@starreel/mcp 0.1.39 → 0.1.41

package/SKILL.md

@@ -180,6 +180,66 @@ content that will be rejected.
     key in an environment variable or secrets manager, never in code or logs.
     The 15-min token auto-re-exchanges; a leaked key is revoked in Settings.
 
+## Script fidelity — auto two-pass rewriting
+
+`rewrite_script` routes automatically based on what the user put into
+`set_script`:
+
+- **Source is already a screenplay** (scene headers / dialogue-line structure):
+  the platform runs a two-pass fidelity rewrite. The user's dialogue is
+  byte-locked by machine gates (a missing line is rejected and redone
+  internally — you never need to babysit this), and the AI adds **no** hooks,
+  twists, or emotional beats of its own. Dramaturgy gaps it detects come back
+  in `get_script` → `dramaturgy_suggestions`; relay them to the user and let
+  the user decide. Ignoring them blocks nothing.
+- **Source is a novel or outline**: the creative single-pass rewrite runs as
+  before (the AI builds hooks and emotional beats).
+
+Overrides via `update_project_settings`: `rewrite_pipeline`
+(`auto` default / `two_pass` force fidelity / `single_forced` force creative)
+and `fidelity_enforce: 1` (hard-reject any rewrite that drops a dialogue line,
+a cast member, or an action beat).
+
+When the user complains "the AI rewrote my script off course": ① confirm the
+full original went into `set_script`; ② set `rewrite_pipeline: "two_pass"` and
+rerun `rewrite_script`; ③ once the user approves a character's look, lock it
+with `update_character` `profile_locked: 1` so later extractions can't
+overwrite appearance (appearance anchors the portrait and face lock — an
+overwrite means face drift across the whole drama).
+
+## After the first rewrite — 点改优先 (point-edit, don't re-run)
+
+**`rewrite_script` is "start over from the source", not "give me another
+revision."** Once a draft exists, re-running it overwrites the current draft
+together with every fix already made — and the new version does **not**
+inherit what the previous one got right. Measured across three production
+runs of the same episode: a long VO block correctly split in v2 was merged
+back in v3, and era-correct wardrobe in v2 drifted a decade in v3. The
+kickoff response says so explicitly via `overwrites_existing_script: true`.
+
+So after the first successful rewrite, **every** change goes through
+`edit_rewritten_script` — free, seconds, deterministic:
+
+1. `get_script` → take the **full** `rewritten_script`.
+2. Change **only the scenes that need changing** — split an over-long scene,
+   fix a line, drop a `motif` that has no matching visual, add a missing
+   `[SFX]`, move a hand-held object out of the wardrobe segment into the
+   scene's `[道具]` line. Every other scene is copied back **verbatim**,
+   punctuation included.
+3. Submit the **whole** script to `edit_rewritten_script` (it is a full-text
+   write, so untouched parts must come back unchanged — never send a fragment,
+   and never let the model "tidy up" scenes it wasn't asked to touch).
+
+Re-run `rewrite_script` only when the user genuinely wants a different take on
+the whole episode. If a re-run (or an edit) went wrong,
+`get_script` with `include_previous: true` returns the previous draft snapshot.
+
+Point-editable by definition — anything the platform can already name: scene
+length, cross-shot dialogue splits, wrong time-of-day or era, orphan motifs,
+props in the wardrobe segment, missing `[SFX]`/`[BGM]`, group-shot head counts.
+Storyboard-level issues do **not** need a script edit at all: use
+`update_shot` / `replace_shot_dialogue`.
+
 ## Failure playbook
 
 Failures surface as a human-readable `message` (MCP throws `Error(message)`;

package/dist/tools/produce.js

@@ -34,6 +34,10 @@ const WORKFLOW_HINT = '★三档执行策略(别把三档混着问客户):' +
     '把自己写好的剧本直接贴进 edit_rewritten_script 绕过改写会被 400 拒(没有改写稿就没有可改的对象),extract_assets 同样要求基于改写稿。' +
     '改写后的所有修改按 AI 产物的结构化格式做:改稿 edit_rewritten_script(润色/纠正)、人物档案 update_character、分镜 update_shot/replace_shot_dialogue——' +
     '别回头整篇替换剧本或在设定字段里另写一套,两套真相源打架是一致性事故的头号根源。' +
+    '★★**改写成功一次后就别再重跑 rewrite_script**:它是从原稿整篇重来,当前稿的所有修正全丢,' +
+    '且新版不保证保留旧版已改好的地方(三版实测会来回摆)。要修就 edit_rewritten_script 点改' +
+    '(get_script 取全文 → 只改那几场、其余逐字照抄 → 提交整篇),免费秒级、结果确定;' +
+    '误重跑用 get_script(include_previous=1) 回捞上一版。' +
     'generate_portraits_and_sheets(定妆图+设定图·分镜后建只给出场角色出图更省)→assign_voices(分配音色)→frames→videos→generate_tts→compose;' +
     '★别先建角色形象/道具设定图/动作模板再分镜——分镜是纯文本步、不依赖任何图;资产在分镜后建更省更准(动作模板本就必须分镜后)。收费步照现有 quote 报价确认流程。' +
     '广告另需 add_product+generate_product_sheet;MV 走 set_mv_lyrics→generate_mv_story→generate_mv_script。' +
@@ -56,6 +60,12 @@ const WORKFLOW_HINT = '★三档执行策略(别把三档混着问客户):' +
     '★两条锁定纪律:①**画幅比例**在 create_drama 即定、drama 级锁定,之后所有出图/出视频/成片都用它、**别中途改**' +
     '(改了已生成内容画幅会不一致、漂移);不设默认 9:16。②**拆镜每镜 5-7 秒是对 AI 出视频优化的正常时长**,' +
     '别因「镜偏长」误判就重拆——generate_storyboards 会**替换整集所有分镜**、已出图白费,已有分镜后端会拦、需 confirm_replace。' +
+    '★改写保真(默认 auto 智能路由):set_script 的原稿**本身已是剧本形态**(场景头/对白行结构)时,rewrite_script 自动走两步保真——' +
+    '客户台词逐句由机器闸锁定(丢一句即内部拒收重做)、AI 绝不加戏;剧作缺口(钩子/情感锚点)不自动补,写进 get_script 返回的 dramaturgy_suggestions 由客户决定采纳。' +
+    '原稿是小说/大纲则自动走创作型改写(AI 铺钩子造情感点),两种客户各得其所、无需手动切换。' +
+    '要覆盖默认用 update_project_settings 的 rewrite_pipeline(auto/two_pass/single_forced)与 fidelity_enforce(1=保真硬闸)。' +
+    '客户说「AI 把我的剧本改偏了」时的处置:①确认完整原稿已进 set_script;②rewrite_pipeline 设 two_pass 强制保真后重跑 rewrite_script;' +
+    '③客户确认角色外观后用 update_character 的 profile_locked=1 锁定档案,防后续提取覆盖外貌导致定妆图换脸。' +
     '用 get_pipeline_status 查进度(按项目类型返回专属步骤)。';
 const ETHNICITY_CODES = [
     'east_asian', 'southeast_asian', 'south_asian', 'central_asian',
@@ -142,6 +152,8 @@ export function registerProduceTools(server, client) {
         project_type: z.enum(['drama', 'ad', 'mv', 'brand_film']).optional()
             .describe('项目类型(默认 drama)。ad=广告(改写走 ad_script_rewriter);mv=音乐(走歌词→故事→剧本子流程);brand_film=品牌微电影(默认16:9)'),
         rewrite_mode: z.enum(['standard', 'director']).optional().describe('AI改写深度:standard 或 director(导演级)'),
+        rewrite_pipeline: z.enum(['auto', 'two_pass', 'single_forced']).optional().describe('改写流水线:auto(默认,按原稿形态智能路由——剧本形态走两步保真,小说/大纲走创作改写)/two_pass(强制两步保真,客户自带成熟剧本必选)/single_forced(强制单步创作)'),
+        fidelity_enforce: z.number().int().min(0).max(1).optional().describe('1=改写保真硬闸:丢台词/丢人物/丢动作节拍直接拒收重做(客户要求逐句保留时开)'),
         director_style: z.string().optional().describe('导演风格包 key'),
         ...PROJECT_SETTINGS_FIELDS,
     }, async (args) => {
@@ -161,13 +173,33 @@ export function registerProduceTools(server, client) {
     }, async ({ episode_id, script }) => jsonResult(await client.producePut(`/episodes/${episode_id}/script`, { content: script })));
     // ---------- AI 改写 / 审阅 / 改稿 ----------
     server.tool('rewrite_script', 'AI 改写:把原始剧本改写成可拍稿(读 content → 写 script_content)。按项目类型自动选改写 agent' +
+        '★保真自动路由:原稿已是剧本形态时自动走两步保真(台词逐句机器锁定、AI 不加戏,剧作缺口进 dramaturgy_suggestions 由客户决定);原稿是小说/大纲则走创作型改写。' +
         '(广告走 ad 改写;MV 不走标准改写会被拦)。后台异步(分钟级),文本步按 token 后付、不欠费,无需报价。' +
         '完成后用 get_script 审阅、edit_rewritten_script 改稿。' +
         '★典型耗时 2~4 分钟(生产实测 ≈169 秒)。**60 秒内查不到结果是正常的,不是失败**——' +
-        '用 get_run_status 判断还在不在跑,别急着重发。' + WORKFLOW_HINT, { episode_id: z.number().int().positive() }, async ({ episode_id }) => jsonResult(await client.producePost(`/episodes/${episode_id}/rewrite`)));
-    server.tool('get_script', '读某一集的原始内容 + AI 改写后的可拍稿 + 改写状态(供审阅、决定是否 edit_rewritten_script)。免费。', { episode_id: z.number().int().positive() }, async ({ episode_id }) => jsonResult(await client.produceGet(`/episodes/${episode_id}/script`)));
-    server.tool('edit_rewritten_script', '客户改 AI 改写后的可拍稿(写 script_content)。只用于修改 AI 改写产出的稿(人工润色/纠正);' +
-        '本集还没跑过 rewrite_script 时会被 400 拒——这不是绕过改写的通道,别把自己写好的剧本直接贴进来。免费。', {
+        '用 get_run_status 判断还在不在跑,别急着重发。' +
+        '★★**本工具是"从原稿整篇重来",不是"再改一版"**:已有可拍稿时重跑会把当前稿连同已做的所有修正' +
+        '一起覆盖,而且新一版**不保证保留旧版已经改好的地方**(生产三版实测:上一版拆好的长旁白段下一版又' +
+        '合回去、上一版正确的年代服装下一版漂走)。响应里的 overwrites_existing_script=true 就是这个意思。' +
+        '所以**首次改写成功之后,后续所有修改一律用 edit_rewritten_script 点改**——免费、秒级、只动指定的' +
+        '那几场,其余逐字不变,结果确定不抽卡;只有"要一个完全不同的版本"才重跑本工具。' +
+        '误重跑后用 get_script(include_previous=1) 取回上一版。' + WORKFLOW_HINT, { episode_id: z.number().int().positive() }, async ({ episode_id }) => jsonResult(await client.producePost(`/episodes/${episode_id}/rewrite`)));
+    server.tool('get_script', '读某一集的原始内容 + AI 改写后的可拍稿 + 改写状态 + dramaturgy_suggestions(两步保真模式下 AI 识别到但未自动补的剧作缺口——钩子/情感锚点;把它转述给客户决定采纳与否,忽略不影响产线)。免费。' +
+        '★改稿工作流:先用本工具取回 rewritten_script 全文 → 只改客户要改的那几场 → 用 edit_rewritten_script 提交(不要重跑改写)。' +
+        'previous_script_available=true 表示有上一版快照;传 include_previous=1 可取回上一版全文(误重跑/改坏了的回捞路径,默认不带以免响应翻倍)。', {
+        episode_id: z.number().int().positive(),
+        include_previous: z.boolean().optional().describe('取回上一版可拍稿全文(previous_script 快照)。默认 false;改坏了或误重跑时用'),
+    }, async ({ episode_id, include_previous }) => jsonResult(await client.produceGet(`/episodes/${episode_id}/script${include_previous ? '?include_previous=1' : ''}`)));
+    server.tool('edit_rewritten_script', '★改稿的**首选通道**:改 AI 改写后的可拍稿(写 script_content)。免费、秒级、结果确定不抽卡——' +
+        '**改写成功一次之后,所有修改都走这里,不要重跑 rewrite_script**(那是从原稿整篇重来,当前稿的修正全丢、' +
+        '已改好的地方会退回去)。' +
+        '★怎么点改(三步,别跳):① get_script 取回 rewritten_script **全文**;② **只改要改的那几场**——' +
+        '目标场次内改动词句/拆场/删挂错的 motif/补 [SFX]/移器物到 [道具] 行,其余场次连标点都逐字照抄;' +
+        '③ 把改完的**整篇**提交本工具(本工具是全文覆盖,所以未改动部分必须原样带回,不能只发片段、也不要让模型' +
+        '"顺手重写"没让它改的场)。典型可点改的问题:场次太长要拆、台词跨镜断句、时段/年代写错、' +
+        'motif 挂在没有该元素的场、[角色档案] 服装段混进手持器物、缺 [SFX]/[BGM] 标注。' +
+        '只用于修改 AI 改写产出的稿;本集还没跑过 rewrite_script 时会被 400 拒——这不是绕过改写的通道,' +
+        '别把自己写好的剧本直接贴进来。改坏了可用 get_script(include_previous=1) 取回上一版。免费。', {
         episode_id: z.number().int().positive(),
         script: z.string().min(1).describe('改好的可拍剧本(覆盖 AI 改写稿)'),
     }, async ({ episode_id, script }) => jsonResult(await client.producePut(`/episodes/${episode_id}/rewritten-script`, { script_content: script })));
@@ -332,6 +364,8 @@ export function registerProduceTools(server, client) {
         ethnicity: z.enum(ETHNICITY_CODES).optional().describe('角色族裔锁'),
         ethnicity_note: z.string().optional(),
         rewrite_mode: z.enum(['standard', 'director']).optional(),
+        rewrite_pipeline: z.enum(['auto', 'two_pass', 'single_forced']).optional().describe('改写流水线:auto 默认智能路由/two_pass 强制两步保真/single_forced 强制单步创作'),
+        fidelity_enforce: z.number().int().min(0).max(1).optional().describe('1=改写保真硬闸(丢台词即拒)'),
         director_style: z.string().optional(),
         theme_statement: z.string().optional().describe('一句话主题'),
         subtitle_preset: z.string().optional(),
@@ -520,7 +554,8 @@ export function registerProduceTools(server, client) {
     server.tool('generate_mv_script', 'MV 项目:据歌词+故事写首集可拍剧本(写 scriptContent)。之后走标准 extract_assets → 分镜 → …。按用量后付。', { drama_id: z.number().int().positive() }, async ({ drama_id }) => jsonResult(await client.producePost(`/dramas/${drama_id}/mv/script`)));
     server.tool('get_mv', 'MV 项目:读歌词/故事/剧本当前状态。免费。', { drama_id: z.number().int().positive() }, async ({ drama_id }) => jsonResult(await client.produceGet(`/dramas/${drama_id}/mv`)));
     // ========== P1 · 资产精修(提取后改设定再出图)==========
-    server.tool('update_character', '改角色设定(名字/外貌/年龄段/族裔/性别/性格/戏份)。改后需重出定妆图/相关图。免费。', {
+    server.tool('update_character', '改角色设定(名字/外貌/年龄段/族裔/性别/性格/戏份/档案锁定)。改后需重出定妆图/相关图。免费。' +
+        '★profile_locked=1:客户确认外观后锁定档案,后续 AI 提取不再覆盖外貌/性格/描述——外貌锚着定妆图与人脸锁定,被覆盖=全片换脸;确认满意即锁,要再改先传 0 解锁。', {
         character_id: z.number().int().positive(),
         name: z.string().optional(),
         description: z.string().optional(),
@@ -530,6 +565,7 @@ export function registerProduceTools(server, client) {
         gender: z.string().optional(),
         personality: z.string().optional(),
         role: z.string().optional(),
+        profile_locked: z.number().int().min(0).max(1).optional().describe('1=锁定档案(AI 提取不再覆盖外貌/性格/描述,防定妆图换脸);0=解锁'),
     }, async ({ character_id, ...fields }) => jsonResult(await client.producePut(`/characters/${character_id}`, fields)));
     server.tool('delete_character', '删除一个角色(提取误建/合并时用)。免费。', { character_id: z.number().int().positive() }, async ({ character_id }) => jsonResult(await client.produceDelete(`/characters/${character_id}`)));
     server.tool('update_scene', '改场景设定(名字/描述/地点/时段/氛围)。免费。', {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@starreel/mcp",
-  "version": "0.1.39",
+  "version": "0.1.41",
   "mcpName": "ai.starreel/starreel",
   "description": "StarReel MCP server — drive the AI short-drama production pipeline (script → storyboards → frames → video → final cut) from Claude, Cursor, or any MCP client",
   "license": "MIT",

package/server.json

@@ -2,13 +2,13 @@
   "$schema": "https://static.modelcontextprotocol.io/schemas/2025-09-29/server.schema.json",
   "name": "ai.starreel/starreel",
   "description": "Turn a script into a finished, downloadable short-drama episode — via MCP or REST.",
-  "version": "0.1.39",
+  "version": "0.1.41",
   "websiteUrl": "https://starreel.ai",
   "packages": [
     {
       "registryType": "npm",
       "identifier": "@starreel/mcp",
-      "version": "0.1.39",
+      "version": "0.1.41",
       "transport": { "type": "stdio" },
       "environmentVariables": [
         {