dsh-more 0.0.2 → 0.0.4

package/CONTRIBUTING.md

@@ -12,7 +12,7 @@
 
 - Node.js `22.19.x` 或 `>= 24`；
 - pnpm；
-- 与项目当前依赖匹配的 DSH。本版本面向 DSH `0.1.0-rc.7`。
+- 与项目当前依赖匹配的 DSH。本版本面向 DSH `0.1.2-rc.1`。
 
 首次检出后执行：
 

package/README.md

@@ -6,7 +6,7 @@ Practical, independently switchable patches for missing context and history cont
 
 `dsh-more` integrates with the existing DSH interface instead of adding a separate management dashboard. Its current patch set can edit a user message and restart from that point, remove one message while preserving the surrounding context, and permanently delete a session while leaving native archive behavior intact.
 
-> **Compatibility:** the current release targets DeepSeek Harness `0.1.0-rc.7` and requires Node.js `>= 24`. DSH is still a release candidate, so upgrades may require changes in the centralized adapter layer described below.
+> **Compatibility:** the current release targets DeepSeek Harness `0.1.2-rc.1` and requires Node.js `>= 24`. DSH is still a release candidate, so upgrades may require changes in the centralized adapter layer described below.
 
 ## Features
 
@@ -23,7 +23,7 @@ All three switches live under **Settings → Plugins → Plugin configuration
 ### Requirements
 
 - Node.js `>= 24`
-- A working `dsh` CLI and a Web profile compatible with DSH `0.1.0-rc.7`
+- A working `dsh` CLI and a Web profile compatible with DSH `0.1.2-rc.1`
 - Access to the npm registry that publishes `dsh-more`
 
 pnpm is needed only when installing or developing from source. For a published release, use the DSH plugin command so the package is installed into the selected DSH profile; do not install it globally with `npm install -g`.
@@ -131,6 +131,17 @@ After confirmation, DSH More cancels a running task, waits for idle, flushes the
 
 > **Irreversible:** permanent deletion does not archive the session and does not move it into a DSH More trash directory.
 
+### Configure model reasoning effort
+
+Open **Settings → Models**, expand a provider with **Edit**, and every model entry in the model catalog carries its own **Thinking effort** button:
+
+1. Click a model's **Thinking effort**; in the dialog, check the levels the model offers (off / minimal / low / medium / high / xhigh / max) and, for each enabled level, fill in the wire spelling sent to the gateway (defaults to the level name); `off` needs no wire value — "supported, send nothing".
+2. Click **Save**. The change is written to `reasoningEfforts` under `llm-pi-ai.providers.<route>.models[]` in `settings.yaml`, and the composer's model picker immediately offers those levels.
+
+You can also check **This model does not reason** (writes `reasoningEfforts: false`) or use **Clear** to drop the field and fall back to the installed catalog. Only `off` may leave its wire value empty, and at least one non-`off` level is required — the same rules the pi-ai adapter enforces when resolving a route. Invalid configurations are rejected before any write, so a bad save never corrupts `settings.yaml`.
+
+**Auto-fill:** models added via **Fetch available models** carry no `reasoningEfforts`, which the adapter treats as non-reasoning. After the provider is saved, this patch auto-fills the default `{ low, medium, high }` onto any model that has no `reasoningEfforts`, so a freshly fetched model can think immediately. To keep a model non-reasoning, set it to `reasoningEfforts: false` (explicit `false` is never auto-filled). The default levels can then be tuned per model from the thinking-effort button.
+
 ## Architecture
 
 ![Manifest-driven DSH More architecture](./docs/assets/dsh-more-architecture.jpg)
@@ -222,7 +233,9 @@ The repository does not currently contain a full browser end-to-end suite agains
 
 ## Compatibility notes
 
-- The current dependency set targets DSH `0.1.0-rc.7`; pre-release API or DOM changes can require adapter updates.
+This update migrates to the split Client services, `useChat` snapshots, settings registration, and session projection APIs, and updates message action row targeting. Dependencies and the lockfile target `0.1.2-rc.1` without the retired `dsh-client-runtime`. For an existing local linked install, rebuild, restart DSH, and reload the page to load both the new Host and Client.
+
+- The current dependency set targets DSH `0.1.2-rc.1`; pre-release API or DOM changes can require adapter updates.
 - In-product DSH More copy is currently primarily Simplified Chinese, although native Copy/Archive selectors recognize both Chinese and English labels.
 - Message actions operate only on ordinary user messages and assistant messages still present in the current model surface; editing requires a completed owning turn.
 - Permanent deletion supports DSH persistence locations of kind `jsonl` only.

package/README.zh-CN.md

@@ -6,7 +6,7 @@
 
 `dsh-more` 不创建平行的“管理中心”，而是把能力直接嵌入 DSH 现有界面。当前补丁可以编辑用户消息并从该处重新开始、在保留前后上下文的前提下删除单条消息，以及在不改变原生归档语义的情况下永久删除会话。
 
-> **兼容性：** 当前发布面向 DeepSeek Harness `0.1.0-rc.7`，要求 Node.js `>= 24`。DSH 仍是 RC 版本，升级后可能需要调整下文所述的集中适配层。
+> **兼容性：** 当前发布面向 DeepSeek Harness `0.1.2-rc.1`，要求 Node.js `>= 24`。DSH 仍是 RC 版本，升级后可能需要调整下文所述的集中适配层。
 
 ## 功能概览
 
@@ -23,7 +23,7 @@
 ### 环境要求
 
 - Node.js `>= 24`
-- 可正常使用的 `dsh` CLI，以及与 DSH `0.1.0-rc.7` 兼容的 Web profile
+- 可正常使用的 `dsh` CLI，以及与 DSH `0.1.2-rc.1` 兼容的 Web profile
 - 能访问发布 `dsh-more` 的 npm registry
 
 只有从源码安装或参与开发时才需要 pnpm。安装已发布版本时，应通过 DSH 插件命令把包加入指定 profile；不要使用 `npm install -g` 全局安装。
@@ -131,6 +131,17 @@ dsh plugin --profile web remove dsh-more
 
 > **不可恢复：** 永久删除不会归档会话，也不会把会话移入 DSH More 回收目录。
 
+### 配置模型思考强度
+
+打开 **设置 → 模型**，点击某个提供方的 **编辑** 展开「模型目录」，每个模型条目下都有一个 **思考强度** 按钮：
+
+1. 点击模型的 **思考强度**，在弹窗中勾选该模型支持的级别（不思考 / 极简 / 低 / 中 / 高 / 极高 / 最高），并为每个开启的级别填写发送给网关的线上值（默认与级别同名）；`不思考` 不需要线上值，表示“支持、但不发送任何参数”；
+2. 点击 **保存**。修改会写入 `settings.yaml` 中 `llm-pi-ai.providers.<route>.models[].reasoningEfforts`，会话里的模型选择器随即按此提供级别。
+
+也可以勾选 **该模型不支持思考**（写入 `reasoningEfforts: false`），或用 **清除配置** 移除该字段、回到内置目录的能力。只有 `off` 允许留空线上值，且必须至少声明一个非 `off` 的级别，与 pi-ai 适配器的解析规则一致；无效配置会在写入前被拒绝，不会破坏 `settings.yaml`。
+
+**自动填充：** 通过 **获取可用模型** 添加的模型默认没有 `reasoningEfforts`，适配器会把它当作不思考。本补丁会在保存提供方后，为**没有任何 `reasoningEfforts` 的模型**自动填入默认思考强度 `{ low, medium, high }`，让新拉取的模型立刻可以思考；如需某个模型不思考，请把它设为 `reasoningEfforts: false`（显式 false 不会被自动填充）。默认级别可用每个模型条目下的思考强度按钮再微调。
+
 ## 架构
 
 ![DSH More 清单驱动架构](./docs/assets/dsh-more-architecture.zh-CN.jpg)
@@ -222,7 +233,9 @@ pnpm run package        # 构建并生成 npm tarball
 
 ## 兼容性说明
 
-- 当前依赖面向 DSH `0.1.0-rc.7`；预发布 API 或 DOM 变化可能要求更新 adapter。
+本次适配迁移到新版拆分后的 Client 服务、`useChat` 聊天快照、设置注册和会话投影接口，并更新消息操作行定位。依赖与锁文件统一到 `0.1.2-rc.1`；不再加载已停止更新的 `dsh-client-runtime`。已有本地链接安装在重新构建后，需要重启 DSH 并刷新页面以加载新的 Host 和 Client。
+
+- 当前依赖面向 DSH `0.1.2-rc.1`；预发布 API 或 DOM 变化可能要求更新 adapter。
 - DSH More 的产品内文案目前以简体中文为主；对 DSH 原生 Copy/Archive 元素的定位同时识别中英文标签。
 - 消息操作只面向仍在当前 model surface 中的普通用户消息和助手消息；编辑要求所属轮次已经完成。
 - 永久删除只支持 DSH 返回的 `jsonl` 持久化定位。