Fix/audit review 4.0.4_r8

CHANGELOG.md

@@ -1,5 +1,14 @@
 # Changelog
 
+## Desktop 1.4.3 / DeepAgent Core V4.0.4_r8 - Production reliability release
+
+- Persist the complete TaskRun lifecycle, including exact-retry admission, generation, owner/lease fencing, phase changes, terminal compare-and-set, result references, and a leased notification outbox.
+- Split structured subagent work into a normal research phase and a bounded single-turn finalizer with an isolated tool surface, per-turn reasoning policy, strict schema validation, and no empty-success fallback.
+- Preserve distinct recoverable terminal reasons for provider, schema, permission, interruption, timeout, takeover, and doom-loop failures; reject invalid tool input before execution.
+- Detect semantic no-progress from tool results, workspace snapshots, and plan state instead of relying on identical command text.
+- Contain subagent supervision failures with valid sibling controls, real terminal-state rendering, a local ErrorBoundary, persisted-mode validation, same-build quarantine, Electron cold-start coverage, and production source-map verification.
+- Regenerate the JavaScript SDK for the durable task status and delivery contracts.
+
 ## V4.0.4 - Contract-gap closure (engine-decoupled)
 
 - Fix Plan Gate deadlock: stale-plan latch now warns (never hard-blocks) on tool execution, aligned with codex exec-policy philosophy. A mutating tool on a stale plan receives a reminder but always runs.

README.md

@@ -14,7 +14,7 @@
   <a href="https://github.com/deepagent-ltd/deepagent-code-enterprise">Enterprise</a>
 </p>
 
-<p align="center"><sub>Desktop 1.4.2 · DeepAgent Core V4.0.4</sub></p>
+<p align="center"><sub>Desktop 1.4.3 · DeepAgent Core V4.0.4_r8</sub></p>
 
 ---
 
@@ -91,16 +91,18 @@ For high-risk decisions, convene an **Expert Panel**. Correctness, security, per
 
 Project IM brings people and agents into the same thread. Mention an agent to start a scoped run with project context, stream its progress, inspect its artifacts, and keep the answer attached to the conversation that requested it.
 
-## DeepAgent Core V4.0.4
+## DeepAgent Core V4.0.4_r8
 
-V4.0.4 closes production contract gaps while keeping the current turn engine stable:
+Desktop 1.4.3 ships the eighth reliability revision of the V4.0.4 contract. This release hardens the boundaries where long-running and delegated work previously risked duplicate execution, ambiguous completion, or renderer-wide failure:
 
-- **Single durable truth:** DocumentStore uses atomic, recoverable writes for documents, plans, learning candidates, governance state, and version conflicts.
-- **Isolated subagents:** write-capable subagents use dedicated worktrees by default and return their changes to the parent workspace through a bounded, conflict-aware path.
-- **Reliable event delivery:** the Event Bus has a transport seam, durable consumer offsets, offline catch-up, real priority ordering, and observable queue depth.
-- **Governed learning and goals:** knowledge promotion is tied to review evidence and ship-gate snapshots; event-driven goal ticks remain idempotent and respect quiet hours.
-- **Secure integrations:** MCP credentials use environment references or native OS secret storage on macOS, Linux, and Windows; capability and source checks fail closed.
-- **Publishing truth:** installation, CLI examples, release metadata, public domains, and supported-version documentation match the product that is actually shipped.
+- **Durable subagent execution:** TaskRun admission, generation, ownership, leases, settlement, and parent delivery are persisted. Exact retries reconcile to the original run, terminal state uses transactional compare-and-set, and a leased outbox makes completion delivery recoverable without re-running provider work.
+- **Two-stage structured finalization:** research and structured output are separate phases. The finalizer is a bounded single turn with thinking disabled when supported, only the `StructuredOutput` tool visible, no historical task or compaction payload, and no empty-success path.
+- **Fail-closed model and tool boundaries:** provider capability decisions no longer silently relax required tool choice, invalid tool input is rejected before execution, and provider, schema, permission, interruption, timeout, and doom-loop failures keep distinct recoverable terminal reasons.
+- **Semantic no-progress protection:** shell activity uses semantic fingerprints, while the no-progress budget compares bounded results, workspace state, and plan progress so cosmetic command changes cannot evade loop detection.
+- **Contained supervision UI:** subagent controls use valid sibling interactions, display real terminal state, and run behind a local ErrorBoundary with retry, close, persisted-mode validation, same-build quarantine, and recovery after build changes.
+- **Release-grade verification:** the affected Core, runtime, app, and desktop paths are covered by exact-retry, failure-injection, production Chromium, Electron cold-start, and source-map smoke tests.
+
+V4.0.4_r8 retains the durable documents, event delivery, governed learning, isolated worktrees, and secure credential boundaries introduced across the earlier V4.0.4 revisions.
 
 ## Installation
 

README.zh.md

@@ -14,7 +14,7 @@
   <a href="https://github.com/deepagent-ltd/deepagent-code-enterprise">Enterprise 版本</a>
 </p>
 
-<p align="center"><sub>桌面版 1.4.2 · DeepAgent Core V4.0.4</sub></p>
+<p align="center"><sub>桌面版 1.4.3 · DeepAgent Core V4.0.4_r8</sub></p>
 
 ---
 
@@ -91,16 +91,18 @@ DeepAgent 可以把独立工作拆分给数量有界、相互隔离的 Worker。
 
 项目 IM 把团队成员和智能体放进同一条讨论。@ 某个智能体即可启动有明确作用域的运行，使用项目上下文、流式展示进度、关联执行工件，并把答案留在发起任务的对话里。
 
-## DeepAgent Core V4.0.4
+## DeepAgent Core V4.0.4_r8
 
-V4.0.4 在保持当前 turn 引擎稳定的前提下，关闭生产合同缺口：
+桌面版 1.4.3 搭载 V4.0.4 合同的第八次可靠性修订。本次发布重点加固长任务和委派任务的关键边界，避免重复执行、模糊终态或局部界面故障演变为整页崩溃：
 
-- **单一持久真相：** DocumentStore 通过原子、可恢复写入统一管理文档、计划、学习候选、治理状态和版本冲突。
-- **隔离子智能体：** 具备写权限的子智能体默认使用独立 worktree，并通过有界、可感知冲突的路径把改动回传到父工作区。
-- **可靠事件投递：** Event Bus 提供可替换 transport、持久 consumer offset、离线补投、真实优先级排序和可观测队列深度。
-- **受治理的学习与目标：** 知识晋升关联审阅证据与 ship-gate snapshot；事件驱动 goal tick 保持幂等并遵守 quiet hours。
-- **安全集成：** MCP 凭据使用环境变量引用或 macOS、Linux、Windows 原生 secret storage；capability 与来源检查逐层失败关闭。
-- **发布真实性：** 安装方式、CLI 示例、发布元数据、公开域名和支持版本文档与实际交付产品一致。
+- **持久化子智能体执行：** TaskRun 的准入、generation、owner、lease、结算与父会话投递均持久化。exact retry 会归并到原始运行，终态通过事务 CAS 结算，租约式 outbox 可在不重复执行 provider work 的前提下恢复完成通知。
+- **两阶段结构化终结：** 研究和结构化输出分为独立阶段。finalizer 是次数有界的单轮执行；在 provider 支持时关闭 thinking，只暴露 `StructuredOutput`，不混入历史 task 或 compaction 内容，也不存在空结果成功路径。
+- **失败关闭的模型与工具边界：** provider capability 决策不再静默放宽 required tool choice；无效工具输入在执行前被拒绝；provider、schema、permission、interruption、timeout 与 doom-loop 保留不同且可恢复的终态原因。
+- **语义级无进展保护：** shell 使用语义 fingerprint；无进展预算同时比较有界结果、工作区状态与计划进度，不能再靠改写命令描述绕过循环检测。
+- **故障隔离的监督面板：** 子智能体控件使用合法的 sibling interaction，展示真实终态，并由局部 ErrorBoundary、重试/关闭、持久 mode 校验、同 build quarantine 与 build 变化后的恢复机制保护。
+- **发布级验证：** 受影响的 Core、运行时、App 与 Desktop 路径已覆盖 exact retry、故障注入、production Chromium、Electron 冷启动和 source-map smoke。
+
+V4.0.4_r8 同时保留此前 V4.0.4 各修订建立的持久文档、可靠事件投递、知识治理、worktree 隔离与安全凭据边界。
 
 ## 安装
 

design/README.md

@@ -1,6 +1,6 @@
 # DeepAgent Code Architecture & Design
 
-> **Public design overview for DeepAgent Core V4.0.4 / Desktop 1.4.2.** Internal implementation details and roadmap documents live in the private `docs/` tree and are intentionally not version-controlled.
+> **Public design overview for DeepAgent Core V4.0.4_r8 / Desktop 1.4.3.** Internal implementation details and roadmap documents live in the private `docs/` tree and are intentionally not version-controlled.
 
 DeepAgent Code is a document-centered, event-driven AI coding system. It combines a coding-agent runtime with a durable control plane that owns context, planning, learning, collaboration, safety, and human oversight.
 
@@ -22,7 +22,13 @@ Sessions, inputs, plans, documents, goals, events, approvals, and learning decis
 
 A user instruction is durably admitted before execution is scheduled. A successful API response therefore means the instruction is recorded, not merely present in a process-local queue.
 
-DeepAgent is built **on top of** the opencode agent/runtime/session/tool/MCP stack. V4.0.4 strengthens the control plane without replacing the current turn engine, tool system, or provider layer.
+DeepAgent is built **on top of** the opencode agent/runtime/session/tool/MCP stack. V4.0.4_r8 strengthens the control plane without replacing the current turn engine, tool system, or provider layer.
+
+### Durable delegated execution
+
+A delegated task is admitted before provider work begins and is identified independently from the tool-call attempt that submitted it. Exact retries reconcile to the same TaskRun; conflicting reuse fails closed. Generation, execution owner, lease, phase, terminal state, result reference, and parent-notification delivery are durable records. Active mutations and terminal settlement require the expected generation, owner, and live lease, while notification delivery uses an attempt-fenced outbox so crash recovery cannot repeat provider work or acknowledge another claimant's delivery.
+
+Structured work uses two explicit stages. The research stage retains the child Agent's normal tools and transcript. The finalizer is a bounded, single provider turn with a narrow ephemeral tool registry, no task or compaction history, no ordinary steering, and a per-turn reasoning/tool-choice decision derived from provider capability. Completion is valid only after schema-validated structured output is persisted; provider, schema, permission, timeout, interruption, and no-progress failures remain distinct recoverable states.
 
 ### 2. One durable authority per concern
 

packages/app/README.md

@@ -2,6 +2,8 @@
 
 SolidJS front-end shell for the DeepAgent Code desktop app (Electron/Tauri).
 
+Current release: Desktop 1.4.3, powered by DeepAgent Core V4.0.4_r8.
+
 ## Stack
 
 - **UI:** SolidJS + Vite (Bun)