AgentEra · Maplemx · Jul 25, 2026 · Jul 23, 2026 · Jul 23, 2026 · Jul 25, 2026
diff --git a/README.md b/README.md
@@ -1,6 +1,6 @@
 <img width="640" alt="Agently" src="https://github.com/user-attachments/assets/c645d031-c8b0-4dba-a515-9d7a4b0a6881" />
 
-# Agently 4.1.4.3 - AI Application Runtime Framework
+# Agently 4.1.4.4 - AI Application Runtime Framework
 
 > Build AI service backends with structured outputs, observable Actions, runtime Skills, MCP capabilities, process streams, and recoverable workflows.
 
@@ -32,14 +32,14 @@ Agently is for teams moving from "the model can do it once" to "the application
 
 The main design question is simple: how do you keep model behavior useful while still giving application code stable contracts, observable execution, and restart-safe workflow boundaries?
 
-Agently 4.1.4.3 is a structured-output compatibility patch over the 4.1.4.2
-owner-boundary release. Direct Pydantic v2 `BaseModel` classes now work as
-first-class `.output(...)` contracts for `ModelRequest` and `AgentExecution`,
-including nested models, while the 4.1.4.2 runtime ownership boundaries remain
-unchanged. Read the
-[4.1.4.3 Release Notes](docs/en/development/release-notes-4.1.4.3.md),
-[4.1.4 Release Notes](docs/en/development/release-notes-4.1.4.md), and
-[4.1.3.9 Release Notes](docs/en/development/release-notes-4.1.3.9.md) for the
+Agently 4.1.4.4 strengthens Pydantic field-constraint prompting and correction
+retries, adds recovery-aware TriggerFlow snapshot projection and bounded local
+snapshot retention, and moves default release validation away from local
+Ollama dependencies. The 4.1.4.2 runtime ownership boundaries and companion
+protocols remain unchanged. Read the
+[4.1.4.4 Release Notes](docs/en/development/release-notes-4.1.4.4.md),
+[4.1.4.3 Release Notes](docs/en/development/release-notes-4.1.4.3.md), and
+[4.1.4 Release Notes](docs/en/development/release-notes-4.1.4.md) for the
 current release and upgrade path.
 
 ## Why Agently
@@ -48,7 +48,7 @@ Many AI frameworks are strong at exploration or at assembling broad integration
 
 Agently is a good fit when you care about:
 
-- **AI services should be runtime executions, not prompt glue** - `AgentExecution` owns one run's prompt, strategy, Actions, Skill bindings, process stream, TaskContext evidence, and result views across direct, flat, and TaskBoard strategies. Read [4.1.4.3 Release Notes](docs/en/development/release-notes-4.1.4.3.md), [4.1.4 Release Notes](docs/en/development/release-notes-4.1.4.md), and [Agent Auto Orchestration examples](examples/agent_auto_orchestration/).
+- **AI services should be runtime executions, not prompt glue** - `AgentExecution` owns one run's prompt, strategy, Actions, Skill bindings, process stream, TaskContext evidence, and result views across direct, flat, and TaskBoard strategies. Read [4.1.4.4 Release Notes](docs/en/development/release-notes-4.1.4.4.md), [4.1.4 Release Notes](docs/en/development/release-notes-4.1.4.md), and [Agent Auto Orchestration examples](examples/agent_auto_orchestration/).
 - **Model switching should not rewrite business logic** - Agently normalizes provider setup, prompt slots, response parsing, action execution, and response reading into one request/runtime contract. Read [Model Setup](docs/en/start/model-setup.md), [Models Overview](docs/en/models/overview.md), and [Requests Overview](docs/en/requests/overview.md).
 - **Structured output should be a framework guarantee, not only a provider feature** - `.output(...)` schemas, required field extraction, parser feedback, retries, `ensure_keys`, `ensure_all_keys`, and validation handlers work together inside Agently. Read [Schema as Prompt](docs/en/requests/schema-as-prompt.md), [Output Control](docs/en/requests/output-control.md), and examples in [`examples/basic/`](examples/basic/).
 - **Streaming should expose structure before the final token** - `instant` mode lets consumers react to structured fields while the model is still streaming, which is useful for UI updates, SSE routes, and workflow signals. Read [Model Response](docs/en/requests/model-response.md), [FastAPI Service Exposure](docs/en/services/fastapi.md), and [`examples/fastapi/`](examples/fastapi/).
@@ -60,7 +60,7 @@ Agently is a good fit when you care about:
 - **Common model-app patterns should be composable** - router, To-Do/dependency execution, planning, reflection, evaluator/reviser, and multi-agent collaboration can be built from the same request/action/signal primitives. Read [Playbooks](docs/en/playbooks/overview.md), [TriggerFlow Model Integration](docs/en/triggerflow/model-integration.md), and [`examples/step_by_step/`](examples/step_by_step/).
 - **Services should keep clean project boundaries** - async APIs, FastAPI helpers, settings files, prompt files, DevTools observation, and companion coding-agent skills fit non-trivial projects. Read [Project Framework](docs/en/start/project-framework.md), [FastAPI Service Exposure](docs/en/services/fastapi.md), and [Observability](docs/en/observability/overview.md).
 
-Current framework version: `4.1.4.3`.
+Current framework version: `4.1.4.4`.
 
 Python: `>=3.10`.
 
@@ -208,7 +208,7 @@ Prompts are composed from named slots. That keeps application intent, constraint
 result = (
     agent
     .role("You are a concise release-note writer.")
-    .info({"version": "4.1.4.3", "audience": "framework users"})
+    .info({"version": "4.1.4.4", "audience": "framework users"})
     .instruct("Return only facts grounded in the input.")
     .input("Summarize this release line for an engineering changelog.")
     .output({
@@ -346,7 +346,7 @@ The older `tool_func` / `use_tools` / `use_mcp` / `use_sandbox` family remains a
 
 ### 5. Runtime Skills
 
-Skills are reusable task guidance packages. In 4.1.4.3, `SkillLibrary` owns
+Skills are reusable task guidance packages. In 4.1.4.4, `SkillLibrary` owns
 immutable installed revisions, AgentExecution owns selection and exact-revision
 binding, and TaskContext owns progressive disclosure. `agent.use_skills(...)`
 is the normal candidate-binding surface; `agent.require_skills(...)` binds a
@@ -517,7 +517,7 @@ pip install agently-devtools
 agently-devtools init my_project
 ```
 
-Agently 4.1.4.3 recommends `agently-devtools >=0.1.10,<0.2.0`.
+Agently 4.1.4.4 recommends `agently-devtools >=0.1.10,<0.2.0`.
 
 ## Architecture
 
@@ -692,7 +692,7 @@ Agently-Skills gives coding agents current Agently implementation guidance.
 - Repository: https://github.com/AgentEra/Agently-Skills
 - Current catalog generation: `v2`
 - Recommended bundle: `app`
-- Agently 4.1.4.3 compatibility: Skills authoring protocol `agently-skills.authoring.v2`
+- Agently 4.1.4.4 compatibility: Skills authoring protocol `agently-skills.authoring.v2`
 
 Use it when asking Codex, Claude Code, Cursor, or another coding agent to implement Agently patterns.
 
@@ -767,8 +767,8 @@ Use the async request APIs directly or wrap agents, requests, generators, Trigge
 
 ## Compatibility Notes
 
-- The current package version is `4.1.4.3`.
-- The current release manifest is `compatibility/releases/4.1.4.3.json`.
+- The current package version is `4.1.4.4`.
+- The current release manifest is `compatibility/releases/4.1.4.4.json`.
 - Development-line planning belongs in `compatibility/in-development.json`; do not treat planned future versions as released.
 - `AgentExecutionResult.get_data()` returns business data; callers that need status, TaskBoard, diagnostics, or other route/task internals use `get_full_data()`.
 - An explicitly captured `AgentExecution` represents one run. Create a fresh execution for the next request after it starts.

diff --git a/README_CN.md b/README_CN.md
@@ -1,6 +1,6 @@
 <img width="640" alt="Agently" src="https://github.com/user-attachments/assets/c645d031-c8b0-4dba-a515-9d7a4b0a6881" />
 
-# Agently 4.1.4.3 - AI 应用运行时框架
+# Agently 4.1.4.4 - AI 应用运行时框架
 
 > 构建具备结构化输出、可观测 Actions、运行时 Skills、MCP 能力、过程流和可恢复工作流的 AI 服务后端。
 
@@ -32,13 +32,12 @@ Agently 面向的是正在从“模型偶尔能做对”走向“应用必须稳
 
 核心设计问题是：怎样保留模型能力，同时让应用代码拥有稳定契约、可观测执行和可重启的工作流边界？
 
-Agently 4.1.4.3 是建立在 4.1.4.2 owner-boundary release 上的结构化输出兼容性
-补丁。Pydantic v2 `BaseModel` 类型现在可以直接作为 `ModelRequest` 和
-`AgentExecution` 的 `.output(...)` 契约，并支持嵌套模型；4.1.4.2 的运行时所有权
-边界保持不变。阅读
-[4.1.4.3 Release Notes](docs/cn/development/release-notes-4.1.4.3.md)、
-[4.1.4 Release Notes](docs/cn/development/release-notes-4.1.4.md) 和
-[4.1.3.9 Release Notes](docs/cn/development/release-notes-4.1.3.9.md)
+Agently 4.1.4.4 加强了 Pydantic 字段约束提示与修正重试，增加具备恢复语义的
+TriggerFlow 快照投影和有界本地快照保留，并让默认发布校验不再依赖本地
+Ollama。4.1.4.2 的运行时所有权边界和 companion protocol 保持不变。阅读
+[4.1.4.4 Release Notes](docs/cn/development/release-notes-4.1.4.4.md)、
+[4.1.4.3 Release Notes](docs/cn/development/release-notes-4.1.4.3.md) 和
+[4.1.4 Release Notes](docs/cn/development/release-notes-4.1.4.md)
 了解当前发布与升级路径。
 
 ## 为什么选择 Agently
@@ -47,7 +46,7 @@ Agently 4.1.4.3 是建立在 4.1.4.2 owner-boundary release 上的结构化输
 
 当你关心这些问题时，Agently 会比较合适：
 
-- **AI 服务应该是运行时执行，不是 prompt glue** - `AgentExecution` 统一管理单次运行的 Prompt、strategy、Actions、Skill bindings、process stream、TaskContext evidence 和 result views，覆盖 direct、flat、TaskBoard strategy。阅读 [4.1.4.3 Release Notes](docs/cn/development/release-notes-4.1.4.3.md)、[4.1.4 Release Notes](docs/cn/development/release-notes-4.1.4.md) 和 [Agent Auto Orchestration 示例](examples/agent_auto_orchestration/)。
+- **AI 服务应该是运行时执行，不是 prompt glue** - `AgentExecution` 统一管理单次运行的 Prompt、strategy、Actions、Skill bindings、process stream、TaskContext evidence 和 result views，覆盖 direct、flat、TaskBoard strategy。阅读 [4.1.4.4 Release Notes](docs/cn/development/release-notes-4.1.4.4.md)、[4.1.4 Release Notes](docs/cn/development/release-notes-4.1.4.md) 和 [Agent Auto Orchestration 示例](examples/agent_auto_orchestration/)。
 - **换模型不应重写业务逻辑** - Agently 把 provider setup、Prompt 槽位、响应解析、Action 执行和响应读取归一到同一套 request/runtime contract。阅读 [模型设置](docs/cn/start/model-setup.md)、[模型概览](docs/cn/models/overview.md) 和 [Requests 概览](docs/cn/requests/overview.md)。
 - **结构化输出应是框架保障，不只是 provider 能力** - `.output(...)` schema、必填字段提取、parser feedback、重试、`ensure_keys`、`ensure_all_keys` 和 validation handlers 在 Agently 内部协同工作。阅读 [Schema as Prompt](docs/cn/requests/schema-as-prompt.md)、[输出控制](docs/cn/requests/output-control.md) 和 [`examples/basic/`](examples/basic/)。
 - **流式输出应在最后一个 token 前暴露结构** - `instant` mode 允许消费者在模型仍在流式输出时响应结构化字段，适合 UI 更新、SSE routes 和 workflow signals。阅读 [模型响应](docs/cn/requests/model-response.md)、[FastAPI 服务封装](docs/cn/services/fastapi.md) 和 [`examples/fastapi/`](examples/fastapi/)。
@@ -59,7 +58,7 @@ Agently 4.1.4.3 是建立在 4.1.4.2 owner-boundary release 上的结构化输
 - **常见模型应用模式应该可组合** - router、To-Do/dependency execution、planning、reflection、evaluator/reviser 和多 Agent 协作，都可以由同一套 request/action/signal primitives 组合出来。阅读 [Playbooks](docs/cn/playbooks/overview.md)、[TriggerFlow 模型集成](docs/cn/triggerflow/model-integration.md) 和 [`examples/step_by_step/`](examples/step_by_step/)。
 - **服务应保持清晰项目边界** - async API、FastAPI helpers、settings 文件、prompt 文件、DevTools 观测和 companion coding-agent skills 适合非一次性项目。阅读 [项目结构](docs/cn/start/project-framework.md)、[FastAPI 服务封装](docs/cn/services/fastapi.md) 和 [观测概览](docs/cn/observability/overview.md)。
 
-当前框架版本：`4.1.4.3`。
+当前框架版本：`4.1.4.4`。
 
 Python：`>=3.10`。
 
@@ -206,7 +205,7 @@ Prompt 由命名槽位组成。这样应用意图、约束、上下文和输出
 result = (
     agent
     .role("你是简洁的 release note 作者。")
-    .info({"version": "4.1.4.3", "audience": "framework users"})
+    .info({"version": "4.1.4.4", "audience": "framework users"})
     .instruct("只基于输入事实作答。")
     .input("为工程 changelog 总结这个发布线。")
     .output({
@@ -341,7 +340,7 @@ raw = agent.action.read_action_artifact(
 
 ### 5. Runtime Skills
 
-Skills 是可复用的任务指导包。在 4.1.4.3 中，`SkillLibrary` 负责不可变的已安装
+Skills 是可复用的任务指导包。在 4.1.4.4 中，`SkillLibrary` 负责不可变的已安装
 revision，AgentExecution 负责选择和精确 revision binding，TaskContext 负责渐进
 disclosure。`agent.use_skills(...)` 是普通候选绑定入口；已知精确 revision 时使用
 `agent.require_skills(...)`。`Agently.skills_executor` 只保留安装、检查、context
@@ -503,7 +502,7 @@ pip install agently-devtools
 agently-devtools init my_project
 ```
 
-Agently 4.1.4.3 推荐 `agently-devtools >=0.1.10,<0.2.0`。
+Agently 4.1.4.4 推荐 `agently-devtools >=0.1.10,<0.2.0`。
 
 ## 架构
 
@@ -678,7 +677,7 @@ Agently-Skills 为 coding agent 提供当前 Agently 实现指导。
 - Repository: https://github.com/AgentEra/Agently-Skills
 - 当前 catalog generation: `v2`
 - 推荐 bundle: `app`
-- Agently 4.1.4.3 compatibility: Skills authoring protocol `agently-skills.authoring.v2`
+- Agently 4.1.4.4 compatibility: Skills authoring protocol `agently-skills.authoring.v2`
 
 当你让 Codex、Claude Code、Cursor 或其他 coding agent 实现 Agently 模式时，应使用它。
 
@@ -753,8 +752,8 @@ CrewAI 和 AutoGen 在以 agent 协作为核心的设计里很强。Agently 是
 
 ## 兼容说明
 
-- 当前 package version 是 `4.1.4.3`。
-- 当前 release manifest 是 `compatibility/releases/4.1.4.3.json`。
+- 当前 package version 是 `4.1.4.4`。
+- 当前 release manifest 是 `compatibility/releases/4.1.4.4.json`。
 - 开发线计划写入 `compatibility/in-development.json`；不要把未来计划版本当作已发布版本。
 - `AgentExecutionResult.get_data()` 返回业务数据；需要 status、TaskBoard、diagnostics 或其他 route/task 内部信息时使用 `get_full_data()`。
 - 显式捕获的 `AgentExecution` 代表一次运行；启动后，下一次请求应创建新的 execution。