feat(models): 模型列表独立成页面，每个模型可独立选提示词风格

- 侧边栏新增「模型列表」页（/admin/models）：自定义模型 + 模型别名，
从设置页移出；删掉随之变成死代码的 ModelSection.jsx
- 自定义模型新增 style 档位：inherit / v2 / writing / legacy / custom，
一档同时决定「内置文案用哪版」与「注不注入特化层」：
v2/writing 固定 v2 文案，legacy 固定旧版短文，custom 用条目正文
且文案跟随全局；与全局 prompt_profile 仍然互斥，不叠加
- 兼容旧配置：只有 prompt、没有 style 的条目按 custom 解释，行为不变
- /admin/model_profiles 与 /admin/settings 返回 style_names
- 设置页按用途分 6 组（提示词 / 上下文与技能 / 运行与行为 / 工具调用 /
安全 / 数据与备份）；提示词风格与思考注入合并成一张卡，
Responses TTL 与 Embeddings 归入运行行为
- 写作规范新增「不要照搬指令原文」一节，并写明两个例外：用户点名要原样
出现的句子/术语/数据/关键词必须一字不改地保留，避免模型连带删掉
- 同步 API(.*).md / README.en.md / prompt-compatibility.md / prompt-design.md

API.en.md

@@ -884,7 +884,7 @@ An unknown value never breaks config loading: it falls back to `v2` with a warni
 
 ### `model_profiles` (custom model list, global)
 
-Expose several public model names on top of the same flash base model, each carrying its own prompt. Whichever name the client sends decides the prompt; entries never affect each other.
+Expose several public model names on top of the same flash base model, each with its own prompt style. Whichever name the client sends decides the style; entries never affect each other.
 
 ```json
 {
@@ -894,7 +894,13 @@ Expose several public model names on top of the same flash base model, each carr
       "base": "deepseek-v4-flash",
       "name": "Writing",
       "remark": "Long-form and rewrites",
-      "prompt": "## Writing rules\n…"
+      "style": "writing"
+    },
+    {
+      "id": "deepseek-chat",
+      "base": "deepseek-v4-flash",
+      "name": "Casual chat",
+      "style": "legacy"
     }
   ]
 }
@@ -905,7 +911,18 @@ Expose several public model names on top of the same flash base model, each carr
 | `id` | Required public model name. Charset `[A-Za-z0-9._-]`, 1-64 characters, must not collide with a built-in model name (e.g. `deepseek-v4-flash`), unique across entries (case-insensitive) |
 | `base` | Required base model: `deepseek-v4-flash` / `-search` / `-nothinking` / `-search-nothinking` |
 | `name` / `remark` | Optional, only for readability in the panel and in exports |
-| `prompt` | Optional prompt that applies to this model only, max 20000 characters; empty adds no extra spec |
+| `style` | Prompt style for this model (see below); empty means `custom` when `prompt` is set, otherwise `inherit` |
+| `prompt` | Optional custom body, **only used when `style=custom`**, max 20000 characters |
+
+`style` values:
+
+| Value | Built-in wording (continuation / contract) | Injected prompt-profile body |
+| --- | --- | --- |
+| `inherit` (default) | Follows global `prompt_style` | Follows global `prompt_profile` |
+| `v2` | Pinned to v2 | None |
+| `writing` | Pinned to v2 | Built-in writing spec |
+| `legacy` | Legacy short wording, no contract | None |
+| `custom` | Follows global `prompt_style` | Uses this entry's `prompt` |
 
 Resolution and injection:
 
@@ -1077,16 +1094,18 @@ Validation runs on write and is independent of `enabled`, so you can author and
 
 ### `GET` / `PUT /admin/model_profiles`
 
-The custom model list (public name + base model + per-model prompt). Like the skill library it is a standalone endpoint with whole-list replacement semantics: the `items` you send become the final list, and an empty array clears it.
+The custom model list (public name + base model + per-model prompt style). Like the skill library it is a standalone endpoint with whole-list replacement semantics: the `items` you send become the final list, and an empty array clears it.
 
 **GET response**:
 
 ```json
 {
   "items": [
-    {"id": "deepseek-writer", "base": "deepseek-v4-flash", "name": "Writing", "remark": "long-form", "prompt": "## Writing rules\n…"}
+    {"id": "deepseek-writer", "base": "deepseek-v4-flash", "name": "Writing", "style": "writing"},
+    {"id": "deepseek-mine", "base": "deepseek-v4-flash", "style": "custom", "prompt": "…"}
   ],
   "base_models": ["deepseek-v4-flash", "deepseek-v4-flash-search", "deepseek-v4-flash-nothinking", "deepseek-v4-flash-search-nothinking"],
+  "style_names": ["inherit", "v2", "writing", "legacy", "custom"],
   "max_count": 50,
   "max_id_chars": 64,
   "max_prompt_chars": 20000
@@ -1096,7 +1115,7 @@ The custom model list (public name + base model + per-model prompt). Like the sk
 **PUT request**:
 
 ```json
-{"items": [{"id": "deepseek-writer", "base": "deepseek-v4-flash", "prompt": "…"}]}
+{"items": [{"id": "deepseek-writer", "base": "deepseek-v4-flash", "style": "writing"}]}
 ```
 
 **Validation** (any failure returns `400` with `{"detail": "..."}`):
@@ -1106,6 +1125,7 @@ The custom model list (public name + base model + per-model prompt). Like the sk
 | `items` | At most 50 entries |
 | `id` | Required, unique (case-insensitive), charset `[A-Za-z0-9._-]`, 1-64 characters, must not shadow a built-in model name |
 | `base` | Required, one of the four `deepseek-v4-flash` names |
+| `style` | Optional; when present must be `inherit` / `v2` / `writing` / `legacy` / `custom` |
 | `prompt` | Optional, ≤ 20000 characters |
 
 Saving takes effect immediately: `/v1/models` and model-name resolution pick up the new entries without a restart.

API.md

@@ -889,7 +889,7 @@ data: {"type":"message_stop"}
 
 ### `model_profiles`（自定义模型列表，全局）
 
-给同一个 flash 基模挂出多个对外模型名，每个名字带自己的提示词。客户端请求里写哪个模型名，就用哪一份提示词，互不影响。
+给同一个 flash 基模挂出多个对外模型名，每个名字可以带自己的提示词风格。客户端请求里写哪个模型名，就用哪一套风格，互不影响。
 
 ```json
 {
@@ -899,7 +899,13 @@ data: {"type":"message_stop"}
       "base": "deepseek-v4-flash",
       "name": "写作",
       "remark": "长文与改稿",
-      "prompt": "## 写作要求\n…"
+      "style": "writing"
+    },
+    {
+      "id": "deepseek-chat",
+      "base": "deepseek-v4-flash",
+      "name": "日常对话",
+      "style": "legacy"
     }
   ]
 }
@@ -910,7 +916,18 @@ data: {"type":"message_stop"}
 | `id` | 必填，对外模型名。字符集 `[A-Za-z0-9._-]`，长度 1-64，不可与内置模型名（如 `deepseek-v4-flash`）冲突，条目间不可重复（大小写不敏感） |
 | `base` | 必填，基模，只能是 `deepseek-v4-flash` / `-search` / `-nothinking` / `-search-nothinking` |
 | `name` / `remark` | 可选，仅用于后台与导出可读性 |
-| `prompt` | 可选，该模型专属提示词，上限 20000 字符；为空表示不额外注入 |
+| `style` | 该模型的提示词风格档位，见下表；留空时：有 `prompt` 按 `custom`，否则按 `inherit` |
+| `prompt` | 可选，自定义正文，**只在 `style=custom` 时生效**，上限 20000 字符 |
+
+`style` 取值：
+
+| 值 | 内置文案（衔接指令 / 执行契约） | 注入的特化层正文 |
+| --- | --- | --- |
+| `inherit`（留空默认） | 跟随全局 `prompt_style` | 跟随全局 `prompt_profile` |
+| `v2` | 固定 v2 | 不注入 |
+| `writing` | 固定 v2 | 内置写作规范正文 |
+| `legacy` | 固定旧版短文案、不注入执行契约 | 不注入 |
+| `custom` | 跟随全局 `prompt_style` | 用本条目 `prompt` |
 
 解析与注入语义：
 
@@ -1082,16 +1099,18 @@ data: {"type":"message_stop"}
 
 ### `GET` / `PUT /admin/model_profiles`
 
-自定义模型列表（模型名 + 基模 + 该模型专属提示词）。与技能库同样是独立接口、整体替换语义：PUT 请求里的 `items` 就是最终列表，传空数组表示清空。
+自定义模型列表（模型名 + 基模 + 该模型的提示词风格）。与技能库同样是独立接口、整体替换语义：PUT 请求里的 `items` 就是最终列表，传空数组表示清空。
 
 **GET 响应**：
 
 ```json
 {
   "items": [
-    {"id": "deepseek-writer", "base": "deepseek-v4-flash", "name": "写作", "remark": "长文", "prompt": "## 写作要求\n…"}
+    {"id": "deepseek-writer", "base": "deepseek-v4-flash", "name": "写作", "style": "writing"},
+    {"id": "deepseek-mine", "base": "deepseek-v4-flash", "style": "custom", "prompt": "…"}
   ],
   "base_models": ["deepseek-v4-flash", "deepseek-v4-flash-search", "deepseek-v4-flash-nothinking", "deepseek-v4-flash-search-nothinking"],
+  "style_names": ["inherit", "v2", "writing", "legacy", "custom"],
   "max_count": 50,
   "max_id_chars": 64,
   "max_prompt_chars": 20000
@@ -1101,7 +1120,7 @@ data: {"type":"message_stop"}
 **PUT 请求**：
 
 ```json
-{"items": [{"id": "deepseek-writer", "base": "deepseek-v4-flash", "prompt": "…"}]}
+{"items": [{"id": "deepseek-writer", "base": "deepseek-v4-flash", "style": "writing"}]}
 ```
 
 **校验规则**（任一条不满足返回 `400`，响应体为 `{"detail": "..."}`）：
@@ -1111,6 +1130,7 @@ data: {"type":"message_stop"}
 | `items` | 最多 50 条 |
 | `id` | 必填、唯一（大小写不敏感），字符集 `[A-Za-z0-9._-]`，长度 1-64，不可与内置模型名冲突 |
 | `base` | 必填，只能是 `deepseek-v4-flash` 系列的四个名字之一 |
+| `style` | 可空；非空时只能是 `inherit` / `v2` / `writing` / `legacy` / `custom` |
 | `prompt` | 可空；≤ 20000 字符 |
 
 保存后立即生效：`/v1/models` 列表与模型名解析都会包含新条目，无需重启。

README.en.md

@@ -340,7 +340,7 @@ Common fields:
 - `thinking_injection`: global thinking-format injection (the "execution contract"); `enabled` defaults to `true` and an empty `prompt` uses the built-in default. A single API key can override it with `api_keys[].thinking_injection` (`custom` / `enabled` / `prompt`), and `custom=false` fully follows the global setting. When a request triggers history splitting, the contract is re-applied to the visible prompt so it never gets buried in the uploaded transcript.
 - `prompt_profile`: prompt-profile layer, **off by default** (`enabled: false`). It prepends one server-side spec block (`name`: `default` / `writing`, or a custom `prompt` up to 20000 characters) to the final prompt, merged with the skill block into a single `System:` section. Presets defer to the user/client persona when they conflict. A single key can override it with `api_keys[].prompt_profile` (`custom` / `enabled` / `name` / `prompt`).
 - `prompt_style`: built-in prompt style, `v2` (default) or `legacy`. `v2` keeps the requirement-carrying continuation prompt and injects the built-in execution contract; `legacy` restores the short pre-`ce5a019` continuation wording and injects no built-in contract (a `thinking_injection.prompt` you set yourself still applies). Switch to `legacy` when replies feel too short or too customer-service-like.
-- `model_profiles`: custom model list (empty by default), at most 50 entries. Each entry maps a public model name (`id`) to a flash-family `base` plus an optional per-model `prompt` (max 20000 characters). Custom names take precedence over built-in aliases, show up in `/v1/models`, and their prompt replaces — never stacks with — `prompt_profile`. Managed through `GET/PUT /admin/model_profiles`.
+- `model_profiles`: custom model list (empty by default), at most 50 entries. Each entry maps a public model name (`id`) to a flash-family `base` plus a `style` (`inherit` / `v2` / `writing` / `legacy` / `custom`) and, for `custom`, its own `prompt` (max 20000 characters). Custom names take precedence over built-in aliases, show up in `/v1/models`, and their resolved body replaces — never stacks with — the global `prompt_profile`. Managed from the **Models** page in the sidebar or via `GET/PUT /admin/model_profiles`.
 - `tools.enabled`: master switch for tool calling, **off by default** (a missing field is treated as `false`). While it is off, requests carrying `tools` are handled as plain chat: no tool instructions are injected and no `tool_calls` are parsed or returned. Turn it on only if you accept the risk — DSML tool output can trigger upstream risk control and get the DeepSeek account banned. Safer alternatives: the plain-text skill library below, or a per-key custom injection prompt.
 - `skills`: experimental plain-text skill library, **off by default** (`enabled: false`). Skill bodies are concatenated into the prompt as a leading system block, so the model needs no tool call at all. `max_chars` (default `8000`) truncates by character and appends a truncation marker; `0` means unlimited. At most 50 items, each with a unique `id` from `[a-z0-9_-]` and a required `body`. A single key can narrow the set with `api_keys[].skills` (`custom` / `enabled` / `ids`). Manage it through `GET/PUT /admin/skills` or `config.json`.
 

docs/prompt-compatibility.md

@@ -232,7 +232,7 @@ DeepSeek web-chat 只接受一段 prompt 字符串和一组文件引用，没有
 | 预设 | 要点 |
 | --- | --- |
 | `default` | 直接回应再按需展开、语气与长度跟着对方走（闲聊就正常聊、不硬上清单小标题）、语言跟随用户、分清确定与推测不编造、不假装能读文件/联网/调工具、用户明确要求优先于本段 |
-| `writing` | 写前确认体裁与读者 → 前几轮要求当硬约束（篇幅/人称/语气/禁用写法逐条满足）→ 具体优先、不写空洞表述 → 分清虚构与非虚构（只对非虚构禁编造）→ 篇幅按用户要求不注水 → 交付前逐条自检（自检不外泄） |
+| `writing` | 写前确认体裁与读者 → 前几轮要求当硬约束（篇幅/人称/语气/禁用写法逐条满足）→ 具体优先、不写空洞表述 → 分清虚构与非虚构（只对非虚构禁编造）→ 篇幅按用户要求不注水 → **不照搬指令原文**（但用户点名要原样出现的句子、术语、数据与关键词一字不改地保留）→ 交付前逐条自检（自检不外泄） |
 
 `default` 与 `writing` 的分工：前者对**所有**请求生效，因此刻意不含「结论先行」这类交付物文风；后者由用户显式开启、面向写作，才可以更具体。这也是「默认关、按需开」的原因。
 
@@ -256,20 +256,33 @@ DeepSeek web-chat 只接受一段 prompt 字符串和一组文件引用，没有
 - `legacy` 的短衔接指令保留了防泄漏要求，只把那三句原文里偏弱的第二句统一成「不要提及文件、上下文或提示词本身」——防泄漏不属于风格约束，不该因为切旧版而放松。
 - 对话记录文件的格式（标题 `# 对话记录`、`[role=...]` 行）不随风格变化。与旧版的实际差距主要来自「有没有执行契约」和「衔接指令有多长」。
 
-### 模型专属提示词（model_profiles）
+### 模型提示词风格（model_profiles）
 
-自定义模型条目把「对外模型名 → 基模 + 提示词」绑在一起，适合把不同风格并排放给客户端选，而不是反复改全局开关。
+自定义模型条目把「对外模型名 → 基模 + 提示词风格」绑在一起，适合把不同风格并排放给客户端选，而不是反复改全局开关。
 
 | 项目 | 行为 |
 | --- | --- |
 | 配置 | 顶层 `model_profiles`，数组，最多 50 条 |
 | 解析优先 | 自定义模型名优先于内置别名映射；命中后按其 `base` 调上游 |
 | 列表 | `/v1/models` 一并列出，`owned_by=ds2api`；Claude / Gemini 的模型列表接口不列 |
+| 风格档位 | `inherit`（跟随全局）/ `v2` / `writing` / `legacy` / `custom`，见下表 |
 | 注入点 | 与 `prompt_profile` 相同（拆分之后、技能之前） |
-| 与特化层关系 | **互斥**：该模型 `prompt` 非空就用它，否则回落到全局 `prompt_profile` |
+| 与特化层关系 | **互斥**：模型档位给出的正文优先，全局 `prompt_profile` 不叠加 |
 | 生效范围 | 只对该模型名生效；客户端换回 `deepseek-v4-flash` 就等于关掉它 |
 | 管理接口 | `GET/PUT /admin/model_profiles`，整体替换语义 |
 
+一个条目同时决定「内置文案用哪版」与「注不注入特化层」：
+
+| `style` | 衔接指令 / 执行契约 | 特化层正文 |
+| --- | --- | --- |
+| `inherit` | 全局 `prompt_style` | 全局 `prompt_profile` |
+| `v2` | 固定 v2 | 不注入 |
+| `writing` | 固定 v2 | 内置写作规范正文 |
+| `legacy` | 固定旧版短文案、不注入执行契约 | 不注入 |
+| `custom` | 全局 `prompt_style` | 条目自己的 `prompt` |
+
+兼容说明：早期条目只有 `prompt` 字段、没有 `style`，这种配置按 `custom` 解释（`EffectiveStyle`），升级后行为不变。
+
 ### 技能注入（skills）
 
 技能是一段服务端配置的纯文本 markdown，直接拼进 prompt。模型全程不需要输出工具调用，看到的只是普通上下文，因此不会碰到 DSML 风控。

docs/prompt-design.md

@@ -148,13 +148,22 @@ L4 技能层        前置 System 段：<skill> 正文块               BuildSki
 5. 保持上下文连续：人名、术语、编号、称呼与上文一致；改稿时说明改动了哪些地方。
 6. 语言跟随用户；中文写作不用「作为AI」「总之」这类套话收尾。
 
+## 不要照搬指令原文
+1. 把用户的指令当成要求，不要当成正文素材：要求句、约束说明、括号里的备注，都不要原样搬进正文，
+   也不要复述「请写一篇…」「要求：…」这类指令框架。
+2. 这一条不适用于用户明确要求原样出现的内容：点名要照抄的句子、口号、标题、术语、数据、引文，
+   以及必须出现的关键词，都要一字不改地写进去，位置按用户要求安排。
+3. 拿不准时分辨这句是「让我怎么写」还是「要我写什么」：前者转成正文表达，后者原样保留。
+
 ## 交付前自检
-逐条核对：篇幅是否符合、硬约束是否满足、有没有漏掉用户点名要写的点、有没有出现用户明确禁止的写法。
-发现缺项先补齐再交付；这份自检只在内部完成，不要输出核对过程。
+逐条核对：篇幅是否符合、硬约束是否满足、有没有漏掉用户点名要写的点、有没有出现用户明确禁止的写法、
+有没有把指令原文抄成正文。发现缺项先补齐再交付；这份自检只在内部完成，不要输出核对过程。
 ```
 
 写作规范可以更具体，因为它是用户**显式开启**的。但「不要编造」必须限定在非虚构：把这条笼统地压在小说、剧本、广告创意上，会直接削弱创作能力。「改稿时说明改动了哪些地方」也是同理——只在改稿场景要求，不要求每轮自我解释。
 
+「不要照搬指令原文」这一条同样必须自带例外：只针对**指令性文字**（「让我怎么写」），不针对**用户点名要原样出现的素材**（「要我写什么」）。写成笼统的「不准引用用户原话」会把「把这句话原样写进标题」「必须出现 XXX 这个词」一起禁掉，那是反效果。第 3 条给模型一个可执行的判据，避免它在边界上左右为难。
+
 ---
 
 ## 4. 拼接与顺序规则
@@ -232,18 +241,28 @@ L4 技能层        前置 System 段：<skill> 正文块               BuildSki
 
 ### 用法 C：自定义模型（把多套风格并排放给客户端）
 
-给同一个 flash 基模挂多个对外模型名，每个名字带自己的提示词。客户端在模型列表里就能直接选风格，不用来回改全局配置，回滚也只是换回原模型名。
+给同一个 flash 基模挂多个对外模型名，每个名字选一个风格档位。客户端在模型列表里直接选风格，不用来回改全局配置，回滚也只是换回原模型名。
 
 ```json
 {
   "model_profiles": [
-    {"id": "deepseek-writer", "base": "deepseek-v4-flash", "name": "写作", "prompt": "<writing 预设正文>"},
-    {"id": "deepseek-chat", "base": "deepseek-v4-flash", "name": "日常对话", "prompt": ""}
+    {"id": "deepseek-chat",   "base": "deepseek-v4-flash", "name": "日常对话", "style": "legacy"},
+    {"id": "deepseek-writer", "base": "deepseek-v4-flash", "name": "写作",     "style": "writing"},
+    {"id": "deepseek-plain",  "base": "deepseek-v4-flash", "name": "跟随全局", "style": "inherit"},
+    {"id": "deepseek-mine",   "base": "deepseek-v4-flash", "name": "我的规范", "style": "custom", "prompt": "…"}
   ]
 }
 ```
 
-规则：模型命中且 `prompt` 非空时，**不会**再叠加全局 `prompt_profile`（两者互斥）；`prompt` 留空则回落到全局特化层。基模带 `-nothinking` / `-search` 后缀时，思考与联网默认值跟着基模走。后台在「自定义模型」板块维护，接口是 `GET/PUT /admin/model_profiles`。
+| `style` | 衔接指令 / 执行契约 | 特化层正文 |
+| --- | --- | --- |
+| `inherit`（缺省） | 跟随全局 `prompt_style` | 跟随全局 `prompt_profile` |
+| `v2` | 固定 v2 | 不注入 |
+| `writing` | 固定 v2 | 内置写作规范正文 |
+| `legacy` | 固定旧版短文案、不注入执行契约 | 不注入 |
+| `custom` | 跟随全局 `prompt_style` | 条目自己的 `prompt` |
+
+规则：模型档位给出的正文与全局 `prompt_profile` **互斥**（不会叠加）；`prompt` 只在 `custom` 档位下生效；基模带 `-nothinking` / `-search` 后缀时，思考与联网默认值跟着基模走。后台在左侧「模型列表」页维护，接口是 `GET/PUT /admin/model_profiles`。
 
 三种用法怎么选：
 

internal/config/model_profile.go

@@ -46,6 +46,54 @@ func IsKnownPromptStyle(style string) bool {
 	}
 }
 
+// 自定义模型条目的风格档位。
+//
+// 一个条目既能自带提示词（custom），也能直接选一个内置风格，或跟随全局配置。
+// 这样同一个 flash 基模可以挂出「日常 v2」「写作」「旧版」等多个模型名并排给客户端选。
+const (
+	// ModelProfileStyleInherit 跟随全局：prompt_style 与 prompt_profile 都按全局设置。
+	ModelProfileStyleInherit = "inherit"
+	// ModelProfileStyleV2 强制 v2 内置文案，不注入特化层。
+	ModelProfileStyleV2 = "v2"
+	// ModelProfileStyleWriting 强制 v2 内置文案 + 写作规范正文。
+	ModelProfileStyleWriting = "writing"
+	// ModelProfileStyleLegacy 强制旧版内置文案，不注入特化层。
+	ModelProfileStyleLegacy = "legacy"
+	// ModelProfileStyleCustom 用本条目 prompt 作为特化层正文，文案风格跟随全局。
+	ModelProfileStyleCustom = "custom"
+)
+
+// ModelProfileStyleNames 返回风格档位名，顺序固定，供后台下拉与文档对照。
+func ModelProfileStyleNames() []string {
+	return []string{
+		ModelProfileStyleInherit,
+		ModelProfileStyleV2,
+		ModelProfileStyleWriting,
+		ModelProfileStyleLegacy,
+		ModelProfileStyleCustom,
+	}
+}
+
+// IsKnownModelProfileStyle 判断是否为可选风格档位。
+func IsKnownModelProfileStyle(style string) bool {
+	switch strings.ToLower(strings.TrimSpace(style)) {
+	case ModelProfileStyleInherit, ModelProfileStyleV2, ModelProfileStyleWriting,
+		ModelProfileStyleLegacy, ModelProfileStyleCustom:
+		return true
+	default:
+		return false
+	}
+}
+
+// NormalizeModelProfileStyle 归一化风格档位：空值按 inherit 处理，未知值同样回落 inherit。
+func NormalizeModelProfileStyle(style string) string {
+	normalized := strings.ToLower(strings.TrimSpace(style))
+	if IsKnownModelProfileStyle(normalized) {
+		return normalized
+	}
+	return ModelProfileStyleInherit
+}
+
 // 自定义模型条目（model_profiles）的限制。
 const (
 	// MaxModelProfiles 是自定义模型条目的数量上限。
@@ -54,10 +102,7 @@ const (
 	MaxModelProfileIDChars = 64
 )
 
-// ModelProfileConfig 是一条自定义模型：对外名称 + 基模 + 该模型专属提示词。
-//
-// 用途：同一个 flash 基模可以挂出多个「模型名」，每个名字带自己的提示词，
-// 客户端按模型名选用，互不影响，比改全局开关更好回滚。
+// ModelProfileConfig 是一条自定义模型：对外名称 + 基模 + 该模型专属的提示词风格/正文。
 type ModelProfileConfig struct {
 	// ID 是对外模型名，客户端在请求里写的就是它。
 	ID string `json:"id"`
@@ -67,10 +112,28 @@ type ModelProfileConfig struct {
 	Name string `json:"name,omitempty"`
 	// Remark 是备注（可选）。
 	Remark string `json:"remark,omitempty"`
-	// Prompt 是该模型专属的提示词正文；为空表示不额外注入（回落到全局 prompt_profile）。
+	// Style 是风格档位：inherit / v2 / writing / legacy / custom。
+	// 留空时：有 Prompt 视为 custom，否则视为 inherit（兼容旧配置）。
+	Style string `json:"style,omitempty"`
+	// Prompt 是该模型专属的提示词正文，只在 Style=custom 时使用。
 	Prompt string `json:"prompt,omitempty"`
 }
 
+// EffectiveStyle 返回生效的风格档位。
+//
+// 兼容旧配置：早期版本只有 prompt 字段，没有 style；那时填写 prompt 的语义就是
+// 「用这段正文当特化层」，所以这里把它解释成 custom。
+func (m ModelProfileConfig) EffectiveStyle() string {
+	style := strings.ToLower(strings.TrimSpace(m.Style))
+	if style == "" {
+		if strings.TrimSpace(m.Prompt) != "" {
+			return ModelProfileStyleCustom
+		}
+		return ModelProfileStyleInherit
+	}
+	return NormalizeModelProfileStyle(style)
+}
+
 // ModelProfileReader 由配置中心实现，供模型解析与列表接口读取自定义模型。
 type ModelProfileReader interface {
 	ModelProfiles() []ModelProfileConfig
@@ -88,13 +151,18 @@ func SupportedBaseModels() []string {
 	}
 }
 
-// Normalize 归一化单条自定义模型：裁剪空白、基模转小写。
+// Normalize 归一化单条自定义模型：裁剪空白、基模转小写、风格档位转小写。
 func (m ModelProfileConfig) Normalize() ModelProfileConfig {
+	style := strings.ToLower(strings.TrimSpace(m.Style))
+	if style != "" {
+		style = NormalizeModelProfileStyle(style)
+	}
 	return ModelProfileConfig{
 		ID:     strings.TrimSpace(m.ID),
 		Base:   strings.ToLower(strings.TrimSpace(m.Base)),
 		Name:   strings.TrimSpace(m.Name),
 		Remark: strings.TrimSpace(m.Remark),
+		Style:  style,
 		Prompt: TruncatePromptProfileText(m.Prompt),
 	}
 }
@@ -202,6 +270,10 @@ func ValidateModelProfiles(items []ModelProfileConfig) error {
 			return fmt.Errorf("model_profiles[%d].base must be one of %s",
 				i, strings.Join(SupportedBaseModels(), ", "))
 		}
+		if rawStyle := strings.TrimSpace(raw.Style); rawStyle != "" && !IsKnownModelProfileStyle(rawStyle) {
+			return fmt.Errorf("model_profiles[%d].style must be one of %s",
+				i, strings.Join(ModelProfileStyleNames(), ", "))
+		}
 	}
 	return nil
 }

internal/config/model_profile_test.go

@@ -161,20 +161,60 @@ func TestResolveModelUsesModelProfiles(t *testing.T) {
 	}
 }
 
-func TestStoreResolveModelPrompt(t *testing.T) {
+func TestStoreResolveModelProfile(t *testing.T) {
 	store := &Store{cfg: Config{ModelProfiles: []ModelProfileConfig{
 		{ID: "deepseek-writer", Base: "deepseek-v4-flash", Prompt: " writing spec "},
 		{ID: "deepseek-plain", Base: "deepseek-v4-flash"},
+		{ID: "deepseek-legacy", Base: "deepseek-v4-flash", Style: "legacy"},
 	}}}
-	if body, ok := store.ResolveModelPrompt("deepseek-writer"); !ok || body != "writing spec" {
-		t.Fatalf("body=%q ok=%v", body, ok)
+	// 老配置只写了 prompt：按 custom 处理，正文生效。
+	got := store.ResolveModelProfile("deepseek-writer")
+	if !got.Found || got.Style != ModelProfileStyleCustom || got.Prompt != "writing spec" {
+		t.Fatalf("unexpected settings: %#v", got)
 	}
-	// 没有提示词的条目不算命中，调用方应回落到全局特化层。
-	if _, ok := store.ResolveModelPrompt("deepseek-plain"); ok {
-		t.Fatal("expected empty prompt to be treated as no match")
+	// 什么都没配：inherit。
+	got = store.ResolveModelProfile("deepseek-plain")
+	if !got.Found || got.Style != ModelProfileStyleInherit || got.Prompt != "" {
+		t.Fatalf("unexpected settings: %#v", got)
 	}
-	if _, ok := store.ResolveModelPrompt("deepseek-v4-flash"); ok {
-		t.Fatal("expected builtin model to have no model prompt")
+	got = store.ResolveModelProfile("deepseek-legacy")
+	if !got.Found || got.Style != ModelProfileStyleLegacy {
+		t.Fatalf("unexpected settings: %#v", got)
+	}
+	// 未命中内置模型时完全跟随全局。
+	if got := store.ResolveModelProfile("deepseek-v4-flash"); got.Found {
+		t.Fatalf("expected no profile for builtin model, got %#v", got)
+	}
+}
+
+func TestModelProfileStylesNormalize(t *testing.T) {
+	cases := map[string]string{
+		"":        ModelProfileStyleInherit,
+		"inherit": ModelProfileStyleInherit,
+		"V2":      ModelProfileStyleV2,
+		"writing": ModelProfileStyleWriting,
+		" legacy": ModelProfileStyleLegacy,
+		"custom":  ModelProfileStyleCustom,
+		"nope":    ModelProfileStyleInherit,
+	}
+	for input, want := range cases {
+		if got := NormalizeModelProfileStyle(input); got != want {
+			t.Fatalf("style %q -> %q, want %q", input, got, want)
+		}
+	}
+	// 老配置兼容：无 style + 有 prompt ⇒ custom；无 style + 无 prompt ⇒ inherit。
+	if got := (ModelProfileConfig{Prompt: "x"}).EffectiveStyle(); got != ModelProfileStyleCustom {
+		t.Fatalf("expected custom for legacy prompt-only entry, got %q", got)
+	}
+	if got := (ModelProfileConfig{}).EffectiveStyle(); got != ModelProfileStyleInherit {
+		t.Fatalf("expected inherit for empty entry, got %q", got)
+	}
+}
+
+func TestValidateModelProfilesRejectsUnknownStyle(t *testing.T) {
+	err := ValidateModelProfiles([]ModelProfileConfig{{ID: "a", Base: "deepseek-v4-flash", Style: "nope"}})
+	if err == nil || !strings.Contains(err.Error(), "style must be one of") {
+		t.Fatalf("error=%v", err)
 	}
 }
 

internal/config/store_accessors.go

@@ -349,25 +349,33 @@ func (s *Store) ModelProfiles() []ModelProfileConfig {
 	return cloneModelProfiles(s.cfg.ModelProfiles)
 }
 
-// ResolveModelPrompt 返回某个模型名对应的专属提示词。
-//
-// 按客户端请求里的模型名（大小写不敏感）匹配自定义模型条目；命中且正文非空时
-// ok=true。未命中或正文为空时调用方应回落到全局 prompt_profile。
-func (s *Store) ResolveModelPrompt(modelID string) (string, bool) {
+// ModelProfileSettings 是某个自定义模型条目对单次请求的提示词覆盖。
+type ModelProfileSettings struct {
+	// Found 表示请求里的模型名命中了自定义模型条目。
+	Found bool
+	// Style 是生效的风格档位：inherit / v2 / writing / legacy / custom。
+	Style string
+	// Prompt 是自定义正文，只在 Style=custom 时有意义。
+	Prompt string
+}
+
+// ResolveModelProfile 按客户端请求里的模型名（大小写不敏感）解析自定义模型条目。
+// 未命中时返回零值（Found=false），调用方应完全跟随全局设置。
+func (s *Store) ResolveModelProfile(modelID string) ModelProfileSettings {
 	if s == nil {
-		return "", false
+		return ModelProfileSettings{}
 	}
 	s.mu.RLock()
 	profile, ok := FindModelProfile(s.cfg.ModelProfiles, modelID)
 	s.mu.RUnlock()
 	if !ok {
-		return "", false
+		return ModelProfileSettings{}
 	}
-	body := strings.TrimSpace(profile.Prompt)
-	if body == "" {
-		return "", false
+	return ModelProfileSettings{
+		Found:  true,
+		Style:  profile.EffectiveStyle(),
+		Prompt: strings.TrimSpace(profile.Prompt),
 	}
-	return body, true
 }
 
 // PromptProfileSettings 是「提示词特化层」对单次请求生效的设置。

internal/httpapi/admin/modelprofiles/handler_model_profiles.go

@@ -12,6 +12,7 @@ import (
 type modelProfilesPayload struct {
 	Items          []config.ModelProfileConfig `json:"items"`
 	BaseModels     []string                    `json:"base_models"`
+	StyleNames     []string                    `json:"style_names"`
 	MaxCount       int                         `json:"max_count"`
 	MaxIDChars     int                         `json:"max_id_chars"`
 	MaxPromptChars int                         `json:"max_prompt_chars"`
@@ -22,6 +23,7 @@ func (h *Handler) getModelProfiles(w http.ResponseWriter, _ *http.Request) {
 	writeJSON(w, http.StatusOK, modelProfilesPayload{
 		Items:          items,
 		BaseModels:     config.SupportedBaseModels(),
+		StyleNames:     config.ModelProfileStyleNames(),
 		MaxCount:       config.MaxModelProfiles,
 		MaxIDChars:     config.MaxModelProfileIDChars,
 		MaxPromptChars: config.MaxPromptProfileChars,

internal/httpapi/admin/modelprofiles/handler_model_profiles_test.go

@@ -38,6 +38,16 @@ func TestGetModelProfilesReturnsItemsAndLimits(t *testing.T) {
 	if len(body.BaseModels) == 0 {
 		t.Fatal("expected base model options")
 	}
+	if len(body.StyleNames) != len(config.ModelProfileStyleNames()) {
+		t.Fatalf("style_names=%#v want %#v", body.StyleNames, config.ModelProfileStyleNames())
+	}
+	// 老配置只有 prompt、没有 style：接口层按 custom 解释，前端要能看到这一点。
+	if body.Items[0].Style != "" {
+		t.Fatalf("stored style should stay empty for legacy entries, got %q", body.Items[0].Style)
+	}
+	if got := body.Items[0].EffectiveStyle(); got != config.ModelProfileStyleCustom {
+		t.Fatalf("effective style=%q want custom", got)
+	}
 	if body.MaxCount != config.MaxModelProfiles {
 		t.Fatalf("max_count=%d want %d", body.MaxCount, config.MaxModelProfiles)
 	}
@@ -98,6 +108,7 @@ func TestUpdateModelProfilesValidatesInput(t *testing.T) {
 		{name: "empty id", items: []map[string]any{{"base": "deepseek-v4-flash"}}},
 		{name: "bad base", items: []map[string]any{{"id": "a", "base": "gpt-4o"}}},
 		{name: "builtin shadow", items: []map[string]any{{"id": "deepseek-v4-flash", "base": "deepseek-v4-flash"}}},
+		{name: "unknown style", items: []map[string]any{{"id": "a", "base": "deepseek-v4-flash", "style": "nope"}}},
 		{name: "duplicate", items: []map[string]any{
 			{"id": "a", "base": "deepseek-v4-flash"},
 			{"id": "a", "base": "deepseek-v4-flash"},

internal/httpapi/admin/settings/handler_settings_read.go

@@ -97,6 +97,7 @@ func (h *Handler) getSettings(w http.ResponseWriter, _ *http.Request) {
 			"count":            len(h.Store.ModelProfiles()),
 			"max_count":        config.MaxModelProfiles,
 			"base_models":      config.SupportedBaseModels(),
+			"style_names":      config.ModelProfileStyleNames(),
 			"max_id_chars":     config.MaxModelProfileIDChars,
 			"max_prompt_chars": config.MaxPromptProfileChars,
 		},

internal/httpapi/openai/history/current_input_file.go

@@ -39,7 +39,7 @@ type Service struct {
 func (s Service) ApplyCurrentInputFile(ctx context.Context, a *auth.RequestAuth, stdReq promptcompat.StandardRequest) (promptcompat.StandardRequest, error) {
 	// 执行契约会写进最新一条用户消息，必须发生在历史拆分之前，
 	// 否则拆分后上传的上下文文件里就缺了这段要求（旧版风格不注入内置契约）。
-	stdReq = applyThinkingInjection(a, stdReq)
+	stdReq = s.applyThinkingInjection(a, stdReq)
 	appliedBefore := stdReq.CurrentInputFileApplied
 	stdReq, err := s.applyCurrentInputFile(ctx, a, stdReq)
 	if err != nil {
@@ -49,7 +49,7 @@ func (s Service) ApplyCurrentInputFile(ctx context.Context, a *auth.RequestAuth,
 	// 在附件里。这里把同一段文本贴回可见 prompt，否则它在长对话中实际上
 	// 等于没注入。只在本轮新发生拆分时贴一次，避免切号重传重复叠加。
 	if !appliedBefore && stdReq.CurrentInputFileApplied {
-		stdReq = applyThinkingInjectionToVisiblePrompt(a, stdReq)
+		stdReq = s.applyThinkingInjectionToVisiblePrompt(a, stdReq)
 	}
 	// 纯文本增强（模型提示词 / 提示词特化层 / 技能库）都是 prompt 级叠加，必须
 	// 放在历史拆分之后：拆分会把 prompt 整段替换成上下文文件说明，提前写入会被覆盖。
@@ -142,7 +142,7 @@ func (s Service) applyCurrentInputFile(ctx context.Context, a *auth.RequestAuth,
 	} else if len(stdReq.ToolNames) == 0 && len(toolNames) > 0 {
 		stdReq.ToolNames = toolNames
 	}
-	style := promptStyleOf(a)
+	style := s.resolvePromptStyle(a, stdReq)
 	continuation := currentInputFilePrompt(style, filename, hasToolsFile)
 	promptMessages = append(promptMessages, map[string]any{
 		"role":    "user",
@@ -237,80 +237,109 @@ func (s Service) ReuploadAppliedCurrentInputFile(ctx context.Context, a *auth.Re
 	return stdReq, nil
 }
 
-// promptStyleOf 返回本次请求生效的内置提示词风格；旧版风格不注入内置执行契约。
-func promptStyleOf(a *auth.RequestAuth) string {
-	if a == nil {
+// modelProfileReader 由配置中心实现，用于读取自定义模型条目的风格与正文。
+// 用可选接口而不是写进 CurrentInputConfigReader，是为了不打断测试里的轻量假实现。
+type modelProfileReader interface {
+	ResolveModelProfile(modelID string) config.ModelProfileSettings
+}
+
+// modelProfile 返回客户端所选自定义模型条目的设置；未命中时返回零值。
+//
+// 用 RequestedModel 而不是 ResolvedModel：解析后的模型名已经是 flash 基模，
+// 拿不回「客户端到底选了哪个自定义模型」。
+func (s Service) modelProfile(stdReq promptcompat.StandardRequest) config.ModelProfileSettings {
+	if s.Store == nil {
+		return config.ModelProfileSettings{}
+	}
+	reader, ok := s.Store.(modelProfileReader)
+	if !ok {
+		return config.ModelProfileSettings{}
+	}
+	modelID := strings.TrimSpace(stdReq.RequestedModel)
+	if modelID == "" {
+		return config.ModelProfileSettings{}
+	}
+	return reader.ResolveModelProfile(modelID)
+}
+
+// resolvePromptStyle 返回本次请求生效的内置提示词风格。
+//
+// 自定义模型条目可以覆盖全局风格：v2 / writing 强制 v2 文案，legacy 强制旧版文案，
+// custom 与 inherit 则跟随全局（custom 只换特化层正文，不改内置文案）。
+func (s Service) resolvePromptStyle(a *auth.RequestAuth, stdReq promptcompat.StandardRequest) string {
+	global := config.PromptStyleV2
+	if a != nil {
+		global = config.NormalizePromptStyle(a.PromptStyle)
+	}
+	switch s.modelProfile(stdReq).Style {
+	case config.ModelProfileStyleV2, config.ModelProfileStyleWriting:
 		return config.PromptStyleV2
+	case config.ModelProfileStyleLegacy:
+		return config.PromptStyleLegacy
+	default:
+		return global
 	}
-	return config.NormalizePromptStyle(a.PromptStyle)
 }
 
 // applyThinkingInjection 应用本次请求生效的「思考格式注入」设置。
 // 设置已在鉴权阶段完成全局 + 单 Key 自定义的合并，这里只负责落盘到请求里。
-func applyThinkingInjection(a *auth.RequestAuth, stdReq promptcompat.StandardRequest) promptcompat.StandardRequest {
+func (s Service) applyThinkingInjection(a *auth.RequestAuth, stdReq promptcompat.StandardRequest) promptcompat.StandardRequest {
 	if a == nil {
 		return stdReq
 	}
 	return promptcompat.ApplyThinkingInjectionToRequest(
-		stdReq, promptStyleOf(a), a.Injection.Enabled, a.Injection.Prompt)
+		stdReq, s.resolvePromptStyle(a, stdReq), a.Injection.Enabled, a.Injection.Prompt)
 }
 
 // applyThinkingInjectionToVisiblePrompt 在历史拆分替换掉 prompt 之后，
 // 把执行契约重新贴到可见 prompt 上（只改 FinalPrompt）。
-func applyThinkingInjectionToVisiblePrompt(a *auth.RequestAuth, stdReq promptcompat.StandardRequest) promptcompat.StandardRequest {
+func (s Service) applyThinkingInjectionToVisiblePrompt(a *auth.RequestAuth, stdReq promptcompat.StandardRequest) promptcompat.StandardRequest {
 	if a == nil {
 		return stdReq
 	}
 	return promptcompat.ReapplyThinkingInjectionToPrompt(
-		stdReq, promptStyleOf(a), a.Injection.Enabled, a.Injection.Prompt)
+		stdReq, s.resolvePromptStyle(a, stdReq), a.Injection.Enabled, a.Injection.Prompt)
 }
 
-// modelPromptReader 由配置中心实现，用于读取自定义模型条目的专属提示词。
-// 用可选接口而不是写进 CurrentInputConfigReader，是为了不打断测试里的轻量假实现。
-type modelPromptReader interface {
-	ResolveModelPrompt(modelID string) (string, bool)
-}
-
-// applyOverlays 把 prompt 级叠加层（模型专属提示词 / 提示词特化层 / 技能库）合进最终 prompt。
+// applyOverlays 把 prompt 级叠加层（模型风格/正文 / 全局特化层 / 技能库）合进最终 prompt。
 //
 // 分段规则：
-//   - 外层规范二选一：命中的自定义模型若带提示词，就用它的，否则回落到全局 prompt_profile；
+//   - 外层规范按「模型条目 > 全局 prompt_profile」解析（见 modelOverlayBody）；
 //   - 内层是技能库（更贴近当前任务）；
 //   - 全部合成一个 System 段。
 func (s Service) applyOverlays(a *auth.RequestAuth, stdReq promptcompat.StandardRequest) promptcompat.StandardRequest {
 	blocks := make([]string, 0, 2)
-	if body := s.modelProfilePrompt(stdReq); body != "" {
-		blocks = append(blocks, body)
-	} else if a != nil {
-		blocks = append(blocks, promptcompat.PromptProfileBlock(a.PromptProfile))
-	}
+	blocks = append(blocks, s.modelOverlayBody(a, stdReq))
 	if a != nil && a.Skills.Enabled && len(a.Skills.Items) > 0 {
 		blocks = append(blocks, promptcompat.BuildSkillsPrompt(a.Skills.Items, a.Skills.MaxChars))
 	}
 	return promptcompat.ApplySystemOverlays(stdReq, blocks...)
 }
 
-// modelProfilePrompt 返回客户端所选自定义模型自带的提示词；没有则返回空串。
+// modelOverlayBody 决定外层规范正文。
 //
-// 用 RequestedModel 而不是 ResolvedModel：解析后的模型名已经是 flash 基模，
-// 拿不回「客户端到底选了哪个自定义模型」。
-func (s Service) modelProfilePrompt(stdReq promptcompat.StandardRequest) string {
-	if s.Store == nil {
-		return ""
-	}
-	reader, ok := s.Store.(modelPromptReader)
-	if !ok {
+//	inherit — 全局 prompt_profile
+//	v2 / legacy — 不注入（这两档只改内置文案）
+//	writing — 写作规范正文
+//	custom — 条目自己的正文；为空时回落到全局
+func (s Service) modelOverlayBody(a *auth.RequestAuth, stdReq promptcompat.StandardRequest) string {
+	profile := s.modelProfile(stdReq)
+	switch profile.Style {
+	case config.ModelProfileStyleV2, config.ModelProfileStyleLegacy:
 		return ""
-	}
-	modelID := strings.TrimSpace(stdReq.RequestedModel)
-	if modelID == "" {
-		return ""
-	}
-	body, ok := reader.ResolveModelPrompt(modelID)
-	if !ok {
+	case config.ModelProfileStyleWriting:
+		return promptcompat.BuildPromptProfilePrompt(config.PromptProfileNameWriting, "")
+	case config.ModelProfileStyleCustom:
+		if body := strings.TrimSpace(profile.Prompt); body != "" {
+			return promptcompat.BuildPromptProfilePrompt("", body)
+		}
 		return ""
+	default:
+		if a == nil {
+			return ""
+		}
+		return promptcompat.PromptProfileBlock(a.PromptProfile)
 	}
-	return strings.TrimSpace(body)
 }
 
 func latestUserInputForFile(messages []any) (int, string) {

internal/httpapi/openai/history/current_input_file_overlay_test.go

@@ -10,26 +10,23 @@ import (
 	"ds2api/internal/promptcompat"
 )
 
-// overlayTestStore 让历史拆分保持开启，并可按模型名返回专属提示词。
+// overlayTestStore 让历史拆分保持开启，并可按模型名返回该条目的风格档位。
 type overlayTestStore struct {
-	enabled      bool
-	minChars     int
-	modelPrompts map[string]string
+	enabled  bool
+	minChars int
+	profiles map[string]config.ModelProfileSettings
 }
 
 func (s overlayTestStore) CurrentInputFileEnabled() bool { return s.enabled }
 func (s overlayTestStore) CurrentInputFileMinChars() int { return s.minChars }
 
-func (s overlayTestStore) ResolveModelPrompt(modelID string) (string, bool) {
-	for id, prompt := range s.modelPrompts {
+func (s overlayTestStore) ResolveModelProfile(modelID string) config.ModelProfileSettings {
+	for id, settings := range s.profiles {
 		if strings.EqualFold(strings.TrimSpace(id), strings.TrimSpace(modelID)) {
-			if strings.TrimSpace(prompt) == "" {
-				return "", false
-			}
-			return prompt, true
+			return settings
 		}
 	}
-	return "", false
+	return config.ModelProfileSettings{}
 }
 
 func overlayTestRequest(requestedModel string) promptcompat.StandardRequest {
@@ -46,83 +43,114 @@ func overlayTestRequest(requestedModel string) promptcompat.StandardRequest {
 	}
 }
 
-// 选中带提示词的自定义模型时，用模型自己的提示词，不再叠加全局特化层。
-func TestModelProfilePromptWinsOverGlobalProfile(t *testing.T) {
-	svc := Service{
-		Store: overlayTestStore{
-			enabled:      true,
-			modelPrompts: map[string]string{"deepseek-writer": "model spec"},
-		},
-		DS: &skillsTestUploader{},
+func applyWithProfile(t *testing.T, a *auth.RequestAuth, store overlayTestStore, model string) promptcompat.StandardRequest {
+	t.Helper()
+	out, err := (Service{Store: store, DS: &skillsTestUploader{}}).
+		ApplyCurrentInputFile(context.Background(), a, overlayTestRequest(model))
+	if err != nil {
+		t.Fatalf("apply current input file failed: %v", err)
 	}
+	return out
+}
+
+// custom 档位：用条目自己的正文，不再叠加全局特化层。
+func TestModelProfileCustomOverridesGlobalProfile(t *testing.T) {
+	store := overlayTestStore{enabled: true, profiles: map[string]config.ModelProfileSettings{
+		"deepseek-writer": {Found: true, Style: config.ModelProfileStyleCustom, Prompt: "model spec"},
+	}}
 	a := &auth.RequestAuth{
 		PromptProfile: config.PromptProfileSettings{Enabled: true, Name: config.PromptProfileNameWriting},
 	}
-	out, err := svc.ApplyCurrentInputFile(context.Background(), a, overlayTestRequest("deepseek-writer"))
-	if err != nil {
-		t.Fatalf("apply current input file failed: %v", err)
-	}
+	out := applyWithProfile(t, a, store, "deepseek-writer")
 	if !strings.Contains(out.FinalPrompt, "model spec") {
-		t.Fatalf("expected model prompt in final prompt, got %q", out.FinalPrompt)
+		t.Fatalf("expected model prompt, got %q", out.FinalPrompt)
 	}
 	globalBody := promptcompat.BuildPromptProfilePrompt(config.PromptProfileNameWriting, "")
 	if strings.Contains(out.FinalPrompt, globalBody) {
-		t.Fatalf("global prompt profile must not stack with the model prompt, got %q", out.FinalPrompt)
+		t.Fatalf("global profile must not stack with the model prompt, got %q", out.FinalPrompt)
 	}
 	if got := strings.Count(out.FinalPrompt, "System: "); got != 1 {
 		t.Fatalf("expected one System marker, got %d", got)
 	}
 }
 
-// 模型名没命中（或该模型没配提示词）时回落到全局特化层。
-func TestModelProfilePromptFallsBackToGlobalProfile(t *testing.T) {
-	cases := []struct {
-		name         string
-		requested    string
-		modelPrompts map[string]string
-	}{
-		{name: "unknown model", requested: "deepseek-v4-flash"},
-		{name: "model without prompt", requested: "deepseek-plain", modelPrompts: map[string]string{"deepseek-plain": "   "}},
-	}
-	for _, tc := range cases {
-		t.Run(tc.name, func(t *testing.T) {
-			svc := Service{
-				Store: overlayTestStore{enabled: true, modelPrompts: tc.modelPrompts},
-				DS:    &skillsTestUploader{},
-			}
-			a := &auth.RequestAuth{
-				PromptProfile: config.PromptProfileSettings{Enabled: true, Name: config.PromptProfileNameWriting},
-			}
-			out, err := svc.ApplyCurrentInputFile(context.Background(), a, overlayTestRequest(tc.requested))
-			if err != nil {
-				t.Fatalf("apply current input file failed: %v", err)
-			}
-			if !strings.Contains(out.FinalPrompt, promptcompat.BuildPromptProfilePrompt(config.PromptProfileNameWriting, "")) {
-				t.Fatalf("expected global profile fallback, got %q", out.FinalPrompt)
-			}
-		})
+// writing 档位：即使全局特化层关闭，也会注入写作规范。
+func TestModelProfileWritingInjectsPreset(t *testing.T) {
+	store := overlayTestStore{enabled: true, profiles: map[string]config.ModelProfileSettings{
+		"deepseek-writer": {Found: true, Style: config.ModelProfileStyleWriting},
+	}}
+	out := applyWithProfile(t, &auth.RequestAuth{}, store, "deepseek-writer")
+	want := promptcompat.BuildPromptProfilePrompt(config.PromptProfileNameWriting, "")
+	if !strings.Contains(out.FinalPrompt, want) {
+		t.Fatalf("expected writing preset body, got %q", out.FinalPrompt)
+	}
+	if !strings.Contains(out.FinalPrompt, "## 执行要求") {
+		t.Fatalf("writing style keeps the v2 continuation, got %q", out.FinalPrompt)
 	}
 }
 
-// 模型提示词与技能层合并成一个 System 段，顺序：模型 → 技能。
-func TestModelProfilePromptCombinesWithSkills(t *testing.T) {
-	svc := Service{
-		Store: overlayTestStore{
-			enabled:      true,
-			modelPrompts: map[string]string{"deepseek-writer": "model spec"},
-		},
-		DS: &skillsTestUploader{},
+// v2 / legacy 档位：只切内置文案，不注入特化层。
+func TestModelProfileV2AndLegacySwapBuiltinText(t *testing.T) {
+	store := overlayTestStore{enabled: true, profiles: map[string]config.ModelProfileSettings{
+		"deepseek-v2":     {Found: true, Style: config.ModelProfileStyleV2},
+		"deepseek-legacy": {Found: true, Style: config.ModelProfileStyleLegacy},
+	}}
+	// 全局是 legacy，但该模型强制 v2。
+	a := &auth.RequestAuth{
+		PromptStyle:   config.PromptStyleLegacy,
+		PromptProfile: config.PromptProfileSettings{Enabled: true, Name: config.PromptProfileNameDefault},
+	}
+	v2Out := applyWithProfile(t, a, store, "deepseek-v2")
+	if !strings.Contains(v2Out.FinalPrompt, "## 执行要求") {
+		t.Fatalf("v2 profile should use the v2 continuation, got %q", v2Out.FinalPrompt)
+	}
+	if strings.Contains(v2Out.FinalPrompt, "System: ") {
+		t.Fatalf("v2 profile must not inject an overlay, got %q", v2Out.FinalPrompt)
+	}
+
+	// 反过来：全局 v2，该模型强制 legacy。
+	b := &auth.RequestAuth{PromptStyle: config.PromptStyleV2}
+	legacyOut := applyWithProfile(t, b, store, "deepseek-legacy")
+	if strings.Contains(legacyOut.FinalPrompt, "## 执行要求") {
+		t.Fatalf("legacy profile should use the short continuation, got %q", legacyOut.FinalPrompt)
 	}
+	if strings.Contains(legacyOut.FinalPrompt, promptcompat.ThinkingInjectionMarker) {
+		t.Fatalf("legacy profile must not inject the builtin contract, got %q", legacyOut.FinalPrompt)
+	}
+}
+
+// inherit 档位（以及未命中）：完全跟随全局。
+func TestModelProfileInheritFollowsGlobal(t *testing.T) {
+	store := overlayTestStore{enabled: true, profiles: map[string]config.ModelProfileSettings{
+		"deepseek-plain": {Found: true, Style: config.ModelProfileStyleInherit},
+	}}
+	a := &auth.RequestAuth{
+		PromptStyle:   config.PromptStyleLegacy,
+		PromptProfile: config.PromptProfileSettings{Enabled: true, Name: config.PromptProfileNameDefault},
+	}
+	for _, model := range []string{"deepseek-plain", "deepseek-v4-flash"} {
+		out := applyWithProfile(t, a, store, model)
+		if !strings.Contains(out.FinalPrompt, promptcompat.BuildPromptProfilePrompt(config.PromptProfileNameDefault, "")) {
+			t.Fatalf("model %q should follow the global profile, got %q", model, out.FinalPrompt)
+		}
+		if strings.Contains(out.FinalPrompt, "## 执行要求") {
+			t.Fatalf("model %q should follow the global legacy style, got %q", model, out.FinalPrompt)
+		}
+	}
+}
+
+// 模型正文与技能层合并成一个 System 段，顺序：模型 → 技能。
+func TestModelProfileCustomCombinesWithSkills(t *testing.T) {
+	store := overlayTestStore{enabled: true, profiles: map[string]config.ModelProfileSettings{
+		"deepseek-writer": {Found: true, Style: config.ModelProfileStyleCustom, Prompt: "model spec"},
+	}}
 	a := &auth.RequestAuth{
 		Skills: config.SkillSettings{
 			Enabled: true,
 			Items:   []config.SkillConfig{{ID: "sql-report", Body: "sql body"}},
 		},
 	}
-	out, err := svc.ApplyCurrentInputFile(context.Background(), a, overlayTestRequest("deepseek-writer"))
-	if err != nil {
-		t.Fatalf("apply current input file failed: %v", err)
-	}
+	out := applyWithProfile(t, a, store, "deepseek-writer")
 	if got := strings.Count(out.FinalPrompt, "System: "); got != 1 {
 		t.Fatalf("expected one System marker, got %d in %q", got, out.FinalPrompt)
 	}
@@ -133,7 +161,7 @@ func TestModelProfilePromptCombinesWithSkills(t *testing.T) {
 	}
 }
 
-// 旧版风格：历史拆分后既不带「执行要求」，也不注入内置执行契约。
+// 旧版风格（全局）：拆分后既不带「执行要求」，也不注入内置执行契约。
 func TestLegacyStyleKeepsSplitPromptMinimal(t *testing.T) {
 	svc := Service{Store: overlayTestStore{enabled: true}, DS: &skillsTestUploader{}}
 	a := &auth.RequestAuth{
@@ -156,7 +184,6 @@ func TestLegacyStyleKeepsSplitPromptMinimal(t *testing.T) {
 	if !strings.Contains(out.FinalPrompt, out.CurrentInputFilename) {
 		t.Fatalf("expected continuation to reference the context file, got %q", out.FinalPrompt)
 	}
-	// 对话记录文件里也不应出现契约：旧版风格下压根没注入。
 	if strings.Contains(out.HistoryText, promptcompat.ThinkingInjectionMarker) {
 		t.Fatalf("legacy transcript must not carry the contract, got %q", out.HistoryText)
 	}

internal/promptcompat/prompt_texts.go

@@ -159,8 +159,15 @@ const promptProfileWritingBody = "以下内容是服务端补充的写作规范
 	"4. 篇幅按用户要求执行；没要求时按内容需要来，不注水、不堆砌同义句。\n" +
 	"5. 保持上下文连续：人名、术语、编号、称呼与上文一致；改稿时说明改动了哪些地方。\n" +
 	"6. 语言跟随用户；中文写作不用「作为AI」「总之」这类套话收尾。\n\n" +
+	"## 不要照搬指令原文\n" +
+	"1. 把用户的指令当成要求，不要当成正文素材：要求句、约束说明、括号里的备注，都不要原样搬进正文，" +
+	"也不要复述「请写一篇…」「要求：…」这类指令框架。\n" +
+	"2. 这一条不适用于用户明确要求原样出现的内容：点名要照抄的句子、口号、标题、术语、数据、引文，" +
+	"以及必须出现的关键词，都要一字不改地写进去，位置按用户要求安排。\n" +
+	"3. 拿不准时分辨这句是「让我怎么写」还是「要我写什么」：前者转成正文表达，后者原样保留。\n\n" +
 	"## 交付前自检\n" +
-	"逐条核对：篇幅是否符合、硬约束是否满足、有没有漏掉用户点名要写的点、有没有出现用户明确禁止的写法。" +
+	"逐条核对：篇幅是否符合、硬约束是否满足、有没有漏掉用户点名要写的点、有没有出现用户明确禁止的写法、" +
+	"有没有把指令原文抄成正文。" +
 	"发现缺项先补齐再交付；这份自检只在内部完成，不要输出核对过程。"
 
 // BuildPromptProfilePrompt 返回特化层正文。

internal/promptcompat/prompt_texts_test.go

@@ -166,6 +166,7 @@ func TestWritingPresetCoversTheTaskComplaints(t *testing.T) {
 		"不要输出核对过程",   // 自检不外泄
 		"分清虚构与非虚构",   // 不编造只针对非虚构，别把创作能力压死
 		"虚构创作可以按设定发挥",
+		"不要照搬指令原文", // 不把指令原样抄成正文
 	} {
 		if !strings.Contains(body, want) {
 			t.Fatalf("writing preset missing %q: %q", want, body)
@@ -173,6 +174,28 @@ func TestWritingPresetCoversTheTaskComplaints(t *testing.T) {
 	}
 }
 
+// 「不要照搬指令原文」必须自带两个例外，否则模型会把「必须原样出现的句子/关键词」也一起删掉。
+func TestWritingPresetAntiCopyKeepsUserMandatedText(t *testing.T) {
+	body := BuildPromptProfilePrompt(config.PromptProfileNameWriting, "")
+	for _, want := range []string{
+		"不适用于用户明确要求原样出现的内容",
+		"点名要照抄的句子",
+		"必须出现的关键词",
+		"一字不改地写进去",
+		"「让我怎么写」还是「要我写什么」",
+	} {
+		if !strings.Contains(body, want) {
+			t.Fatalf("writing preset anti-copy rule missing %q: %q", want, body)
+		}
+	}
+	// 不允许写成笼统的「禁止引用用户原话」：那会连用户点名要写的内容一起禁掉。
+	for _, banned := range []string{"禁止引用", "不得使用用户的措辞", "不准出现用户的词"} {
+		if strings.Contains(body, banned) {
+			t.Fatalf("anti-copy rule must not be blanket (%q): %q", banned, body)
+		}
+	}
+}
+
 // 通用预设对每一次请求生效（含闲聊），所以不允许写死交付物文风；语气交给「跟着对方走」。
 func TestDefaultPresetStaysChatFriendly(t *testing.T) {
 	body := BuildPromptProfilePrompt(config.PromptProfileNameDefault, "")

webui/src/features/models/ModelAliasesSection.jsx

@@ -0,0 +1,101 @@
+import { useCallback, useEffect, useState } from 'react'
+import { Save } from 'lucide-react'
+
+import { fetchSettings, putSettings } from '../settings/settingsApi'
+
+// 模型别名在 /admin/settings 里是一个 JSON 对象；这里单独读写，避免与设置页表单耦合。
+export default function ModelAliasesSection({ t, apiFetch, onMessage }) {
+    const [text, setText] = useState('{}')
+    const [loading, setLoading] = useState(false)
+    const [saving, setSaving] = useState(false)
+    const [error, setError] = useState('')
+
+    const load = useCallback(async () => {
+        setLoading(true)
+        try {
+            const { res, data } = await fetchSettings(apiFetch, t)
+            if (!res.ok) {
+                setError(data?.detail || t('settings.loadFailed'))
+                return
+            }
+            const aliases = data?.config_snapshot?.model_aliases || data?.model_aliases || {}
+            setText(JSON.stringify(aliases || {}, null, 2))
+            setError('')
+        } catch (_e) {
+            setError(t('settings.loadFailed'))
+        } finally {
+            setLoading(false)
+        }
+    }, [apiFetch, t])
+
+    useEffect(() => {
+        load()
+    }, [load])
+
+    const save = useCallback(async () => {
+        let parsed
+        try {
+            parsed = JSON.parse(text || '{}')
+        } catch (_e) {
+            const detail = t('settings.invalidJsonField', { field: 'model_aliases' })
+            setError(detail)
+            onMessage?.('error', detail)
+            return
+        }
+        if (!parsed || typeof parsed !== 'object' || Array.isArray(parsed)) {
+            const detail = t('settings.invalidJsonField', { field: 'model_aliases' })
+            setError(detail)
+            onMessage?.('error', detail)
+            return
+        }
+        setSaving(true)
+        try {
+            const { res, data } = await putSettings(apiFetch, { model_aliases: parsed })
+            if (!res.ok) {
+                const detail = data?.detail || t('settings.saveFailed')
+                setError(detail)
+                onMessage?.('error', detail)
+                return
+            }
+            setError('')
+            onMessage?.('success', t('settings.modelAliasesSaved'))
+        } catch (_e) {
+            setError(t('settings.saveFailed'))
+            onMessage?.('error', t('settings.saveFailed'))
+        } finally {
+            setSaving(false)
+        }
+    }, [apiFetch, onMessage, text, t])
+
+    return (
+        <div className="bg-card border border-border rounded-xl p-5 space-y-4">
+            <div className="space-y-1">
+                <h3 className="font-semibold">{t('settings.modelTitle')}</h3>
+                <p className="text-xs text-muted-foreground">{t('settings.modelAliasesDesc')}</p>
+            </div>
+            {error && <p className="text-xs text-destructive">{error}</p>}
+            <label className="text-sm space-y-2 block">
+                <span className="text-muted-foreground">{t('settings.modelAliases')}</span>
+                <textarea
+                    value={text}
+                    onChange={(e) => setText(e.target.value)}
+                    rows={12}
+                    spellCheck={false}
+                    className="w-full bg-background border border-border rounded-lg px-3 py-2 font-mono text-xs"
+                />
+            </label>
+            <div className="flex items-center gap-2">
+                <button
+                    type="button"
+                    onClick={save}
+                    disabled={saving || loading}
+                    className="h-9 px-3 rounded-lg bg-primary text-primary-foreground text-sm hover:bg-primary/90 disabled:opacity-50 flex items-center gap-1"
+                >
+                    <Save className="w-4 h-4" />
+                    {saving ? t('settings.saving') : t('settings.modelAliasesSave')}
+                </button>
+                <span className="text-xs text-muted-foreground">{t('settings.modelAliasesHint')}</span>
+            </div>
+        </div>
+    )
+}

webui/src/features/models/ModelsContainer.jsx

@@ -0,0 +1,21 @@
+import { useI18n } from '../../i18n'
+import ModelProfilesSection from '../settings/ModelProfilesSection'
+import ModelAliasesSection from './ModelAliasesSection'
+
+export default function ModelsContainer({ onMessage, authFetch }) {
+    const { t } = useI18n()
+    const apiFetch = authFetch || fetch
+
+    return (
+        <div className="space-y-6">
+            <div className="space-y-1">
+                <h1 className="text-lg font-semibold">{t('modelsPage.title')}</h1>
+                <p className="text-sm text-muted-foreground">{t('modelsPage.desc')}</p>
+            </div>
+
+            <ModelProfilesSection t={t} apiFetch={apiFetch} onMessage={onMessage} />
+
+            <ModelAliasesSection t={t} apiFetch={apiFetch} onMessage={onMessage} />
+        </div>
+    )
+}

webui/src/features/settings/BehaviorSection.jsx

@@ -28,84 +28,6 @@ export default function BehaviorSection({ t, form, setForm }) {
                         className="w-full bg-background border border-border rounded-lg px-3 py-2"
                     />
                 </label>
-                <label className="flex items-start gap-3 rounded-lg border border-border bg-background/60 p-4">
-                    <input
-                        type="checkbox"
-                        checked={Boolean(form.thinking_injection?.enabled ?? true)}
-                        onChange={(e) => setForm((prev) => ({
-                            ...prev,
-                            thinking_injection: {
-                                ...prev.thinking_injection,
-                                enabled: e.target.checked,
-                            },
-                        }))}
-                        className="mt-1 h-4 w-4 rounded border-border"
-                    />
-                    <div className="space-y-1">
-                        <span className="text-sm font-medium block">{t('settings.thinkingInjectionEnabled')}</span>
-                        <span className="text-xs text-muted-foreground block">{t('settings.thinkingInjectionDesc')}</span>
-                    </div>
-                </label>
-                <div className="text-sm space-y-2 md:col-span-2">
-                    <div className="flex items-center justify-between gap-3">
-                        <label htmlFor="thinking-injection-prompt" className="text-muted-foreground">
-                            {t('settings.thinkingInjectionPrompt')}
-                        </label>
-                        <button
-                            type="button"
-                            disabled={!form.thinking_injection?.prompt}
-                            onClick={() => setForm((prev) => ({
-                                ...prev,
-                                thinking_injection: {
-                                    ...prev.thinking_injection,
-                                    // 清空即回到内置默认提示词，开关状态保持不变。
-                                    prompt: '',
-                                },
-                            }))}
-                            className="h-8 px-2.5 rounded-lg border border-border bg-background text-xs text-muted-foreground hover:text-foreground hover:bg-secondary/70 disabled:opacity-50 disabled:hover:text-muted-foreground disabled:hover:bg-background"
-                        >
-                            {t('settings.thinkingInjectionRestoreDefault')}
-                        </button>
-                    </div>
-                    <textarea
-                        id="thinking-injection-prompt"
-                        rows={8}
-                        value={form.thinking_injection?.prompt || ''}
-                        placeholder={form.thinking_injection?.default_prompt || ''}
-                        onChange={(e) => setForm((prev) => ({
-                            ...prev,
-                            thinking_injection: {
-                                ...prev.thinking_injection,
-                                prompt: e.target.value,
-                            },
-                        }))}
-                        className="w-full bg-background border border-border rounded-lg px-3 py-2 resize-y min-h-48 text-sm leading-6"
-                    />
-                    <p className="text-xs text-muted-foreground">{t('settings.thinkingInjectionPromptHelp')}</p>
-                </div>
-                <div className="text-sm space-y-2 md:col-span-2">
-                    <label htmlFor="prompt-style" className="text-muted-foreground">
-                        {t('settings.promptStyleTitle')}
-                    </label>
-                    <select
-                        id="prompt-style"
-                        value={form.prompt_style?.mode || 'v2'}
-                        onChange={(e) => setForm((prev) => ({
-                            ...prev,
-                            prompt_style: { ...prev.prompt_style, mode: e.target.value },
-                        }))}
-                        className="w-full bg-background border border-border rounded-lg px-3 py-2"
-                    >
-                        {(form.prompt_style?.modes || ['v2', 'legacy']).map((mode) => (
-                            <option key={mode} value={mode}>
-                                {mode === 'legacy'
-                                    ? t('settings.promptStyleLegacy')
-                                    : t('settings.promptStyleV2')}
-                            </option>
-                        ))}
-                    </select>
-                    <p className="text-xs text-muted-foreground">{t('settings.promptStyleHelp')}</p>
-                </div>
             </div>
         </div>
     )

webui/src/features/settings/ModelProfilesSection.jsx

@@ -10,8 +10,23 @@ const FALLBACK_BASE_MODELS = [
     'deepseek-v4-flash-search-nothinking',
 ]
 
-function emptyProfile() {
-    return { id: '', base: FALLBACK_BASE_MODELS[0], name: '', remark: '', prompt: '' }
+const STYLE_LABEL_KEYS = {
+    inherit: 'settings.modelProfilesStyleInherit',
+    v2: 'settings.modelProfilesStyleV2',
+    writing: 'settings.modelProfilesStyleWriting',
+    legacy: 'settings.modelProfilesStyleLegacy',
+    custom: 'settings.modelProfilesStyleCustom',
+}
+
+function emptyProfile(styleNames) {
+    return {
+        id: '',
+        base: FALLBACK_BASE_MODELS[0],
+        name: '',
+        remark: '',
+        style: styleNames.includes('inherit') ? 'inherit' : styleNames[0] || 'inherit',
+        prompt: '',
+    }
 }
 
 function normalizeItems(data) {
@@ -21,6 +36,8 @@ function normalizeItems(data) {
         base: String(item?.base || FALLBACK_BASE_MODELS[0]),
         name: String(item?.name || ''),
         remark: String(item?.remark || ''),
+        // 旧配置只有 prompt：展示成 custom，与后端 EffectiveStyle 的解释一致。
+        style: String(item?.style || (item?.prompt ? 'custom' : 'inherit')),
         prompt: String(item?.prompt || ''),
     }))
 }
@@ -30,6 +47,7 @@ export default function ModelProfilesSection({ t, apiFetch, onMessage }) {
     const [saving, setSaving] = useState(false)
     const [items, setItems] = useState([])
     const [baseModels, setBaseModels] = useState(FALLBACK_BASE_MODELS)
+    const [styleNames, setStyleNames] = useState(Object.keys(STYLE_LABEL_KEYS))
     const [maxCount, setMaxCount] = useState(50)
     const [maxPromptChars, setMaxPromptChars] = useState(20000)
     const [expandedRows, setExpandedRows] = useState({})
@@ -47,6 +65,9 @@ export default function ModelProfilesSection({ t, apiFetch, onMessage }) {
             if (Array.isArray(data?.base_models) && data.base_models.length > 0) {
                 setBaseModels(data.base_models)
             }
+            if (Array.isArray(data?.style_names) && data.style_names.length > 0) {
+                setStyleNames(data.style_names)
+            }
             if (Number(data?.max_count) > 0) {
                 setMaxCount(Number(data.max_count))
             }
@@ -75,8 +96,8 @@ export default function ModelProfilesSection({ t, apiFetch, onMessage }) {
     }, [])
 
     const addItem = useCallback(() => {
-        setItems((prev) => (prev.length >= maxCount ? prev : [...prev, emptyProfile()]))
-    }, [maxCount])
+        setItems((prev) => (prev.length >= maxCount ? prev : [...prev, emptyProfile(styleNames)]))
+    }, [maxCount, styleNames])
 
     const toggleRow = useCallback((index) => {
         setExpandedRows((prev) => ({ ...prev, [index]: !prev[index] }))
@@ -123,6 +144,7 @@ export default function ModelProfilesSection({ t, apiFetch, onMessage }) {
                     base: String(item.base || '').trim(),
                     name: String(item.name || '').trim(),
                     remark: String(item.remark || '').trim(),
+                    style: String(item.style || 'inherit'),
                     prompt: String(item.prompt || ''),
                 })),
             }
@@ -192,6 +214,20 @@ export default function ModelProfilesSection({ t, apiFetch, onMessage }) {
                                         ))}
                                     </select>
                                 </label>
+                                <label className="text-xs space-y-1">
+                                    <span className="text-muted-foreground block">{t('settings.modelProfilesStyle')}</span>
+                                    <select
+                                        value={item.style}
+                                        onChange={(e) => updateItem(index, { style: e.target.value })}
+                                        className="w-full bg-background border border-border rounded-lg px-3 py-2 text-sm"
+                                    >
+                                        {styleNames.map((style) => (
+                                            <option key={style} value={style}>
+                                                {STYLE_LABEL_KEYS[style] ? t(STYLE_LABEL_KEYS[style]) : style}
+                                            </option>
+                                        ))}
+                                    </select>
+                                </label>
                                 <label className="text-xs space-y-1">
                                     <span className="text-muted-foreground block">{t('settings.modelProfilesName')}</span>
                                     <input
@@ -202,16 +238,12 @@ export default function ModelProfilesSection({ t, apiFetch, onMessage }) {
                                         className="w-full bg-background border border-border rounded-lg px-3 py-2 text-sm"
                                     />
                                 </label>
-                                <label className="text-xs space-y-1">
-                                    <span className="text-muted-foreground block">{t('settings.modelProfilesRemark')}</span>
-                                    <input
-                                        type="text"
-                                        value={item.remark}
-                                        onChange={(e) => updateItem(index, { remark: e.target.value })}
-                                        className="w-full bg-background border border-border rounded-lg px-3 py-2 text-sm"
-                                    />
-                                </label>
                             </div>
+                            <p className="text-xs text-muted-foreground">
+                                {STYLE_LABEL_KEYS[item.style]
+                                    ? t(`${STYLE_LABEL_KEYS[item.style]}Hint`)
+                                    : t('settings.modelProfilesStyleInheritHint')}
+                            </p>
                             <div className="flex items-center justify-between gap-2">
                                 <button
                                     type="button"
@@ -219,7 +251,7 @@ export default function ModelProfilesSection({ t, apiFetch, onMessage }) {
                                     className="text-xs text-muted-foreground hover:text-foreground flex items-center gap-1"
                                 >
                                     {expanded ? <ChevronUp className="w-3.5 h-3.5" /> : <ChevronDown className="w-3.5 h-3.5" />}
-                                    {expanded ? t('settings.skillsCollapse') : t('settings.modelProfilesPrompt')}
+                                    {expanded ? t('settings.skillsCollapse') : t('settings.modelProfilesExtra')}
                                 </button>
                                 <button
                                     type="button"
@@ -231,17 +263,31 @@ export default function ModelProfilesSection({ t, apiFetch, onMessage }) {
                                 </button>
                             </div>
                             {expanded && (
-                                <div className="space-y-1">
-                                    <textarea
-                                        rows={8}
-                                        value={item.prompt}
-                                        onChange={(e) => updateItem(index, { prompt: e.target.value })}
-                                        placeholder={t('settings.modelProfilesPromptPlaceholder')}
-                                        className="w-full bg-background border border-border rounded-lg px-3 py-2 resize-y min-h-32 text-sm leading-6"
-                                    />
-                                    <p className="text-xs text-muted-foreground">
-                                        {t('settings.modelProfilesPromptHint')}
-                                    </p>
+                                <div className="space-y-3">
+                                    <label className="text-xs space-y-1 block">
+                                        <span className="text-muted-foreground block">{t('settings.modelProfilesRemark')}</span>
+                                        <input
+                                            type="text"
+                                            value={item.remark}
+                                            onChange={(e) => updateItem(index, { remark: e.target.value })}
+                                            className="w-full bg-background border border-border rounded-lg px-3 py-2 text-sm"
+                                        />
+                                    </label>
+                                    <div className="space-y-1">
+                                        <span className="text-xs text-muted-foreground block">{t('settings.modelProfilesPrompt')}</span>
+                                        <textarea
+                                            rows={8}
+                                            value={item.prompt}
+                                            onChange={(e) => updateItem(index, { prompt: e.target.value })}
+                                            placeholder={t('settings.modelProfilesPromptPlaceholder')}
+                                            className="w-full bg-background border border-border rounded-lg px-3 py-2 resize-y min-h-32 text-sm leading-6"
+                                        />
+                                        <p className="text-xs text-muted-foreground">
+                                            {item.style === 'custom'
+                                                ? t('settings.modelProfilesPromptHintCustom')
+                                                : t('settings.modelProfilesPromptHintPreset')}
+                                        </p>
+                                    </div>
                                 </div>
                             )}
                         </div>

webui/src/features/settings/ModelSection.jsx

@@ -1,16 +0,0 @@
-export default function ModelSection({ t, form, setForm }) {
-    return (
-        <div className="bg-card border border-border rounded-xl p-5 space-y-4">
-            <h3 className="font-semibold">{t('settings.modelTitle')}</h3>
-            <label className="text-sm space-y-2 block">
-                <span className="text-muted-foreground">{t('settings.modelAliases')}</span>
-                <textarea
-                    value={form.model_aliases_text}
-                    onChange={(e) => setForm((prev) => ({ ...prev, model_aliases_text: e.target.value }))}
-                    rows={12}
-                    className="w-full bg-background border border-border rounded-lg px-3 py-2 font-mono text-xs"
-                />
-            </label>
-        </div>
-    )
-}

webui/src/features/settings/PromptInjectionSection.jsx

@@ -0,0 +1,87 @@
+export default function PromptInjectionSection({ t, form, setForm }) {
+    return (
+        <div className="bg-card border border-border rounded-xl p-5 space-y-4">
+            <div className="space-y-1">
+                <h3 className="font-semibold">{t('settings.promptInjectionTitle')}</h3>
+                <p className="text-xs text-muted-foreground">{t('settings.promptInjectionDesc')}</p>
+            </div>
+            <div className="grid grid-cols-1 md:grid-cols-2 gap-4">
+                <label className="text-sm space-y-2">
+                    <span className="text-muted-foreground">{t('settings.promptStyleTitle')}</span>
+                    <select
+                        value={form.prompt_style?.mode || 'v2'}
+                        onChange={(e) => setForm((prev) => ({
+                            ...prev,
+                            prompt_style: { ...prev.prompt_style, mode: e.target.value },
+                        }))}
+                        className="w-full bg-background border border-border rounded-lg px-3 py-2"
+                    >
+                        {(form.prompt_style?.modes || ['v2', 'legacy']).map((mode) => (
+                            <option key={mode} value={mode}>
+                                {mode === 'legacy'
+                                    ? t('settings.promptStyleLegacy')
+                                    : t('settings.promptStyleV2')}
+                            </option>
+                        ))}
+                    </select>
+                    <span className="text-xs text-muted-foreground block">{t('settings.promptStyleHelp')}</span>
+                </label>
+                <label className="flex items-start gap-3 rounded-lg border border-border bg-background/60 p-4">
+                    <input
+                        type="checkbox"
+                        checked={Boolean(form.thinking_injection?.enabled ?? true)}
+                        onChange={(e) => setForm((prev) => ({
+                            ...prev,
+                            thinking_injection: {
+                                ...prev.thinking_injection,
+                                enabled: e.target.checked,
+                            },
+                        }))}
+                        className="mt-1 h-4 w-4 rounded border-border"
+                    />
+                    <div className="space-y-1">
+                        <span className="text-sm font-medium block">{t('settings.thinkingInjectionEnabled')}</span>
+                        <span className="text-xs text-muted-foreground block">{t('settings.thinkingInjectionDesc')}</span>
+                    </div>
+                </label>
+                <div className="text-sm space-y-2 md:col-span-2">
+                    <div className="flex items-center justify-between gap-3">
+                        <label htmlFor="thinking-injection-prompt" className="text-muted-foreground">
+                            {t('settings.thinkingInjectionPrompt')}
+                        </label>
+                        <button
+                            type="button"
+                            disabled={!form.thinking_injection?.prompt}
+                            onClick={() => setForm((prev) => ({
+                                ...prev,
+                                thinking_injection: {
+                                    ...prev.thinking_injection,
+                                    // 清空即回到内置默认提示词，开关状态保持不变。
+                                    prompt: '',
+                                },
+                            }))}
+                            className="h-8 px-2.5 rounded-lg border border-border bg-background text-xs text-muted-foreground hover:text-foreground hover:bg-secondary/70 disabled:opacity-50 disabled:hover:text-muted-foreground disabled:hover:bg-background"
+                        >
+                            {t('settings.thinkingInjectionRestoreDefault')}
+                        </button>
+                    </div>
+                    <textarea
+                        id="thinking-injection-prompt"
+                        rows={8}
+                        value={form.thinking_injection?.prompt || ''}
+                        placeholder={form.thinking_injection?.default_prompt || ''}
+                        onChange={(e) => setForm((prev) => ({
+                            ...prev,
+                            thinking_injection: {
+                                ...prev.thinking_injection,
+                                prompt: e.target.value,
+                            },
+                        }))}
+                        className="w-full bg-background border border-border rounded-lg px-3 py-2 resize-y min-h-48 text-sm leading-6"
+                    />
+                    <p className="text-xs text-muted-foreground">{t('settings.thinkingInjectionPromptHelp')}</p>
+                </div>
+            </div>
+        </div>
+    )
+}

webui/src/features/settings/SettingsContainer.jsx

@@ -2,18 +2,18 @@ import { AlertTriangle, Save } from 'lucide-react'
 
 import { useI18n } from '../../i18n'
 import { useSettingsForm } from './useSettingsForm'
+import SettingsGroup from './SettingsGroup'
 import SecuritySection from './SecuritySection'
 import PlatformSection from './PlatformSection'
 import RuntimeSection from './RuntimeSection'
 import BehaviorSection from './BehaviorSection'
+import PromptInjectionSection from './PromptInjectionSection'
 import PromptProfileSection from './PromptProfileSection'
 import ToolCallsSection from './ToolCallsSection'
 import SkillsSection from './SkillsSection'
-import ModelProfilesSection from './ModelProfilesSection'
 import CurrentInputFileSection from './CurrentInputFileSection'
 import FileAuditSection from './FileAuditSection'
 import AutoDeleteSection from './AutoDeleteSection'
-import ModelSection from './ModelSection'
 import BackupSection from './BackupSection'
 
 export default function SettingsContainer({ onRefresh, onMessage, authFetch, onForceLogout, isVercel = false }) {
@@ -56,7 +56,7 @@ export default function SettingsContainer({ onRefresh, onMessage, authFetch, onF
     })
 
     return (
-        <div className="space-y-6">
+        <div className="space-y-8">
             {autoFetchPaused && (
                 <div className="p-4 rounded-lg border border-destructive/30 bg-destructive/10 text-destructive flex items-center justify-between gap-4">
                     <div className="flex items-center gap-2">
@@ -87,51 +87,55 @@ export default function SettingsContainer({ onRefresh, onMessage, authFetch, onF
                 </div>
             )}
 
-            <SecuritySection
-                t={t}
-                form={form}
-                setForm={setForm}
-                newPassword={newPassword}
-                setNewPassword={setNewPassword}
-                changingPassword={changingPassword}
-                onUpdatePassword={updatePassword}
-            />
-
-            <PlatformSection t={t} form={form} setForm={setForm} />
-
-            <RuntimeSection t={t} form={form} setForm={setForm} />
-
-            <BehaviorSection t={t} form={form} setForm={setForm} />
-
-            <PromptProfileSection t={t} form={form} setForm={setForm} />
-
-            <ToolCallsSection t={t} form={form} setForm={setForm} />
-
-            <ModelProfilesSection t={t} apiFetch={apiFetch} onMessage={onMessage} />
-
-            <SkillsSection t={t} apiFetch={apiFetch} onMessage={onMessage} />
-
-            <CurrentInputFileSection t={t} form={form} setForm={setForm} />
-
-            <FileAuditSection t={t} form={form} setForm={setForm} />
-
-            <AutoDeleteSection t={t} form={form} setForm={setForm} />
-
-            <ModelSection t={t} form={form} setForm={setForm} />
-
-            <BackupSection
-                t={t}
-                importMode={importMode}
-                setImportMode={setImportMode}
-                importing={importing}
-                onLoadExportData={loadExportData}
-                onDownloadExportFile={downloadExportFile}
-                onImport={doImport}
-                onImportFileChange={loadImportFile}
-                importText={importText}
-                setImportText={setImportText}
-                exportData={exportData}
-            />
+            <SettingsGroup title={t('settings.groupPrompt')} description={t('settings.groupPromptDesc')}>
+                <PromptInjectionSection t={t} form={form} setForm={setForm} />
+                <PromptProfileSection t={t} form={form} setForm={setForm} />
+            </SettingsGroup>
+
+            <SettingsGroup title={t('settings.groupContext')} description={t('settings.groupContextDesc')}>
+                <CurrentInputFileSection t={t} form={form} setForm={setForm} />
+                <SkillsSection t={t} apiFetch={apiFetch} onMessage={onMessage} />
+            </SettingsGroup>
+
+            <SettingsGroup title={t('settings.groupRuntime')} description={t('settings.groupRuntimeDesc')}>
+                <RuntimeSection t={t} form={form} setForm={setForm} />
+                <BehaviorSection t={t} form={form} setForm={setForm} />
+                <PlatformSection t={t} form={form} setForm={setForm} />
+                <AutoDeleteSection t={t} form={form} setForm={setForm} />
+                <FileAuditSection t={t} form={form} setForm={setForm} />
+            </SettingsGroup>
+
+            <SettingsGroup title={t('settings.groupTools')} description={t('settings.groupToolsDesc')}>
+                <ToolCallsSection t={t} form={form} setForm={setForm} />
+            </SettingsGroup>
+
+            <SettingsGroup title={t('settings.groupSecurity')} description={t('settings.groupSecurityDesc')}>
+                <SecuritySection
+                    t={t}
+                    form={form}
+                    setForm={setForm}
+                    newPassword={newPassword}
+                    setNewPassword={setNewPassword}
+                    changingPassword={changingPassword}
+                    onUpdatePassword={updatePassword}
+                />
+            </SettingsGroup>
+
+            <SettingsGroup title={t('settings.groupData')} description={t('settings.groupDataDesc')}>
+                <BackupSection
+                    t={t}
+                    importMode={importMode}
+                    setImportMode={setImportMode}
+                    importing={importing}
+                    onLoadExportData={loadExportData}
+                    onDownloadExportFile={downloadExportFile}
+                    onImport={doImport}
+                    onImportFileChange={loadImportFile}
+                    importText={importText}
+                    setImportText={setImportText}
+                    exportData={exportData}
+                />
+            </SettingsGroup>
 
             <div className="flex justify-end">
                 <button

webui/src/features/settings/SettingsGroup.jsx

@@ -0,0 +1,11 @@
+export default function SettingsGroup({ title, description, children }) {
+    return (
+        <section className="space-y-4">
+            <div className="space-y-1 border-l-2 border-primary/40 pl-3">
+                <h2 className="text-sm font-semibold tracking-wide">{title}</h2>
+                {description && <p className="text-xs text-muted-foreground">{description}</p>}
+            </div>
+            <div className="space-y-4">{children}</div>
+        </section>
+    )
+}

webui/src/layout/DashboardShell.jsx

@@ -11,6 +11,7 @@ import {
     Server,
     Users,
     Globe,
+    Layers,
     History,
     Loader2
 } from 'lucide-react'
@@ -25,6 +26,7 @@ const ChatHistoryContainer = lazy(() => import('../features/chatHistory/ChatHist
 const BatchImport = lazy(() => import('../components/BatchImport'))
 const VercelSyncContainer = lazy(() => import('../features/vercel/VercelSyncContainer'))
 const SettingsContainer = lazy(() => import('../features/settings/SettingsContainer'))
+const ModelsContainer = lazy(() => import('../features/models/ModelsContainer'))
 const ProxyManagerContainer = lazy(() => import('../features/proxy/ProxyManagerContainer'))
 
 function TabLoadingFallback({ label }) {
@@ -46,6 +48,7 @@ export default function DashboardShell({ token, onLogout, config, fetchConfig, s
 
     const navItems = [
         { id: 'accounts', label: t('nav.accounts.label'), icon: Users, description: t('nav.accounts.desc') },
+        { id: 'models', label: t('nav.models.label'), icon: Layers, description: t('nav.models.desc') },
         { id: 'proxies', label: t('nav.proxies.label'), icon: Globe, description: t('nav.proxies.desc') },
         { id: 'test', label: t('nav.test.label'), icon: Server, description: t('nav.test.desc') },
         { id: 'history', label: t('nav.history.label'), icon: History, description: t('nav.history.desc') },
@@ -111,6 +114,8 @@ export default function DashboardShell({ token, onLogout, config, fetchConfig, s
         switch (activeTab) {
             case 'accounts':
                 return <AccountManagerContainer config={config} onRefresh={fetchConfig} onMessage={showMessage} authFetch={authFetch} />
+            case 'models':
+                return <ModelsContainer onMessage={showMessage} authFetch={authFetch} />
             case 'proxies':
                 return <ProxyManagerContainer config={config} onRefresh={fetchConfig} onMessage={showMessage} authFetch={authFetch} />
             case 'test':

webui/src/locales/en.json

@@ -4,11 +4,19 @@
         "english": "English",
         "chinese": "中文"
     },
+    "modelsPage": {
+        "title": "Models",
+        "desc": "Expose several public model names on top of the same flash base model, each with its own style (follow global / v2 / writing / legacy / custom body). Clients pick a name; entries never affect each other."
+    },
     "nav": {
         "accounts": {
             "label": "Account Management",
             "desc": "Manage the DeepSeek account pool"
         },
+        "models": {
+            "label": "Models",
+            "desc": "Custom models and model aliases"
+        },
         "proxies": {
             "label": "Proxy IPs",
             "desc": "Manage outbound proxy nodes for accounts"
@@ -477,6 +485,20 @@
         "thinkingInjectionPrompt": "Thinking format prompt",
         "thinkingInjectionRestoreDefault": "Restore built-in default",
         "thinkingInjectionPromptHelp": "Leave empty to use the built-in default prompt shown as the input placeholder.",
+        "promptInjectionTitle": "Prompt style and thinking injection",
+        "promptInjectionDesc": "These two decide the built-in wording the model sees in long conversations. They only affect built-in prompts, never text you wrote yourself.",
+        "groupPrompt": "Prompt",
+        "groupPromptDesc": "Built-in style, thinking injection and the prompt profile live in this group.",
+        "groupContext": "Context and skills",
+        "groupContextDesc": "History-splitting policy and the plain-text skill library.",
+        "groupRuntime": "Runtime and behaviour",
+        "groupRuntimeDesc": "Concurrency, caching, platform fingerprint, session cleanup and file audit.",
+        "groupTools": "Tool calling",
+        "groupToolsDesc": "Master switch for DSML tool calling, off by default.",
+        "groupSecurity": "Security",
+        "groupSecurityDesc": "Admin panel password and JWT lifetime.",
+        "groupData": "Data and backup",
+        "groupDataDesc": "Export, restore and bulk import.",
         "promptStyleTitle": "Built-in prompt style",
         "promptStyleV2": "v2 (current: continuation carries requirements, contract injected)",
         "promptStyleLegacy": "legacy (short continuation, no built-in contract)",
@@ -492,6 +514,24 @@
         "promptProfilePrompt": "Custom body",
         "promptProfileRestoreDefault": "Restore built-in preset",
         "promptProfilePromptHelp": "Leave empty to use the selected preset's built-in body, shown as the placeholder. Custom text is capped at 20000 characters; when it conflicts with a user or client role definition, the user's setting wins.",
+        "modelProfilesStyle": "Style",
+        "modelProfilesStyleInherit": "Follow global",
+        "modelProfilesStyleV2": "v2 wording",
+        "modelProfilesStyleWriting": "Writing",
+        "modelProfilesStyleLegacy": "Legacy",
+        "modelProfilesStyleCustom": "Custom body",
+        "modelProfilesStyleInheritHint": "Follows the built-in prompt style and prompt profile from Settings; changing the global setting changes this model too.",
+        "modelProfilesStyleV2Hint": "Pins the v2 built-in wording (requirement-carrying continuation + execution contract) and injects no prompt profile.",
+        "modelProfilesStyleWritingHint": "v2 wording plus the built-in writing spec, good for long-form and rewrites.",
+        "modelProfilesStyleLegacyHint": "Legacy short wording, no execution contract — replies come out looser.",
+        "modelProfilesStyleCustomHint": "Uses the custom body below as the prompt profile; built-in wording still follows the global style.",
+        "modelProfilesExtra": "Remark and model body (expand to edit)",
+        "modelProfilesPromptHintCustom": "This body only applies when the style is \"Custom body\"; max 20000 characters.",
+        "modelProfilesPromptHintPreset": "The current style uses built-in wording or follows the global setting, so this body is ignored. Switch the style to \"Custom body\" to use it.",
+        "modelAliasesDesc": "Maps model names sent by clients onto models this project supports, e.g. point gpt-4o at deepseek-v4-flash. Custom models take precedence over these aliases.",
+        "modelAliasesSave": "Save aliases",
+        "modelAliasesSaved": "Model aliases saved",
+        "modelAliasesHint": "A JSON object: keys are the model names clients send, values are the target models.",
         "modelProfilesTitle": "Custom models",
         "modelProfilesDesc": "Expose several model names on top of the same flash base model, each with its own prompt. Clients pick a name, entries do not affect each other, and rollback is simpler than flipping a global switch. When the selected model carries a prompt it wins; otherwise the prompt profile above applies.",
         "modelProfilesCount": "{count} total",

webui/src/locales/zh.json

@@ -4,11 +4,19 @@
         "english": "English",
         "chinese": "中文"
     },
+    "modelsPage": {
+        "title": "模型列表",
+        "desc": "给同一个 flash 基模挂出多个对外模型名，每个名字可以独立选风格（应用全局 / v2 / 写作 / 旧版 / 自定义正文）。客户端按模型名选用，互不影响。"
+    },
     "nav": {
         "accounts": {
             "label": "账号管理",
             "desc": "管理 DeepSeek 账号池"
         },
+        "models": {
+            "label": "模型列表",
+            "desc": "自定义模型与模型别名"
+        },
         "proxies": {
             "label": "代理 IP",
             "desc": "管理账号可用的代理出口"
@@ -477,6 +485,20 @@
         "thinkingInjectionPrompt": "思考格式提示词",
         "thinkingInjectionRestoreDefault": "恢复内置默认提示词",
         "thinkingInjectionPromptHelp": "留空时使用内置默认提示词；默认内容会显示在输入框占位文本中。",
+        "promptInjectionTitle": "提示词风格与思考注入",
+        "promptInjectionDesc": "这两项决定模型在长对话里拿到的内置文案；只影响内置提示词，不会改动你自己填的内容。",
+        "groupPrompt": "提示词",
+        "groupPromptDesc": "内置风格、思考注入与提示词特化层，都在这一组里。",
+        "groupContext": "上下文与技能",
+        "groupContextDesc": "历史拆分策略与纯文本技能库。",
+        "groupRuntime": "运行与行为",
+        "groupRuntimeDesc": "并发、缓存、平台指纹、会话清理与文件审核。",
+        "groupTools": "工具调用",
+        "groupToolsDesc": "DSML 工具调用的总开关，默认关闭。",
+        "groupSecurity": "安全",
+        "groupSecurityDesc": "管理面板密码与 JWT 有效期。",
+        "groupData": "数据与备份",
+        "groupDataDesc": "配置导出、恢复与批量导入。",
         "promptStyleTitle": "内置提示词风格",
         "promptStyleV2": "v2（当前版：衔接指令带执行要求，注入执行契约）",
         "promptStyleLegacy": "legacy（旧版：短衔接指令，不注入执行契约）",
@@ -492,6 +514,24 @@
         "promptProfilePrompt": "自定义正文",
         "promptProfileRestoreDefault": "恢复内置预设",
         "promptProfilePromptHelp": "留空时使用所选预设的内置正文，占位文本里就是该正文；自定义内容上限 20000 字符。与用户或客户端设定的角色冲突时以用户设定为准。",
+        "modelProfilesStyle": "风格",
+        "modelProfilesStyleInherit": "应用全局",
+        "modelProfilesStyleV2": "v2 文案",
+        "modelProfilesStyleWriting": "写作",
+        "modelProfilesStyleLegacy": "旧版",
+        "modelProfilesStyleCustom": "自定义正文",
+        "modelProfilesStyleInheritHint": "跟随设置里的内置提示词风格与提示词特化层，全局改它就跟着改。",
+        "modelProfilesStyleV2Hint": "固定用 v2 内置文案（衔接指令带执行要求 + 执行契约），不注入特化层。",
+        "modelProfilesStyleWritingHint": "v2 文案 + 内置写作规范正文，适合长文与改稿。",
+        "modelProfilesStyleLegacyHint": "旧版短文案，不注入执行契约，回答更放得开。",
+        "modelProfilesStyleCustomHint": "用下面的自定义正文当特化层；内置文案风格仍跟随全局。",
+        "modelProfilesExtra": "备注与该模型正文（展开编辑）",
+        "modelProfilesPromptHintCustom": "这段正文只在风格为「自定义正文」时生效，上限 20000 字符。",
+        "modelProfilesPromptHintPreset": "当前风格使用内置文案或跟随全局，这里填的正文不会生效；要自定义请把风格改成「自定义正文」。",
+        "modelAliasesDesc": "把客户端发来的模型名映射到本项目支持的模型，例如把 gpt-4o 指向 deepseek-v4-flash。自定义模型列表优先于这里的映射。",
+        "modelAliasesSave": "保存别名",
+        "modelAliasesSaved": "模型别名已保存",
+        "modelAliasesHint": "JSON 对象，键是客户端请求的模型名，值是目标模型。",
         "modelProfilesTitle": "自定义模型（模型列表）",
         "modelProfilesDesc": "给同一个 flash 基模挂出多个对外模型名，每个名字带自己的提示词。客户端按模型名选用，互不影响，比改全局开关更好回滚。命中带提示词的模型时优先用它，没命中才回落到上面的提示词特化层。",
         "modelProfilesCount": "共 {count} 个",