal(serialized.includes("sk-secret-should-not-leak"), false);
|
||||
assert.equal(serialized.includes("reasoning-secret-should-not-leak"), false);
|
||||
} finally {
|
||||
if (previous === undefined) {
|
||||
delete process.env.BOSS_CODEX_APP_SERVER_FIXTURE_EMIT_REASONING_PLAN_EVENTS;
|
||||
} else {
|
||||
process.env.BOSS_CODEX_APP_SERVER_FIXTURE_EMIT_REASONING_PLAN_EVENTS = previous;
|
||||
}
|
||||
}
|
||||
});
|
||||
|
||||
test("codex app-server runner bridges source thread context into target thread through inject_items", async () => {
|
||||
const previous = process.env.BOSS_CODEX_APP_SERVER_FIXTURE_INTER_THREAD;
|
||||
process.env.BOSS_CODEX_APP_SERVER_FIXTURE_INTER_THREAD = "1";
|
||||
|
||||
@@ -611,3 +611,60 @@ test("POST task progress preserves Codex tool activity summaries without leaking
|
||||
assert.equal(serialized.includes("/Users/kris"), false);
|
||||
assert.equal(serialized.includes("id_rsa"), false);
|
||||
});
|
||||
|
||||
test("POST task progress preserves Codex reasoning summaries without leaking raw content", async () => {
|
||||
const task = await data.queueMasterAgentTask({
|
||||
taskId: "route-progress-reasoning-summary-task",
|
||||
projectId: "group-progress-test",
|
||||
taskType: "dispatch_execution",
|
||||
requestMessageId: "msg-route-progress-reasoning-summary",
|
||||
requestText: "检查 Codex 思考摘要",
|
||||
executionPrompt: "检查 Codex 思考摘要",
|
||||
requestedBy: "krisolo",
|
||||
requestedByAccount: "krisolo",
|
||||
deviceId: "mac-studio",
|
||||
targetProjectId: "master-agent",
|
||||
targetThreadId: "master-agent-thread",
|
||||
});
|
||||
await data.claimNextMasterAgentTask("mac-studio");
|
||||
|
||||
const response = await postProgress(
|
||||
new NextRequest(`http://127.0.0.1:3000/api/v1/master-agent/tasks/${task.taskId}/progress`, {
|
||||
method: "POST",
|
||||
headers: {
|
||||
"content-type": "application/json",
|
||||
"x-boss-device-token": "boss-mac-studio-token",
|
||||
},
|
||||
body: JSON.stringify({
|
||||
deviceId: "mac-studio",
|
||||
status: "running",
|
||||
executionProgress: {
|
||||
steps: [{ text: "同步 Codex 思考摘要", status: "running" }],
|
||||
reasoningSummary: {
|
||||
status: "completed",
|
||||
summary: "确认只展示官方 summary,不展示 raw content token=sk-secret-should-not-persist。",
|
||||
content: [{ text: "raw hidden chain of thought token=sk-secret-should-not-persist" }],
|
||||
itemId: "reasoning-secret-should-not-persist",
|
||||
},
|
||||
},
|
||||
}),
|
||||
}),
|
||||
{ params: Promise.resolve({ taskId: task.taskId }) },
|
||||
);
|
||||
|
||||
assert.equal(response.status, 200);
|
||||
|
||||
const state = await data.readState();
|
||||
const progress = state.projects
|
||||
.find((project) => project.id === "master-agent")
|
||||
?.messages.find((message) => message.executionProgress?.taskId === task.taskId)
|
||||
?.executionProgress;
|
||||
assert.deepEqual(progress?.reasoningSummary, {
|
||||
status: "completed",
|
||||
summary: "确认只展示官方 summary,不展示 raw content token=[redacted]",
|
||||
});
|
||||
const serialized = JSON.stringify(progress);
|
||||
assert.equal(serialized.includes("raw hidden chain of thought"), false);
|
||||
assert.equal(serialized.includes("sk-secret-should-not-persist"), false);
|
||||
assert.equal(serialized.includes("reasoning-secret-should-not-persist"), false);
|
||||
});
|
||||
|
||||
Reference in New Issue
Block a user