dsh-codex-subscription 2.1.0 → 2.1.1-beta.2

package/README.en.md

@@ -130,10 +130,9 @@ Restart DSH manually after installation, then:
 - Optional percentage, progress bar, or Beta runway estimate for the selected Codex model (off by default);
 - Standard or Fast mode for supported Codex models directly in the composer;
 - Standard, Extended, and per-model Custom context windows; Custom accepts a full numeric token count, stays within each audited model capacity, and feeds DSH's native agent compaction policy;
-- A copyable support report and direct feedback link in Settings; the report includes bounded request stages, HTTP/transport classes, elapsed ranges, and route source types while excluding OAuth credentials, account identifiers, proxy addresses, and authorization timestamps;
+- A copyable support report and direct feedback link in Settings; the report includes bounded request stages, HTTP/transport classes, elapsed ranges, and route source types while excluding OAuth credentials, account identifiers, and authorization timestamps;
 - Visible errors when subscription routing is unavailable, with no silent paid fallback.
 
-The plugin can follow an existing HTTPS proxy from the process environment or operating-system proxy settings for official OpenAI and ChatGPT requests. It does not provide a proxy, relay, node list, or system-proxy configuration.
 
 ### GPT-6 Astra context
 
@@ -245,6 +244,13 @@ With a supported Codex model selected, open the composer's model menu to choose
 Standard adds no icon; only Fast shows a lightning icon before the model name. Spark does not show the speed entry. Fast mode increases speed and uses more Credits;
 see the [OpenAI Codex Speed documentation](https://learn.chatgpt.com/docs/agent-configuration/speed) for the current rules.
 
+### Advanced experiments (2.1.1 Beta)
+
+Opt in under **Advanced & diagnostics**. SSE and DSH subtasks remain the defaults:
+
+- **WebSocket** reuses connections and eligible context transfers. Failed handshakes can fall back to SSE; interrupted responses surface an error without automatic replay. Applies to the next request, does not expand context limits, and is not guaranteed to be faster.
+- **Codex independent subtasks** reuse your subscription login and the official DSH Codex runtime, without a separate login or CLI setup. They follow the current subscription model and workspace permissions; other model sessions use Luna low. Shared-context subtasks remain with DSH.
+
 ## Update and uninstall
 
 ### Update and verify
@@ -289,7 +295,7 @@ npx -y @deepseek-ai/dsh@0.1.5-rc.1 plugin --profile web remove dsh-codex-subscri
 - **`dsh` is not recognized:** the official npm route does not create a global `dsh` command; use the complete `npx -y @deepseek-ai/dsh@0.1.5-rc.1 ...` command above;
 - **More than one DSH exists:** run the standard command from the intended DSH environment so that product selects the corresponding profile;
 - **Setup still fails:** confirm the command is running in the intended DSH environment. Do not delete the profile or change the system PATH to force an install.
-- **Need to report a problem:** generate a **Support diagnostics** report at the bottom of Settings, then open the [bug report form](https://github.com/WSL043/dsh-codex-subscription/issues/new?template=install-problem.yml). The report includes the OS/runtime, bounded sign-in phase, and safe request-failure categories, but excludes credentials, account identifiers, proxy addresses, raw responses, and full logs. Paste it into the required diagnostics field; never attach sign-in URLs, authorization codes, or browser callback addresses.
+- **Need to report a problem:** generate a **Support diagnostics** report at the bottom of Settings, then open the [bug report form](https://github.com/WSL043/dsh-codex-subscription/issues/new?template=install-problem.yml). The report includes the OS/runtime, bounded sign-in phase, and safe request-failure categories, but excludes credentials, account identifiers, raw responses, and full logs. Paste it into the required diagnostics field; never attach sign-in URLs, authorization codes, or browser callback addresses.
 
 The ChatGPT Codex backend and DSH can change independently. This community project is not affiliated with or endorsed by DeepSeek or OpenAI.
 

package/README.md

@@ -131,10 +131,9 @@ dsh --profile web --dump-config
 - 输入框可用百分比、进度条或可选的 Beta 续航预测显示当前 Codex 模型的剩余额度（默认关闭）；
 - 输入框可为支持的 Codex 模型切换标准或高速模式；
 - 上下文窗口提供标准、扩展和逐模型自定义；自定义直接填写完整 Token 数值，并在已审核的模型容量内交给 DSH 原生 Agent 压缩策略处理；
-- 设置页可生成并复制无敏感信息的支持诊断，并直接打开反馈入口；报告包含有限的请求阶段、HTTP/网络分类、耗时区间和路由来源类型，但不包含 OAuth 凭据、账号标识、代理地址或授权时间；
+- 设置页可生成并复制无敏感信息的支持诊断，并直接打开反馈入口；报告包含有限的请求阶段、HTTP/网络分类、耗时区间和路由来源类型，但不包含 OAuth 凭据、账号标识或授权时间；
 - 订阅路由不可用时明确报错，不会静默切换到其他付费路由。
 
-对于 OpenAI 与 ChatGPT 官方请求，插件可以沿用进程环境或操作系统中已有的 HTTPS 代理。插件本身不提供代理、转发服务、节点列表，也不会修改系统代理设置。
 
 ### GPT-6 Astra 上下文
 
@@ -244,6 +243,13 @@ ChatGPT 返回可用重置卡时，设置页会把每张卡分别显示为紧凑
 只有高速模式会在模型名称左侧显示闪电；Spark 不显示速度入口。高速模式会提高速度，也会消耗更多 Credits；具体规则见
 [OpenAI Codex Speed 文档](https://learn.chatgpt.com/docs/agent-configuration/speed)。
 
+### 高级实验选项（2.1.1 Beta）
+
+在 **高级与诊断** 中按需开启，默认仍使用 SSE 和 DSH 子任务：
+
+- **WebSocket**：复用连接和可复用的上下文传输；握手失败可回退 SSE，已开始的响应中断会报错，不自动重放。下次请求生效，不增加模型上下文容量，也不保证更快。
+- **Codex 独立子任务**：复用订阅登录和官方 DSH Codex 运行时，无需另行登录或配置 CLI。跟随当前订阅模型及工作区权限；其他模型会话使用 Luna low。共享上下文的子任务仍交给 DSH。
+
 ## 更新与卸载
 
 ### 更新并检查
@@ -288,7 +294,7 @@ npx -y @deepseek-ai/dsh@0.1.5-rc.1 plugin --profile web remove dsh-codex-subscri
 - **`dsh` 无法识别**：官方 npm 方式本来就不会创建全局 `dsh` 命令，请使用上面的完整 `npx -y @deepseek-ai/dsh@0.1.5-rc.1 ...` 命令；
 - **电脑上有多个 DSH**：请从目标 DSH 环境运行标准命令，由该产品自身选择对应 profile；
 - **安装仍然失败**：确认命令是在目标 DSH 环境中运行，不要删除 profile 或随意修改系统 PATH。
-- **需要提交问题**：在设置页底部生成“支持诊断”，然后打开[使用问题表单](https://github.com/WSL043/dsh-codex-subscription/issues/new?template=install-problem.yml)。报告包含系统/运行时、有限的登录阶段和安全的请求失败分类，但不含凭据、账号标识、代理地址、原始响应或完整日志；请粘贴到必填诊断栏，且不要附上登录链接、授权码或浏览器回调地址。
+- **需要提交问题**：在设置页底部生成“支持诊断”，然后打开[使用问题表单](https://github.com/WSL043/dsh-codex-subscription/issues/new?template=install-problem.yml)。报告包含系统/运行时、有限的登录阶段和安全的请求失败分类，但不含凭据、账号标识、原始响应或完整日志；请粘贴到必填诊断栏，且不要附上登录链接、授权码或浏览器回调地址。
 
 ## 边界与支持
 

package/THIRD_PARTY_NOTICES.md

@@ -5,11 +5,13 @@ This project depends on software distributed under its own terms. The dependency
 | Project | Role | License | Source |
 | --- | --- | --- | --- |
 | DeepSeek Harness packages | Plugin host, client slots, credential and LLM interfaces | MIT | https://github.com/deepseek-ai/deepseek-harness |
+| `@openai/codex` 0.153.4 | Optional official DSH Codex subagent runtime, installed as a dependency | Apache-2.0 | https://github.com/openai/codex |
 | `@earendil-works/pi-ai` 0.82.1 | OpenAI Codex OAuth, model catalog, Responses transport, and WebSocket continuation | MIT | https://github.com/earendil-works/pi |
 | React | DSH settings component runtime | MIT | https://github.com/facebook/react |
 | dsh-image-viewer 0.1.0-beta.9 (3d121d3), with 0.1.0 (7f3be2c) zoom and annotation improvements | Adapted built-in image viewer, interaction state, and styles | MIT, Copyright (c) 2026 WSL043; full license included in LICENSE | https://github.com/WSL043/dsh-image-viewer |
 | Heroicons | Composer speed icon | MIT | https://github.com/tailwindlabs/heroicons |
 | `https-proxy-agent` 7.0.6 | HTTPS proxy transport for Codex OAuth token requests | MIT | https://github.com/TooTallNate/proxy-agents |
+| `ws` 8.21.3 | Request-scoped subscription WebSocket transport | MIT | https://github.com/websockets/ws |
 | QuotaPin for Codex | Design basis for the opt-in local quota runway forecast | MIT | https://github.com/WSL043/QuotaPin-for-Codex |
 | tsdown | Development-time bundler | MIT | https://github.com/rolldown/tsdown |
 

package/compatibility.json

@@ -6,6 +6,7 @@
   ],
   "previews": [
     "0.1.5-alpha.1",
-    "0.1.5-alpha.2"
+    "0.1.5-alpha.2",
+    "0.1.6-alpha.1"
   ]
 }