dsh-prompt-enhance 0.1.4 → 0.1.6

package/README.md

@@ -28,7 +28,7 @@ Good agent prompts state *who the model should be*, *what to deliver*, *in what
 | ⚙️ **Live settings** | Every knob (model override, temperature, budgets, system prompt, shortcut) hot-applies from Settings → 插件配置 |
 | 🛡️ **Draft safety** | Empty, over-length, images-only, and command-chip inputs are rejected locally; upstream failures are mapped to readable messages; the draft is never mutated on failure |
 
-![The preview panel running in dsh web: original and enhanced prompt side by side with model info, fill-back and copy actions](docs/evidence-prompt-enhance-panel.png)
+![The preview panel running in dsh web: original and enhanced prompt side by side with model info, fill-back and copy actions](https://raw.githubusercontent.com/rongxingda/dsh-prompt-enhance/main/docs/evidence-prompt-enhance-panel.png)
 
 ## How it works
 
@@ -99,7 +99,7 @@ dsh plugin --profile web remove dsh-prompt-enhance
 1. Local guards run first: empty drafts, over-length drafts (never auto-truncated — that would change your meaning), images-only drafts, and drafts containing command or file-reference chips are refused with a clear message. Chips are rejected because filling back would destroy them.
 2. The preview panel opens with a cancellable spinner. Your draft stays untouched — the hint says so.
 3. The result phase shows both texts side by side. **Apply** fills the enhanced text back and raises the undo bar; **Copy** puts it on the clipboard; **Cancel** (or `Esc`, or clicking the overlay) discards everything.
-4. The undo bar sits above the composer: one click restores the original. If you keep typing after applying, the bar quietly retires itself so stale text can never overwrite newer edits.
+4. The undo bar sits above the composer: one click restores the original. If you keep typing after applying, the bar quietly retires itself so stale text can never overwrite newer edits. Undo entries live **only in the current page's memory** (up to 3 per session) — a page reload, a web restart, or a session switch clears them.
 
 **`/enhance <text>`** — rewrite any text from the slash menu. The result renders in the command plane (copyable) and never enters the conversation history or the model's context. To enhance the *composer draft itself*, use the button or shortcut — the draft lives in the browser. Cancelling a running `/enhance` follows the harness command plane; if the client offers no cancel affordance, the call simply runs to completion or times out.
 
@@ -117,10 +117,11 @@ Everything lives in the `prompt-enhance` settings namespace, edited from the web
 | `maxOutputTokens` | `2048` | Output token budget of one enhancement call |
 | `maxInputChars` | `12000` | Input character cap (counted in Unicode code points — an emoji is one character); over-limit drafts are **rejected, never truncated** |
 | `timeoutMs` | `60000` | End-to-end deadline of one call |
-| `systemPrompt` | built-in strategy | Replace the whole enhancement strategy if you prefer your own |
+| `systemPrompt` | built-in strategy | Custom strategy text; how it combines with the built-in strategy is set by `strategyMode` |
+| `strategyMode` | `replace-default` | How a custom strategy combines with the built-in one: `replace-default` **swaps it out entirely** (backward compatible, but the built-in hard rules — preserve intent, never fabricate, body-only output, language mirroring — are not retained and must be carried into your own text); `extend-default` **appends your text after the built-in strategy**, keeping those rules in force |
 | `shortcut` | `ctrl+alt+e` | Global shortcut spec (at least one modifier + one alphanumeric/function key — bare keys are ignored so normal typing can never be swallowed); empty disables it |
-| `maxConcurrent` | `2` | Host-side concurrency cap; extra requests answer `429` |
-| `rateLimitPerMinute` | `10` | Sliding-window rate cap per minute; extra requests answer `429` |
+| `maxConcurrent` | `2` | Concurrency cap **within a single host process**; extra requests answer `429` (`concurrency-limit`). The browser UI admits exactly one in-flight request (a single preview panel), so this cap mainly protects the `/enhance` command plane and multi-client callers |
+| `rateLimitPerMinute` | `10` | Sliding-window rate cap per minute, **within a single host process**, counting **successful calls only** — failures (timeout / upstream error / cancellation) never consume the window, so a run of failures cannot rate-limit you out; extra requests answer `429` (`rate-limit`) with a `Retry-After` in seconds |
 | `provider` + `model` values | — | Match the harness settings: each key under `llm-pi-ai.providers` (e.g. `zhipu`, `muyuu`) is a provider and each `models[].id` under it (e.g. `glm-5.3-flash`) is a model. Example pair: `provider: zhipu` + `model: glm-5.3-flash` |
 
 **Model routing precedence:** explicit settings pair → the route recorded in the current session's request header → the harness-wide default model (`agent-default-model`). If none of them names a route (e.g. a fresh session with no default model), the plugin fails with an actionable message instead of guessing.
@@ -138,7 +139,7 @@ The default system prompt instructs the model to be a prompt-rewriting expert an
 
 Hard rules: preserve intent exactly (never remove, alter, or contradict user information); never fabricate — insert an explicit placeholder like `(待补充：…)` / `(TBD: …)` for unknown details; keep the scope unchanged; output **only** the rewritten body (no explanations, fences, or pleasantries); mirror the input language; stay within roughly 1–3× the original length; lightly polish already-well-formed prompts instead of inflating them.
 
-Set `systemPrompt` in the settings to replace all of this with your own strategy.
+Set `systemPrompt` in the settings to use your own strategy; `strategyMode` decides how it combines. The default `replace-default` is a **complete replacement** — the built-in safety constraints (preserve intent, never fabricate, body-only output, mirror the input language, treat the framed draft as data) are not retained automatically, so your own strategy must include them. Switching to `extend-default` appends your text after the built-in strategy instead, keeping those rules in force.
 
 ## Error handling & edge cases
 
@@ -149,6 +150,8 @@ Set `systemPrompt` in the settings to replace all of this with your own strategy
 | Images attached but no text | Refused: text-only feature |
 | Command or file-reference chips in the draft | Refused: fill-back would destroy the chips |
 | Submitting / busy phase or a request already in flight | Refused with "try again in a moment" |
+| Per-minute cap exceeded | `429` (`rate-limit`) with a `Retry-After` in seconds; the host message names the exact wait. The window counts successful calls only — failures do not consume it |
+| Concurrency cap full | `429` (`concurrency-limit`), no `Retry-After` — a slot frees whenever an in-flight call settles |
 | Upstream model failure | Stable codes mapped to readable hints: `AUTH` → check API key, `RATE_LIMIT` → retry later, `QUOTA_EXCEEDED` → check balance, `CONTEXT_WINDOW_EXCEEDED` → shorten input, `NO_ADAPTER`/unconfigured → configure a model |
 | Output reaches `maxOutputTokens` | Refused with a hint to raise the cap or shorten the draft |
 | Model returns empty / fence-wrapped / tool-call output | Normalized (fences stripped) or refused; retryable |
@@ -158,14 +161,49 @@ Set `systemPrompt` in the settings to replace all of this with your own strategy
 
 Every failure surfaces inside the plugin's own panel; the composer draft is never modified by a failed call, so manual input continues undisturbed.
 
+## Error codes & localization
+
+The host route answers structured errors of the shape `{ code, message?, params? }`: the browser renders its localized primary line from `code` + `params` using the current language dictionary, and `message` — when present — is an optional diagnostic detail (the provider's raw failure text, a config error) shown verbatim beneath the primary line. The `/enhance` command plane has no locale dictionary; a host-side renderer of the same errors produces the Chinese text directly.
+
+| Error code | HTTP status | Browser primary copy (dictionary key) | Params |
+|---|---|---|---|
+| `rejected` | 403 / 413 / 415 / 422 | generic `error.rejected`; with `{ count, max }` it reuses `error.tooLong` | over-length input: `{ count, max }` |
+| `rate-limit` | 429 | `error.rateLimit` | `{ limit, retryAfterSeconds }` |
+| `concurrency-limit` | 429 | `error.concurrencyLimit` | `{ max }` |
+| `timeout` | 504 | `error.timeout` | `{ seconds }` |
+| `unconfigured` | 409 | `error.unconfigured` | — |
+| `upstream` | 502 | generic `error.upstream`; with `reason` it uses `error.upstream.{reason}` (`auth` / `invalidCredential` / `rateLimit` / `quota` / `empty` / `contextWindow` / `toolCall` / `maxTokens`) | `{ reason }` |
+| `internal` | 500 / 502 | `error.internal` | — |
+
+## Troubleshooting
+
+**Button missing / shortcut dead**
+Settings → 插件配置 → `prompt-enhance` section: is `enabled` true? Is the plugin installed (`dsh plugin --profile web list`) and `dsh web` restarted? Any apply error in the browser console?
+
+**"No model resolved for the enhancement"**
+Routing follows settings pair → session model → harness default; if all three are empty there is nothing to call. Pair `provider`/`model` in the settings, or send a message in the current session first so it carries a model route. Check the `agent-default-model` settings section.
+
+**Authentication failures**
+The `message` detail line carries the concrete cause (e.g. 401). Check the provider's API key in the harness credential store; quota/balance problems surface as the `quota` hint.
+
+**Rate-limited (429) right after a run of failures**
+Should not happen — the window counts successes only. If it still does, check whether several profiles/processes are mounted (each counts independently and stacks up) or whether `rateLimitPerMinute` is set too low.
+
+**Poor rewrites (fabrication, dropped requirements, broken formatting)**
+With `strategyMode` set to `replace-default`, a custom `systemPrompt` replaces the built-in strategy entirely — its hard rules (never fabricate, body-only output, mirror the input language) are not retained automatically; carry them into your own text or switch to `extend-default`.
+
+**Behavior changed after upgrading**
+Since 0.1.6 the host no longer sends Chinese primary-line copy: the browser localizes errors by code (`message` is now only a diagnostic detail), and the `rate-limit` window counts successful calls only. No configuration change needed.
+
 ## Security model
 
 The enhance route is served by your own dsh host and reachable **only from this machine**:
 
 - **Socket fence** — requests from non-loopback addresses are refused (`127.0.0.1` / `::1` only). Note this means *any local process* can call the route; it carries no user authentication.
 - **Host allowlist** — the route also validates the `Host` header against `localhost` / `127.0.0.1` / `[::1]`, which defeats DNS-rebinding (a rebound attacker domain keeps the loopback socket address but carries the attacker's hostname and is refused). Responses are `cache-control: no-store`.
-- **Abuse caps** — an `Origin` gate refuses browser calls from non-local pages, and the route enforces a concurrency cap (`maxConcurrent`, default 2) and a per-minute rate limit (`rateLimitPerMinute`, default 10), answering `429` beyond either.
+- **Abuse caps** — an `Origin` gate refuses browser calls from non-local pages, and the route enforces a concurrency cap (`maxConcurrent`, default 2) and a per-minute rate limit (`rateLimitPerMinute`, default 10), answering `429` beyond either. Both counters live **in a single host process's memory**: multiple processes, a cluster, or several profiles mounting the plugin each count independently, so the global cap would be exceeded — this plugin supports single-process deployments only and ships no shared rate-limit store.
 - **Proxy rejection** — requests carrying `X-Forwarded-For` / `Forwarded` headers are refused outright: those headers only exist when a proxy is in the path, which the trust model does not cover.
+- **Request-level timeouts are the host's and Node's job** — the plugin bounds the body size (including a Content-Length fast reject) and the per-call `timeoutMs`; connection-level timeouts (headers / request timeout / keep-alive) are server-level settings on the shared `http.Server` that a prefix route must not touch, so they fall to the host and Node's defaults (60 s headers / 300 s request).
 - **Not for reverse-proxy exposure** — if you put dsh web behind a proxy that listens on the LAN, external callers appear as loopback to the route and the fence is moot. Do not expose a proxied host without adding your own authentication at the proxy.
 - **Prompt-injection boundary** — the draft is framed between `<raw_prompt>` tags, literal closing tags inside the draft are neutralized, and the strategy prompt treats the framed text as pure data. This lowers the risk of simple tag-escape; prompt-based boundaries are best-effort, not a guarantee. Enhancements run with your own credentials and the result is only ever shown back to you.
 
@@ -177,7 +215,7 @@ src/
 ├── config.ts           schemastery schema + resolution (paired route validation)
 ├── prompts.ts          built-in strategy system prompt + <raw_prompt> framing
 ├── enhancer.ts         the ctx.llm auxiliary call (route resolution, deadline
-│                       racing, finish validation, code → message mapping)
+│                       racing, finish validation, structured errors + host render)
 ├── enhance-routes.ts   POST /prompt-enhance/enhance (loopback fence, body cap)
 ├── enhance-command.ts  /enhance slash command (host command registry)
 ├── loopback.ts         127.0.0.1/::1 fence for the route
@@ -191,7 +229,7 @@ src/
     ├── UndoBar         conversation.input.dock entry: restore affordance
     ├── ui-state.ts     external store shared by components (panel, undo, sessions)
     ├── enhance-client  fetch client with abort + typed errors
-    ├── undo-stack.ts   per-session LIFO (depth 3)
+    ├── undo-stack.ts   per-session LIFO (depth 3, global cap 60, LRU eviction)
     ├── shortcut.ts     pure combo parsing / matching
     ├── settings.ts     client mirror of the settings namespace
     ├── locales.ts      zh + en dictionaries (harness locale namespace)

package/README.zh-CN.md

@@ -28,7 +28,7 @@
 | ⚙️ **配置热生效** | 模型覆盖、温度、预算、系统提示词、快捷键,全部在 设置 → 插件配置 即改即用 |
 | 🛡️ **草稿安全** | 空输入、超长、仅图片、含命令块在本地拦截;上游失败映射为可读提示;失败绝不改动草稿 |
 
-![dsh web 中实机运行的预览面板:原文与增强结果并排,含模型信息与回填/复制操作](docs/evidence-prompt-enhance-panel.png)
+![dsh web 中实机运行的预览面板:原文与增强结果并排,含模型信息与回填/复制操作](https://raw.githubusercontent.com/rongxingda/dsh-prompt-enhance/main/docs/evidence-prompt-enhance-panel.png)
 
 ## 工作原理
 
@@ -99,7 +99,7 @@ dsh plugin --profile web remove dsh-prompt-enhance
 1. 先跑本地守卫:空草稿、超长草稿(**不自动截断**——截断会改变原意)、仅图片草稿、含命令或文件引用块的草稿,都会被拒绝并给出明确提示。引用块被拒是因为回填会破坏它们。
 2. 预览面板打开,显示可取消的加载动画。此时草稿原封不动——面板上写明了这一点。
 3. 结果阶段左右并排展示两份文本。**回填**把增强文本写回输入框并升起撤销条;**复制**进剪贴板;**取消**(`Esc` 或点击遮罩)丢弃一切。
-4. 撤销条停在输入框上方:一键恢复原文。回填后你继续输入,撤销条会安静地自我退位——过期的原文永远不会覆盖你更新的编辑。
+4. 撤销条停在输入框上方:一键恢复原文。回填后你继续输入,撤销条会安静地自我退位——过期的原文永远不会覆盖你更新的编辑。撤销记录**只存在于当前页面内存**(每个会话最多 3 条),刷新页面、重启 Web 或切换会话后即失效。
 
 **`/enhance <文本>`** —— 从斜杠菜单改写任意文本。结果在命令面板渲染(可复制),不进入会话历史与模型上下文。要增强**输入框里的草稿**请用按钮或快捷键——草稿在浏览器里。正在运行的 `/enhance` 取消跟随 harness 命令面板能力;客户端未提供取消入口时,调用会执行到完成或超时。
 
@@ -115,12 +115,13 @@ dsh plugin --profile web remove dsh-prompt-enhance
 | `provider` + `model` | 空 | 显式路由覆盖;必须**成对**填写(或都留空以跟随当前会话模型) |
 | `temperature` | `0.3` | 低温使改写更忠实于原意 |
 | `maxOutputTokens` | `2048` | 单次增强调用的输出 token 预算 |
-| `maxInputChars` | `12000` | 输入字数上限(按 Unicode 码点统计,一个 emoji 算一个字);超限**拒绝而不截断** |
+| `maxInputChars` | `12000` | 输入字符数上限(按 Unicode 字符统计,不是 token 数;一个 emoji 算一个字符);超限**拒绝而不截断** |
 | `timeoutMs` | `60000` | 单次调用的端到端超时 |
-| `systemPrompt` | 内置策略 | 想用自己的增强策略时整体替换 |
+| `systemPrompt` | 内置策略 | 自定义增强策略文本;与内置策略如何组合见 `strategyMode` |
+| `strategyMode` | `replace-default` | 自定义策略的组合方式:`replace-default` **整体替换**内置策略(向后兼容,但「不改变原意、不编造、只输出正文、镜像输入语言」等硬性约束不会自动保留,需自行写入);`extend-default` 把自定义文本**追加在内置策略之后**,硬性约束继续生效 |
 | `shortcut` | `ctrl+alt+e` | 快捷键规格(至少一个修饰键 + 单个字母/数字/功能键——纯裸键会被忽略,绝不会吞掉正常打字);留空禁用 |
-| `maxConcurrent` | `2` | 宿主侧并发上限;超出的请求返回 `429` |
-| `rateLimitPerMinute` | `10` | 每分钟滑动窗口限流;超出的请求返回 `429` |
+| `maxConcurrent` | `2` | **单个 Host 进程内**的并发上限;超出的请求返回 `429`(`concurrency-limit`)。浏览器 UI 因单预览面板天然只允许 1 个在途请求,此上限主要保护 `/enhance` 命令入口与多客户端调用 |
+| `rateLimitPerMinute` | `10` | **单个 Host 进程内**的每分钟滑动窗口限流,只统计**成功完成**的增强——失败(超时 / 上游错误 / 取消)不消耗窗口,连续失败不会把自己限流;超出的请求返回 `429`(`rate-limit`),响应附 `Retry-After` 秒数 |
 | `provider` + `model` 取值 | — | 与 harness 设置一致:设置文件 `llm-pi-ai.providers` 下的键就是 provider(如 `zhipu`、`muyuu`),其 `models[].id` 就是 model(如 `glm-5.3-flash`)。示例:`provider: zhipu` + `model: glm-5.3-flash` |
 
 **模型路由优先级:** 设置成对覆盖 → 当前会话请求头中记录的路由 → 全局默认模型(`agent-default-model`)。三者都指不出路由时(例如全新会话且无默认模型),插件给出可操作的报错而不是乱猜。
@@ -138,17 +139,19 @@ dsh plugin --profile web remove dsh-prompt-enhance
 
 硬性规则:精确保留原意(不删除、不篡改、不与用户信息冲突);绝不编造——未知细节插入显式占位符如 `(待补充:…)` / `(TBD: …)`;保持任务范围不变;**只输出**改写后的正文(无解释、无围栏、无客套);镜像输入语言;长度约为原文 1–3 倍;已经成形的提示词只做轻度润色而不注水。
 
-在设置中填写 `systemPrompt` 即可整体替换为你的策略。
+在设置中填写 `systemPrompt` 即可自定义策略,组合方式由 `strategyMode` 决定:默认 `replace-default` 是**完全替换**——内置策略中的安全约束(不改变原意、不编造、只输出正文、镜像输入语言、原文视为纯数据)不会自动保留,自定义策略需要自行包含这些约束;改为 `extend-default` 则把你的文本追加在内置策略之后,硬性约束继续生效。
 
 ## 异常与边界
 
 | 情形 | 行为 |
 |---|---|
 | 空 / 纯空白 / 零宽字符草稿 | 本地拒绝:「输入框为空」 |
-| 草稿超过 `maxInputChars` | 本地与路由双重拒绝并给出精确字数;**不自动截断** |
+| 草稿超过 `maxInputChars` | 本地与路由双重拒绝并给出精确字符数;**不自动截断** |
 | 只附加了图片没有文本 | 拒绝:仅支持文本 |
 | 草稿含命令 / 文件引用块 | 拒绝:回填会破坏引用块 |
 | 提交中 / 占用阶段 / 已有请求在途 | 拒绝并提示「稍后再试」 |
+| 每分钟次数超限 | `429`（`rate-limit`），响应携带 `Retry-After` 秒数，宿主消息给出精确等待时长;窗口只统计成功完成的调用——失败不消耗窗口 |
+| 并发已满 | `429`（`concurrency-limit`），无 `Retry-After`——空出的时机不可预测，等一个在途调用结束即可 |
 | 上游模型失败 | 稳定错误码映射为可读提示:`AUTH` → 检查 API Key、`RATE_LIMIT` → 稍后重试、`QUOTA_EXCEEDED` → 检查余额、`CONTEXT_WINDOW_EXCEEDED` → 精简输入、`NO_ADAPTER`/未配置 → 先配置模型 |
 | 输出达到 `maxOutputTokens` | 拒绝并提示调大上限或精简原文 |
 | 模型返回空 / 围栏包裹 / 工具调用 | 规范化(剥围栏)或拒绝;可重试 |
@@ -158,14 +161,49 @@ dsh plugin --profile web remove dsh-prompt-enhance
 
 所有失败只出现在插件自己的面板里;失败的调用绝不修改输入框草稿,手动输入不受任何干扰。
 
+## 错误码与本地化
+
+宿主路由返回的结构化错误形如 `{ code, message?, params? }`:主文案由浏览器按 `code` + `params` 从当前语言字典渲染,`message` 只是可选的诊断细节(如模型服务商的原始报错、配置错误原文),原样展示在主文案下方。`/enhance` 命令平面没有 locale 字典,由宿主侧的同源渲染函数直接产出中文文本。
+
+| 错误码 | HTTP 状态 | 浏览器主文案(字典键) | 参数 |
+|---|---|---|---|
+| `rejected` | 403 / 413 / 415 / 422 | `error.rejected` 通用;携带 `{ count, max }` 时复用 `error.tooLong` | 超长输入: `{ count, max }` |
+| `rate-limit` | 429 | `error.rateLimit` | `{ limit, retryAfterSeconds }` |
+| `concurrency-limit` | 429 | `error.concurrencyLimit` | `{ max }` |
+| `timeout` | 504 | `error.timeout` | `{ seconds }` |
+| `unconfigured` | 409 | `error.unconfigured` | — |
+| `upstream` | 502 | `error.upstream` 通用;携带 `reason` 时用 `error.upstream.{reason}`(如 `auth` / `quota` / `rateLimit` / `empty` / `contextWindow` / `toolCall` / `maxTokens` / `invalidCredential`) | `{ reason }` |
+| `internal` | 500 / 502 | `error.internal` | — |
+
+## 故障排查
+
+**按钮不出现 / 快捷键无响应**
+设置 → 插件配置 → `prompt-enhance` 节的 `enabled` 是否为 `true`;插件是否成功安装(`dsh plugin --profile web list`)并重启了 `dsh web`;浏览器控制台是否有插件应用报错。
+
+**「尚未确定增强用的模型」**
+插件遵循设置成对覆盖 → 当前会话模型 → 全局默认模型的路由优先级,三者都为空时无法调用。在设置中成对填写 `provider`/`model`,或先在当前会话发一条消息让会话带上模型路由。检查 `agent-default-model` 设置节是否配置。
+
+**「鉴权失败」**
+`message` 细节行会带出具体原因(如 401)。检查对应 provider 在 harness 凭据存储中的 API Key;配额/余额问题对应 `quota` 提示。
+
+**连续失败后立刻被限流(429)**
+不应发生——限流窗口只统计成功调用,失败不消耗窗口。若仍遇到,确认同时挂载了多个 profile / 多进程(各自独立计数会叠加),或 `rateLimitPerMinute` 配置过低。
+
+**结果不理想(编造、丢要求、格式乱)**
+`strategyMode` 为 `replace-default` 时自定义 `systemPrompt` 会整体替换内置策略——内置的「不编造、只输出正文、镜像输入语言」等硬性约束不会自动保留,请自行写入;或改用 `extend-default`。
+
+**升级后行为变了**
+0.1.6 起宿主错误不再携带中文主文案,浏览器端按错误码本地化渲染(非中文界面不再混入中文);`rate-limit` 窗口改为只统计成功调用。升级后无需改配置。
+
 ## 安全模型
 
 增强路由由你自己的 dsh 宿主提供服务,**仅限本机访问**:
 
 - **套接字栅栏** —— 非回环地址(`127.0.0.1` / `::1` 之外)的请求一律拒绝。注意这意味着*本机任意进程*都可以调用该路由;路由本身不带用户鉴权。
 - **Host 白名单** —— 路由同时校验 `Host` 头是否为 `localhost` / `127.0.0.1` / `[::1]`,可防御 DNS 重绑定(重绑定的攻击者域名虽然落在回环套接字上,但携带的是攻击者主机名,会被拒绝)。响应均为 `cache-control: no-store`。
-- **滥用上限** —— `Origin` 门拒绝来自非本地页面的浏览器调用;路由还执行并发上限(`maxConcurrent`,默认 2)与每分钟限流(`rateLimitPerMinute`,默认 10),超限返回 `429`。
+- **滥用上限** —— `Origin` 门拒绝来自非本地页面的浏览器调用;路由还执行并发上限(`maxConcurrent`,默认 2)与每分钟限流(`rateLimitPerMinute`,默认 10),超限返回 `429`。这两项计数保存在**单个宿主进程的内存中**——多进程、cluster 或多个 profile 同时挂载时各自独立计数,全局上限会被突破;本插件只支持单进程部署,也不提供共享限流存储。
 - **代理请求拒绝** —— 携带 `X-Forwarded-For` / `Forwarded` 头的请求直接拒绝:这些头只会在链路上存在代理时出现,而代理场景不在本路由的信任模型内。
+- **请求级超时由宿主与 Node 兜底** —— 插件层只限制 body 字节数(含 Content-Length 快速拒绝)与单次模型调用的 `timeoutMs`;连接级超时(headers timeout / request timeout / keep-alive)是共享 `http.Server` 的 server 级配置,插件不覆盖,由宿主与 Node 默认值(headers 60 秒 / request 300 秒)兜底。
 - **不要经反向代理暴露** —— 若把 dsh web 放在监听局域网的代理后面,外部调用者在路由看来就是回环地址,栅栏形同虚设。除非在代理层自行加鉴权,否则不要暴露。
 - **提示注入边界** —— 草稿被框在 `<raw_prompt>` 标签之间,草稿内的字面闭合标签会被中和,策略提示词把框内文本视为纯数据——降低简单标签逃逸的风险;基于提示词的边界是尽力而为,并非完整防护。增强只使用你自己的凭据,结果也只回显给你本人。
 
@@ -177,7 +215,7 @@ src/
 ├── config.ts           schemastery 模式 + 解析校验(成对路由)
 ├── prompts.ts          内置策略系统提示词 + <raw_prompt> 框架
 ├── enhancer.ts         ctx.llm 辅助调用(路由解析、超时竞速、finish 校验、
-│                       错误码 → 文案映射)
+│                       结构化错误 code+params + 宿主侧渲染)
 ├── enhance-routes.ts   POST /prompt-enhance/enhance(回环栅栏、限长)
 ├── enhance-command.ts  /enhance 斜杠命令(宿主命令注册表)
 ├── loopback.ts         路由的 127.0.0.1/::1 栅栏
@@ -190,7 +228,7 @@ src/
     ├── UndoBar         conversation.input.dock 条目:恢复入口
     ├── ui-state.ts     组件共享的外部 store(面板、撤销、会话注册表)
     ├── enhance-client  fetch 客户端(可中止 + 类型化错误)
-    ├── undo-stack.ts   按会话的 LIFO 栈(深度 3)
+    ├── undo-stack.ts   按会话的 LIFO 栈(每会话深度 3,全局 60 条,LRU 淘汰)
     ├── shortcut.ts     纯函数的组合键解析 / 匹配
     ├── settings.ts     设置命名空间的客户端镜像
     ├── locales.ts      中英文字典(harness locale 命名空间)
@@ -230,7 +268,7 @@ CI 在每次 push/PR 上运行类型检查 + 测试 + 构建,并在 `lib/` 与
 回填通过 `inputActions.setDraft` 写纯文本,会破坏引用块。先移除它们,增强后再插回。
 
 **为什么不自动截断超长输入?**
-截断会无声地改变你的原意——插件宁可拒绝并给出精确字数。
+截断会无声地改变你的原意——插件宁可拒绝并给出精确字符数。
 
 **能固定用某个模型吗?**
 在设置中**成对**填写 `provider` 与 `model`(例如 harness 已配置的 provider 路由)。都留空则跟随当前会话模型。

package/lib/client.js

@@ -39,12 +39,15 @@ var import_react_dom = require("react-dom");
 
 // src/shared/validate.ts
 var INVISIBLE_CHARS = /[\u200B-\u200D\uFEFF\u202A-\u202E\u2066-\u2069]/g;
+function countText(text) {
+  return [...text].length;
+}
 function checkInputText(text, maxChars) {
   const stripped = text.replace(INVISIBLE_CHARS, "");
   if (stripped.trim().length === 0) {
     return { ok: false, code: "empty" };
   }
-  const codePoints = [...text].length;
+  const codePoints = countText(text);
   if (codePoints > maxChars) {
     return { ok: false, code: "too-long", count: codePoints, max: maxChars };
   }
@@ -61,12 +64,22 @@ var EnhanceClientError = class extends Error {
     this.detail = detail;
   }
 };
-var KNOWN_ERROR_CODES = /* @__PURE__ */ new Set(["rejected", "rate", "timeout", "upstream", "unconfigured", "internal"]);
+var KNOWN_ERROR_CODES = /* @__PURE__ */ new Set([
+  "rejected",
+  "rate-limit",
+  "concurrency-limit",
+  "timeout",
+  "upstream",
+  "unconfigured",
+  "internal"
+]);
 function parseError(value) {
   const record = value;
-  if (record !== null && typeof record === "object" && typeof record.message === "string" && record.message !== "") {
+  if (record !== null && typeof record === "object") {
     const code = typeof record.code === "string" && KNOWN_ERROR_CODES.has(record.code) ? record.code : "internal";
-    return { code, message: record.message };
+    const message = typeof record.message === "string" && record.message !== "" ? record.message : void 0;
+    const params = record.params !== null && typeof record.params === "object" ? record.params : void 0;
+    return { code, ...message !== void 0 ? { message } : {}, ...params !== void 0 ? { params } : {} };
   }
   return { code: "internal", message: "\u5BBF\u4E3B\u670D\u52A1\u8FD4\u56DE\u5F02\u5E38\u3002" };
 }
@@ -114,16 +127,117 @@ async function requestEnhance(body, signal) {
 
 // src/client/ResultPanel.tsx
 var import_react = require("react");
+
+// src/client/locales.ts
+var zh = {
+  "button.title": "\u63D0\u793A\u8BCD\u589E\u5F3A\uFF08\u91CD\u5199\u4E3A\u7ED3\u6784\u5316\u63D0\u793A\u8BCD\uFF09",
+  "button.busy": "\u6B63\u5728\u589E\u5F3A\u2026",
+  "panel.title": "\u63D0\u793A\u8BCD\u589E\u5F3A",
+  "panel.loading": "\u6B63\u5728\u589E\u5F3A\uFF0C\u7A0D\u5019\u2026",
+  "panel.loading.hint": "\u539F\u6587\u4FDD\u7559\u5728\u8F93\u5165\u6846\u4E2D\uFF0C\u4E0D\u4F1A\u88AB\u52A8\u4FEE\u6539\u3002",
+  "panel.cancel": "\u53D6\u6D88",
+  "panel.original": "\u539F\u59CB\u63D0\u793A\u8BCD",
+  "panel.enhanced": "\u589E\u5F3A\u7ED3\u679C",
+  "panel.apply": "\u56DE\u586B\u5230\u8F93\u5165\u6846",
+  "panel.copy": "\u590D\u5236\u7ED3\u679C",
+  "panel.copied": "\u5DF2\u590D\u5236",
+  "panel.copyFailed": "\u590D\u5236\u5931\u8D25",
+  "panel.close": "\u5173\u95ED",
+  "panel.retry": "\u91CD\u8BD5",
+  "panel.stale.warn": "\u8349\u7A3F\u5728\u589E\u5F3A\u671F\u95F4\u6709\u6539\u52A8\u2014\u2014\u589E\u5F3A\u7ED3\u679C\u57FA\u4E8E\u589E\u5F3A\u524D\u7684\u6587\u672C\u3002\u56DE\u586B\u5C06\u8986\u76D6\u4F60\u7684\u6700\u65B0\u7F16\u8F91(\u64A4\u9500\u53EF\u6062\u590D\u56DE\u586B\u524D\u7684\u8349\u7A3F)\u3002",
+  "panel.model": "\u6A21\u578B\uFF1A{provider} / {model}",
+  "panel.elapsed": "\u8017\u65F6 {ms}",
+  "undo.applied": "\u5DF2\u7528\u589E\u5F3A\u7ED3\u679C\u66FF\u6362\u539F\u63D0\u793A\u8BCD",
+  "undo.undo": "\u64A4\u9500",
+  "undo.dismiss": "\u5173\u95ED\u63D0\u793A",
+  "error.empty": "\u8F93\u5165\u6846\u4E3A\u7A7A\uFF0C\u8BF7\u5148\u8F93\u5165\u8981\u589E\u5F3A\u7684\u63D0\u793A\u8BCD\u3002",
+  "error.tooLong": "\u5185\u5BB9\u5171 {count} \u4E2A\u5B57\u7B26\uFF0C\u8D85\u8FC7 {max} \u4E2A\u5B57\u7B26\u4E0A\u9650\uFF08\u6309 Unicode \u5B57\u7B26\u6570\u7EDF\u8BA1\uFF0C\u4E0D\u662F token \u6570\uFF09\u3002\u4E3A\u907F\u514D\u6539\u53D8\u539F\u610F\u4E0D\u4F1A\u81EA\u52A8\u622A\u65AD\uFF0C\u8BF7\u7CBE\u7B80\u540E\u518D\u8BD5\u3002",
+  "error.imagesOnly": "\u5F53\u524D\u53EA\u9644\u52A0\u4E86\u56FE\u7247\uFF0C\u4EC5\u652F\u6301\u589E\u5F3A\u6587\u672C\u5185\u5BB9\u3002",
+  "error.occurrences": "\u8F93\u5165\u5185\u5BB9\u5305\u542B\u547D\u4EE4\u6216\u6587\u4EF6\u5F15\u7528\uFF0C\u6682\u4E0D\u652F\u6301\u589E\u5F3A\uFF1B\u8BF7\u5148\u79FB\u9664\u540E\u518D\u8BD5\u3002",
+  "error.phase": "\u5F53\u524D\u8F93\u5165\u6B63\u88AB\u5360\u7528\uFF08\u63D0\u4EA4\u4E2D\uFF09\uFF0C\u8BF7\u7A0D\u540E\u518D\u8BD5\u3002",
+  "error.disabled": "\u63D0\u793A\u8BCD\u589E\u5F3A\u5DF2\u5728\u63D2\u4EF6\u8BBE\u7F6E\u4E2D\u5173\u95ED\u3002",
+  "error.rejected": "\u589E\u5F3A\u8BF7\u6C42\u88AB\u62D2\u7EDD\u3002",
+  "error.timeout": "\u589E\u5F3A\u8D85\u65F6\uFF08{seconds} \u79D2\uFF09\uFF0C\u8BF7\u91CD\u8BD5\uFF1B\u539F\u8F93\u5165\u672A\u6539\u52A8\u3002",
+  "error.unconfigured": "\u5C1A\u672A\u786E\u5B9A\u589E\u5F3A\u7528\u7684\u6A21\u578B\uFF1A\u8BF7\u5728\u63D2\u4EF6\u8BBE\u7F6E\u4E2D\u6210\u5BF9\u586B\u5199 provider/model\uFF0C\u6216\u5148\u5728\u5F53\u524D\u4F1A\u8BDD\u53D1\u9001\u4E00\u6761\u6D88\u606F\uFF08\u5C06\u8DDF\u968F\u4F1A\u8BDD\u6A21\u578B\uFF09\u3002",
+  "error.upstream": "\u6A21\u578B\u670D\u52A1\u8FD4\u56DE\u9519\u8BEF\uFF0C\u8BF7\u91CD\u8BD5\uFF1B\u539F\u8F93\u5165\u672A\u6539\u52A8\u3002",
+  "error.internal": "\u589E\u5F3A\u5931\u8D25\uFF0C\u8BF7\u91CD\u8BD5\uFF1B\u539F\u8F93\u5165\u672A\u6539\u52A8\u3002",
+  "error.rateLimit": "\u8BF7\u6C42\u8FC7\u4E8E\u9891\u7E41\uFF1A\u6BCF\u5206\u949F\u6700\u591A {limit} \u6B21\u589E\u5F3A\uFF0C\u8BF7\u5728 {retryAfterSeconds} \u79D2\u540E\u518D\u8BD5\u3002",
+  "error.concurrencyLimit": "\u5DF2\u6709 {max} \u4E2A\u589E\u5F3A\u6B63\u5728\u8FDB\u884C\uFF0C\u8BF7\u7B49\u5176\u4E2D\u4E00\u4E2A\u5B8C\u6210\u540E\u518D\u8BD5\u3002",
+  "error.upstream.auth": "\u9274\u6743\u5931\u8D25\uFF1A\u8BF7\u68C0\u67E5\u8BE5 provider \u7684 API Key \u914D\u7F6E\u3002",
+  "error.upstream.invalidCredential": "\u9274\u6743\u5931\u8D25\uFF1A\u5B58\u50A8\u7684 API Key \u4E0D\u53EF\u7528\uFF0C\u8BF7\u4FEE\u6B63\u540E\u91CD\u8BD5\u3002",
+  "error.upstream.rateLimit": "\u6A21\u578B\u670D\u52A1\u9650\u6D41\uFF0C\u8BF7\u7A0D\u540E\u91CD\u8BD5\u3002",
+  "error.upstream.quota": "\u6A21\u578B\u670D\u52A1\u914D\u989D/\u4F59\u989D\u4E0D\u8DB3\uFF0C\u8BF7\u68C0\u67E5\u8D26\u6237\u3002",
+  "error.upstream.empty": "\u6A21\u578B\u8FD4\u56DE\u4E86\u7A7A\u54CD\u5E94\uFF0C\u8BF7\u91CD\u8BD5\u3002",
+  "error.upstream.contextWindow": "\u8F93\u5165\u8D85\u51FA\u6A21\u578B\u4E0A\u4E0B\u6587\u7A97\u53E3\uFF0C\u8BF7\u7CBE\u7B80\u539F\u6587\u6216\u66F4\u6362\u6A21\u578B\u3002",
+  "error.upstream.toolCall": "\u6A21\u578B\u8BF7\u6C42\u4E86\u5DE5\u5177\u8C03\u7528\uFF0C\u63D0\u793A\u8BCD\u589E\u5F3A\u53EA\u9700\u8981\u7EAF\u6587\u672C\uFF1B\u8BF7\u66F4\u6362\u6A21\u578B\u540E\u91CD\u8BD5\u3002",
+  "error.upstream.maxTokens": "\u91CD\u5199\u7ED3\u679C\u8FBE\u5230\u8F93\u51FA\u4E0A\u9650\uFF08maxOutputTokens\uFF09\uFF0C\u8BF7\u5728\u8BBE\u7F6E\u4E2D\u8C03\u5927\u4E0A\u9650\u6216\u7CBE\u7B80\u539F\u6587\u540E\u91CD\u8BD5\u3002"
+};
+var en = {
+  "button.title": "Enhance prompt (rewrite into a structured prompt)",
+  "button.busy": "Enhancing\u2026",
+  "panel.title": "Prompt Enhance",
+  "panel.loading": "Enhancing, please wait\u2026",
+  "panel.loading.hint": "Your draft stays untouched until you apply the result.",
+  "panel.cancel": "Cancel",
+  "panel.original": "Original prompt",
+  "panel.enhanced": "Enhanced result",
+  "panel.apply": "Fill into input box",
+  "panel.copy": "Copy result",
+  "panel.copied": "Copied",
+  "panel.copyFailed": "Copy failed",
+  "panel.close": "Close",
+  "panel.retry": "Retry",
+  "panel.stale.warn": "The draft changed while enhancing \u2014 the result is based on the pre-enhance text. Applying will overwrite your latest edits (undo restores the draft as it was before applying).",
+  "panel.model": "Model: {provider} / {model}",
+  "panel.elapsed": "{ms} elapsed",
+  "undo.applied": "Original prompt replaced by the enhanced version",
+  "undo.undo": "Undo",
+  "undo.dismiss": "Dismiss",
+  "error.empty": "The input box is empty \u2014 type a prompt to enhance first.",
+  "error.tooLong": "The draft is {count} characters, above the {max} cap (Unicode characters, not tokens). It is never auto-truncated (that would change your meaning) \u2014 please shorten it.",
+  "error.imagesOnly": "Only images are attached; prompt enhance supports text only.",
+  "error.occurrences": "The draft contains commands or file references, which are not supported yet \u2014 remove them first.",
+  "error.phase": "The input box is busy (submitting) \u2014 try again in a moment.",
+  "error.disabled": "Prompt enhance is disabled in the plugin settings.",
+  "error.rejected": "The enhance request was rejected.",
+  "error.timeout": "The enhancement timed out ({seconds}s). Retry; your draft is untouched.",
+  "error.unconfigured": "No model resolved for the enhancement: pair provider/model in the plugin settings, or send a message in the current session first (the enhancement will follow the session model).",
+  "error.upstream": "The model provider returned an error. Retry; your draft is untouched.",
+  "error.internal": "Enhancement failed. Retry; your draft is untouched.",
+  "error.rateLimit": "Too many requests \u2014 at most {limit} enhancements per minute; retry in {retryAfterSeconds} seconds.",
+  "error.concurrencyLimit": "{max} enhancements are already running \u2014 wait for one to finish, then retry.",
+  "error.upstream.auth": "Authentication failed \u2014 check the API key configured for this provider.",
+  "error.upstream.invalidCredential": "Authentication failed \u2014 the stored API key is invalid; fix it and retry.",
+  "error.upstream.rateLimit": "The model provider is rate-limiting; retry later.",
+  "error.upstream.quota": "The model provider reports a quota/balance issue \u2014 check your account.",
+  "error.upstream.empty": "The model returned an empty response; retry.",
+  "error.upstream.contextWindow": "The input exceeds the model context window \u2014 shorten it or switch models.",
+  "error.upstream.toolCall": "The model requested tool calls, but prompt enhance needs plain text \u2014 switch models and retry.",
+  "error.upstream.maxTokens": "The rewrite hit the output cap (maxOutputTokens) \u2014 raise it in settings or shorten the draft."
+};
+var dictionaries = { zh, en };
+var NS = "prompt-enhance";
+
+// src/client/ResultPanel.tsx
 var import_jsx_runtime = require("react/jsx-runtime");
 var retryable = (code) => code === "upstream" || code === "timeout" || code === "internal";
-function localizedErrorMessage(t, code) {
+function localizedErrorMessage(t, code, params) {
+  if (code === "rejected" && typeof params?.count === "number" && typeof params?.max === "number") {
+    return t("error.tooLong", { count: params.count, max: params.max });
+  }
+  if (code === "upstream" && typeof params?.reason === "string") {
+    const specific = `error.upstream.${params.reason}`;
+    if (specific in zh) return t(specific);
+  }
   switch (code) {
     case "rejected":
       return t("error.rejected");
-    case "rate":
-      return t("error.rate");
+    case "rate-limit":
+      return t("error.rateLimit", params);
+    case "concurrency-limit":
+      return t("error.concurrencyLimit", params);
     case "timeout":
-      return t("error.timeout");
+      return t("error.timeout", params);
     case "unconfigured":
       return t("error.unconfigured");
     case "upstream":
@@ -223,10 +337,13 @@ function ResultPanel(props) {
           /* @__PURE__ */ (0, import_jsx_runtime.jsx)("div", { children: t("panel.loading") }),
           /* @__PURE__ */ (0, import_jsx_runtime.jsx)("div", { className: "dsh-pe-hint", children: t("panel.loading.hint") })
         ] }),
-        state.phase === "error" && state.error !== void 0 && /* @__PURE__ */ (0, import_jsx_runtime.jsxs)("div", { className: "dsh-pe-error", children: [
-          /* @__PURE__ */ (0, import_jsx_runtime.jsx)("div", { children: state.error.localized ?? localizedErrorMessage(t, state.error.code) }),
-          state.error.localized !== state.error.message && state.error.message !== "" && /* @__PURE__ */ (0, import_jsx_runtime.jsx)("div", { className: "dsh-pe-error-detail", children: state.error.message })
-        ] }),
+        state.phase === "error" && state.error !== void 0 && /* @__PURE__ */ (0, import_jsx_runtime.jsx)("div", { className: "dsh-pe-error", children: state.error.localized !== void 0 ? (
+          // Client-side guard failures carry their own pre-localized copy.
+          /* @__PURE__ */ (0, import_jsx_runtime.jsx)("div", { children: state.error.localized })
+        ) : /* @__PURE__ */ (0, import_jsx_runtime.jsxs)(import_jsx_runtime.Fragment, { children: [
+          /* @__PURE__ */ (0, import_jsx_runtime.jsx)("div", { children: localizedErrorMessage(t, state.error.code, state.error.params) }),
+          state.error.message !== void 0 && state.error.message !== "" && /* @__PURE__ */ (0, import_jsx_runtime.jsx)("div", { className: "dsh-pe-error-detail", children: state.error.message })
+        ] }) }),
         state.phase === "result" && state.result !== void 0 && /* @__PURE__ */ (0, import_jsx_runtime.jsxs)("div", { className: "dsh-pe-body", children: [
           /* @__PURE__ */ (0, import_jsx_runtime.jsxs)("div", { className: "dsh-pe-col", children: [
             /* @__PURE__ */ (0, import_jsx_runtime.jsx)("div", { className: "dsh-pe-col-title", children: t("panel.original") }),
@@ -256,14 +373,25 @@ function ResultPanel(props) {
 }
 
 // src/client/undo-stack.ts
-function createUndoStack(maxDepth = 3) {
+function totalOf(stacks) {
+  let n = 0;
+  for (const stack of stacks.values()) n += stack.length;
+  return n;
+}
+function createUndoStack(maxDepth = 3, maxTotalEntries = 60) {
   const stacks = /* @__PURE__ */ new Map();
   return {
     push(sessionId, entry) {
       const stack = stacks.get(sessionId) ?? [];
       stack.push(entry);
       while (stack.length > maxDepth) stack.shift();
+      stacks.delete(sessionId);
       stacks.set(sessionId, stack);
+      while (totalOf(stacks) > maxTotalEntries) {
+        const oldest = stacks.keys().next().value;
+        if (oldest === void 0) break;
+        stacks.delete(oldest);
+      }
     },
     peek(sessionId) {
       const stack = stacks.get(sessionId);
@@ -543,78 +671,6 @@ function UndoBar(props) {
   ] });
 }
 
-// src/client/locales.ts
-var zh = {
-  "button.title": "\u63D0\u793A\u8BCD\u589E\u5F3A\uFF08\u91CD\u5199\u4E3A\u7ED3\u6784\u5316\u63D0\u793A\u8BCD\uFF09",
-  "button.busy": "\u6B63\u5728\u589E\u5F3A\u2026",
-  "panel.title": "\u63D0\u793A\u8BCD\u589E\u5F3A",
-  "panel.loading": "\u6B63\u5728\u589E\u5F3A\uFF0C\u7A0D\u5019\u2026",
-  "panel.loading.hint": "\u539F\u6587\u4FDD\u7559\u5728\u8F93\u5165\u6846\u4E2D\uFF0C\u4E0D\u4F1A\u88AB\u52A8\u4FEE\u6539\u3002",
-  "panel.cancel": "\u53D6\u6D88",
-  "panel.original": "\u539F\u59CB\u63D0\u793A\u8BCD",
-  "panel.enhanced": "\u589E\u5F3A\u7ED3\u679C",
-  "panel.apply": "\u56DE\u586B\u5230\u8F93\u5165\u6846",
-  "panel.copy": "\u590D\u5236\u7ED3\u679C",
-  "panel.copied": "\u5DF2\u590D\u5236",
-  "panel.copyFailed": "\u590D\u5236\u5931\u8D25",
-  "panel.close": "\u5173\u95ED",
-  "panel.retry": "\u91CD\u8BD5",
-  "panel.stale.warn": "\u8349\u7A3F\u5728\u589E\u5F3A\u671F\u95F4\u6709\u6539\u52A8\u2014\u2014\u589E\u5F3A\u7ED3\u679C\u57FA\u4E8E\u589E\u5F3A\u524D\u7684\u6587\u672C\u3002\u56DE\u586B\u5C06\u8986\u76D6\u4F60\u7684\u6700\u65B0\u7F16\u8F91(\u64A4\u9500\u53EF\u6062\u590D\u56DE\u586B\u524D\u7684\u8349\u7A3F)\u3002",
-  "panel.model": "\u6A21\u578B\uFF1A{provider} / {model}",
-  "panel.elapsed": "\u8017\u65F6 {ms}",
-  "undo.applied": "\u5DF2\u7528\u589E\u5F3A\u7ED3\u679C\u66FF\u6362\u539F\u63D0\u793A\u8BCD",
-  "undo.undo": "\u64A4\u9500",
-  "undo.dismiss": "\u5173\u95ED\u63D0\u793A",
-  "error.empty": "\u8F93\u5165\u6846\u4E3A\u7A7A\uFF0C\u8BF7\u5148\u8F93\u5165\u8981\u589E\u5F3A\u7684\u63D0\u793A\u8BCD\u3002",
-  "error.tooLong": "\u5185\u5BB9\u5171 {count} \u5B57\uFF0C\u8D85\u8FC7 {max} \u5B57\u4E0A\u9650\u3002\u4E3A\u907F\u514D\u6539\u53D8\u539F\u610F\u4E0D\u4F1A\u81EA\u52A8\u622A\u65AD\uFF0C\u8BF7\u7CBE\u7B80\u540E\u518D\u8BD5\u3002",
-  "error.imagesOnly": "\u5F53\u524D\u53EA\u9644\u52A0\u4E86\u56FE\u7247\uFF0C\u4EC5\u652F\u6301\u589E\u5F3A\u6587\u672C\u5185\u5BB9\u3002",
-  "error.occurrences": "\u8F93\u5165\u5185\u5BB9\u5305\u542B\u547D\u4EE4\u6216\u6587\u4EF6\u5F15\u7528\uFF0C\u6682\u4E0D\u652F\u6301\u589E\u5F3A\uFF1B\u8BF7\u5148\u79FB\u9664\u540E\u518D\u8BD5\u3002",
-  "error.phase": "\u5F53\u524D\u8F93\u5165\u6B63\u88AB\u5360\u7528\uFF08\u63D0\u4EA4\u4E2D\uFF09\uFF0C\u8BF7\u7A0D\u540E\u518D\u8BD5\u3002",
-  "error.disabled": "\u63D0\u793A\u8BCD\u589E\u5F3A\u5DF2\u5728\u63D2\u4EF6\u8BBE\u7F6E\u4E2D\u5173\u95ED\u3002",
-  "error.rejected": "\u589E\u5F3A\u8BF7\u6C42\u88AB\u62D2\u7EDD\u3002",
-  "error.timeout": "\u589E\u5F3A\u8D85\u65F6\uFF0C\u8BF7\u91CD\u8BD5\uFF1B\u539F\u8F93\u5165\u672A\u6539\u52A8\u3002",
-  "error.unconfigured": "\u5C1A\u672A\u786E\u5B9A\u589E\u5F3A\u7528\u7684\u6A21\u578B\uFF1A\u8BF7\u5728\u63D2\u4EF6\u8BBE\u7F6E\u4E2D\u6210\u5BF9\u586B\u5199 provider/model\u3002",
-  "error.upstream": "\u6A21\u578B\u670D\u52A1\u8FD4\u56DE\u9519\u8BEF\uFF0C\u8BF7\u91CD\u8BD5\uFF1B\u539F\u8F93\u5165\u672A\u6539\u52A8\u3002",
-  "error.internal": "\u589E\u5F3A\u5931\u8D25\uFF0C\u8BF7\u91CD\u8BD5\uFF1B\u539F\u8F93\u5165\u672A\u6539\u52A8\u3002",
-  "error.rate": "\u8BF7\u6C42\u8FC7\u4E8E\u9891\u7E41\u6216\u5E76\u53D1\u5DF2\u6EE1\uFF0C\u8BF7\u7A0D\u540E\u518D\u8BD5\u3002"
-};
-var en = {
-  "button.title": "Enhance prompt (rewrite into a structured prompt)",
-  "button.busy": "Enhancing\u2026",
-  "panel.title": "Prompt Enhance",
-  "panel.loading": "Enhancing, please wait\u2026",
-  "panel.loading.hint": "Your draft stays untouched until you apply the result.",
-  "panel.cancel": "Cancel",
-  "panel.original": "Original prompt",
-  "panel.enhanced": "Enhanced result",
-  "panel.apply": "Fill into input box",
-  "panel.copy": "Copy result",
-  "panel.copied": "Copied",
-  "panel.copyFailed": "Copy failed",
-  "panel.close": "Close",
-  "panel.retry": "Retry",
-  "panel.stale.warn": "The draft changed while enhancing \u2014 the result is based on the pre-enhance text. Applying will overwrite your latest edits (undo restores the draft as it was before applying).",
-  "panel.model": "Model: {provider} / {model}",
-  "panel.elapsed": "{ms} elapsed",
-  "undo.applied": "Original prompt replaced by the enhanced version",
-  "undo.undo": "Undo",
-  "undo.dismiss": "Dismiss",
-  "error.empty": "The input box is empty \u2014 type a prompt to enhance first.",
-  "error.tooLong": "The draft is {count} characters, above the {max} cap. It is never auto-truncated (that would change your meaning) \u2014 please shorten it.",
-  "error.imagesOnly": "Only images are attached; prompt enhance supports text only.",
-  "error.occurrences": "The draft contains commands or file references, which are not supported yet \u2014 remove them first.",
-  "error.phase": "The input box is busy (submitting) \u2014 try again in a moment.",
-  "error.disabled": "Prompt enhance is disabled in the plugin settings.",
-  "error.rejected": "The enhance request was rejected.",
-  "error.timeout": "The enhancement timed out. Retry; your draft is untouched.",
-  "error.unconfigured": "No model resolved for the enhancement: pair provider/model in the plugin settings.",
-  "error.upstream": "The model provider returned an error. Retry; your draft is untouched.",
-  "error.internal": "Enhancement failed. Retry; your draft is untouched.",
-  "error.rate": "Too many requests or the concurrency cap is full \u2014 retry in a moment."
-};
-var dictionaries = { zh, en };
-var NS = "prompt-enhance";
-
 // src/client/styles.ts
 var STYLE_ID = "dsh-prompt-enhance-styles";
 var CSS = `