dsh-soul 0.3.3 → 0.5.0

package/README.md

@@ -7,6 +7,7 @@ DeepSeek Harness 个性化设置插件，用于配置 Agent 的昵称、回复
 ## 功能
 
 - Web UI 个性化设置页面
+- 设置页体验：中英双语文案（随界面语言切换）、dirty 检测（无改动禁用保存）、保存结果提示、「查看当前生效提示词」只读面板
 - 启用或禁用个性化设置
 - 「关于你」：设置用户昵称、职业和介绍，回复时结合你的背景
 - 选择回复风格和语调（合并为单一选项）：`professional`（专业严谨）、`casual`（轻松自然）、`humorous`（幽默风趣）、`roast`（吐槽达人）、`efficient`（高效干练）
@@ -14,17 +15,22 @@ DeepSeek Harness 个性化设置插件，用于配置 Agent 的昵称、回复
   - 标题和列表：`default`（默认）、`more`（增强，采用清晰格式和列表结构）、`less`（减弱，使用更多段落文本）
   - 表情符号：`default`（默认）、`more`（增强，使用较多表情符号）、`less`（减弱，尽量减少使用表情符号）
 - 选择输出语言（Agent 回复语言 + `/soul` 命令输出语言）：中文 / English
+- 提示词随输出语言本地化：`language=en` 时 system prompt 使用英文描述
 - 输入自定义指令
 - Agent 可调用工具 `set_persona`，让模型在对话中直接调整人设
+- 人设预设：多套命名人设一键切换（保存 / 使用 / 列表 / 删除，Web UI 与命令双入口）
+- `set_persona` 确认模式（`requireToolConfirmation`）：Agent 的人设修改需经 `/soul confirm` 确认后才生效
+- `/soul set` 键值方式修改配置项
 - 配置持久化保存
 - 配置输入校验：字段白名单、类型、长度上限（昵称/职业 50、介绍 500、自定义指令 2000 字符）与枚举校验，非法或超限字段整单拒绝
 - 配置更新后同步到所有活动 Agent
+- 变更检测：仅在配置实际变化时刷新提示词并注入会话，无变化的保存不产生注入消息
 
 ## 安装
 
 ```powershell
 dsh plugin --profile web add dsh-soul
-dsh --profile web web
+dsh plugin --profile web update dsh-soul
 ```
 
 插件配置由 `cordis.patch.yml` 提供：
@@ -60,16 +66,23 @@ dsh --profile web web
 也可以使用斜杠命令：
 
 ```text
-/soul show       查看当前配置
-/soul reset      重置配置
-/soul enable     启用个性化设置
-/soul disable    禁用个性化设置
-/soul 小明       设置昵称
+/soul show        查看当前配置（含确认模式与预设数量）
+/soul set k=v     修改配置项（如 /soul set style=humorous language=en）
+/soul save <名>   保存当前配置为人设预设
+/soul use <名>    应用人设预设
+/soul list        查看人设预设（✔ 标记当前匹配项）
+/soul del <名>    删除人设预设（delete / rm 别名）
+/soul confirm     应用待确认的人设变更（确认模式）
+/soul reject      拒绝待确认的人设变更
+/soul reset       重置配置（保留人设预设库）
+/soul enable      启用个性化设置
+/soul disable     禁用个性化设置
+/soul 小明        设置昵称
 ```
 
-配置保存后会同步到所有活动 Agent，当前会话下一次请求即可使用最新配置。
+配置保存后会同步到所有活动 Agent，当前会话下一次请求即可使用最新配置；无实际变化的保存不产生注入消息。
 
-Agent 也可以通过工具 `set_persona` 在对话中直接调整你的人设（昵称、回复风格和语调、特质、回复语言、自定义指令）。模型只会在明确请求改变称呼、语气、风格或语言时调用该工具。
+Agent 也可以通过工具 `set_persona` 在对话中直接调整你的人设（昵称、回复风格和语调、特质、回复语言、自定义指令）。模型只会在明确请求改变称呼、语气、风格或语言时调用该工具。开启「人设变更需确认」（`requireToolConfirmation`）后，该工具的修改会以待确认提议返回，需使用 `/soul confirm` 确认或 `/soul reject` 拒绝后才会生效。
 
 ## 配置文件
 
@@ -89,10 +102,13 @@ soul-config.json
   "bio": "对编程和技术感兴趣",
   "style": "professional",
   "language": "zh",
-  "customInstructions": "请保持简洁，优先给出结论。"
+  "customInstructions": "请保持简洁，优先给出结论。",
+  "requireToolConfirmation": false
 }
 ```
 
+人设预设保存在同一文件的 `personas` 字段：名称 → 人设字段快照（昵称/职业/介绍/风格/特质/语言/自定义指令）+ `updatedAt`；预设库变更不影响活动配置，使用预设时才应用到配置并同步会话。
+
 字段长度上限：昵称 / 职业 50 字符，介绍 500 字符，自定义指令 2000 字符；未知字段会被丢弃，非法或超限字段整单拒绝（HTTP 返回 400 与字段级错误明细）。
 
 ## 实现原理

package/README_EN.md

@@ -7,6 +7,7 @@ A personalization plugin for DeepSeek Harness (DSH). Configure your agent's nick
 ## Features
 
 - Personalization settings page in the Web UI
+- Localized settings UI (English/Chinese, follows the interface language), dirty tracking (save disabled when unchanged), save-result toasts, and a read-only "active prompt" viewer
 - Enable or disable personalization
 - "About you": set your nickname, occupation and bio so replies fit your background
 - Combined style & tone option: `professional`, `casual`, `humorous`, `roast`, `efficient`
@@ -14,17 +15,22 @@ A personalization plugin for DeepSeek Harness (DSH). Configure your agent's nick
   - Headings & lists: `default`, `more` (clear formatting with headings and lists), `less` (more paragraph text)
   - Emoji: `default`, `more` (frequent emoji usage), `less` (minimal emoji usage)
 - Output language (agent reply language + `/soul` command output language): Chinese or English
+- The compiled system prompt follows the output language (English descriptions when `language=en`)
 - Custom instructions
 - Agent-callable tool `set_persona` to let the model adjust persona during a conversation
+- Named personas: save the current config under a name and switch with one click (save / use / list / delete, in both the Web UI and slash commands)
+- `set_persona` confirmation mode (`requireToolConfirmation`): agent persona changes take effect only after `/soul confirm`
+- `/soul set` key=value field updates
 - Configuration persisted to disk
 - Input validation: field whitelist, types, length limits (nickname/occupation 50, bio 500, custom instructions 2000 chars) and enum checks; invalid or oversized fields reject the whole write
 - Configuration synced to all active agents after every update
+- Change detection: the prompt is refreshed and sessions injected only when the configuration actually changed — no-op saves inject nothing
 
 ## Installation
 
 ```powershell
 dsh plugin --profile web add dsh-soul
-dsh --profile web web
+dsh plugin --profile web update dsh-soul
 ```
 
 The plugin is registered via `cordis.patch.yml`:
@@ -60,16 +66,23 @@ After starting DSH, open the **Personalization** section on the settings page, e
 Slash commands are also available:
 
 ```text
-/soul show       Show current configuration
-/soul reset      Reset configuration
-/soul enable     Enable personalization
-/soul disable    Disable personalization
-/soul Bob        Set nickname
+/soul show        Show current configuration (confirmation mode & persona count)
+/soul set k=v     Change config fields (e.g. /soul set style=humorous language=en)
+/soul save <name> Save the current config as a persona
+/soul use <name>  Apply a persona
+/soul list        List personas (✔ marks the active match)
+/soul del <name>  Delete a persona (delete / rm aliases)
+/soul confirm     Apply the pending persona proposal (confirmation mode)
+/soul reject      Discard the pending persona proposal
+/soul reset       Reset configuration (keeps the persona library)
+/soul enable      Enable personalization
+/soul disable     Disable personalization
+/soul Bob         Set nickname
 ```
 
-Once saved, the configuration is synced to all active agents and takes effect on the next request in the current session.
+Once saved, the configuration is synced to all active agents and takes effect on the next request in the current session; no-op saves inject nothing.
 
-The agent can also use the `set_persona` tool to adjust your persona (nickname, style & tone, traits, reply language, custom instructions) during a conversation. The model only invokes this tool when you explicitly ask to change how it addresses or responds to you.
+The agent can also use the `set_persona` tool to adjust your persona (nickname, style & tone, traits, reply language, custom instructions). The model only invokes this tool when you explicitly ask to change how it addresses or responds to you. With **Require confirmation for persona changes** (`requireToolConfirmation`) enabled, tool changes come back as a pending proposal and take effect only after `/soul confirm` (or are discarded by `/soul reject`).
 
 ## Configuration File
 
@@ -89,10 +102,13 @@ Example:
   "bio": "Interested in programming and technology",
   "style": "professional",
   "language": "en",
-  "customInstructions": "Be concise and lead with the conclusion."
+  "customInstructions": "Be concise and lead with the conclusion.",
+  "requireToolConfirmation": false
 }
 ```
 
+Personas are stored in the same file under the `personas` field: name → persona field snapshot (nickname / occupation / bio / style / traits / language / custom instructions) + `updatedAt`. Persona library changes never touch the active config; a persona is applied to the config (and synced to sessions) only when used.
+
 Length limits: nickname / occupation 50 chars, bio 500 chars, custom instructions 2000 chars. Unknown fields are dropped; invalid or oversized fields reject the whole write (HTTP returns 400 with per-field error details).
 
 ## How It Works