@kolbo/mcp 1.70.2 → 1.70.4

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@kolbo/mcp",
-  "version": "1.70.2",
+  "version": "1.70.4",
   "description": "Kolbo AI MCP Server - Generate images, videos, music, speech, and sound effects from Claude Code",
   "main": "src/index.js",
   "bin": {

package/skill/GENERATED.md

@@ -1,6 +1,6 @@
 # AUTO-GENERATED — do not edit
 
-This tree is mirrored from kolbo-code@0fbb0ce, the single source of truth.
+This tree is mirrored from kolbo-code@9e4903a, the single source of truth.
 Canonical source: packages/opencode/skills/kolbo/
 Distribution: .github/workflows/sync-skill-to-plugin.yml
 

package/skill/SKILL.md

@@ -1,5 +1,5 @@
 ---
-version: 0.8.1
+version: 0.8.2
 name: kolbo
 description: |
   Generate, edit, analyze, and direct creative media through Kolbo AI: images,
@@ -155,10 +155,11 @@ A user-named tool — in any language — overrides every other rule. Recognized
 **Preset contract:** if the user asks for a preset, names one, or says to use one of their/Kolbo presets, call `list_presets` with the matching type before generation and pass the selected exact `id` as `preset_id`. Use `image` for `generate_image` and `image_edit` for `generate_image_edit`. Never invent an id or silently continue without the requested preset.
 
 1. **Check credits** ONCE per conversation (Step 0). Skip if already checked.
-2. **Discover models** with `list_models` using a `type` filter — but **skip when the user names a specific model**.
+2. **Discover models** with `list_models` using a `type` filter — but **skip when the user names a specific model** (this turn **or** earlier in the conversation / compaction `## Locked choices`).
 3. **Pick the model**:
-   - User named one → use it. Model identifiers resolve leniently — shorthand like `"z-image"` or `"nano banana 2"` auto-resolves to the exact identifier, so don't over-engineer exact-id lookups (`list_models` is still authoritative for constraints, caps, and pricing).
-   - Auto-select → only from "Auto-selectable" section (models with a `summary`). Cheapest fit. Prefer `[RECOMMENDED]` when cost is similar.
+   - User named one → that name is a **family lock**, not a single catalog row. Use it. Identifiers resolve leniently — `"z-image"` / `"nano banana 2"` / `"grok imagine"` auto-resolve, including to the sibling for the tool you are calling (`grok-imagine-text-to-video` on `generate_video_from_image` becomes `grok-imagine-image-to-video`). `list_models` is still authoritative for constraints, caps, and pricing — not for swapping brands.
+   - **Never cheapest-swap a named family.** After compaction, "animate those images" is still Grok if the user said Grok. Seedance / Kling / Veo are not a "best balance" substitute. If the named family has no variant for this modality, ASK — do not silently switch.
+   - Auto-select → **only when no model was named on this task**. Then pick from "Auto-selectable" (models with a `summary`). Cheapest fit. Prefer `[RECOMMENDED]` when cost is similar.
    - Never auto-select from "Named-only" section.
 4. **Validate inputs** against model caps — see `references/workflows/cost-and-validation.md`.
 5. **How calls work**: each tool blocks until generation is fully complete. Images: seconds. Video: minutes. Multiple tool calls in one response run concurrently. On hosts with live widgets the tool instead returns `submitted` instantly — the card updates on its own; you only need `get_generation_status` when a follow-up step needs the output URLs.

package/skill/VERSION

@@ -1 +1 @@
-0.8.1
+0.8.2

package/skill/references/workflows/production-log.md

@@ -4,7 +4,7 @@ Load this file when starting a multi-step production, or before any continuation
 
 ## Why It Exists
 
-Every URL, id, and brief produced by a Kolbo MCP tool MUST be recorded in `.kolbo/production.md` in the user's workspace. This file — not chat history — is your source of truth for prior artifacts: URLs scattered across `tool_result` blobs are unreliable to re-scan and disappear entirely on context compaction.
+Every URL, id, and brief produced by a Kolbo MCP tool MUST be recorded in `.kolbo/production.md` in the user's workspace. This file — not chat history — is your source of truth for prior artifacts: URLs scattered across `tool_result` blobs are unreliable to re-scan and disappear entirely on context compaction. If the user named a model, write that name into `## 🎯 Now` and keep using that family on every follow-up — compaction is not permission to cheapest-swap.
 
 ## When to READ it
 

package/src/apps/index.js

@@ -321,6 +321,26 @@ function pickForType(candidates, types) {
   return pool.reduce((a, b) => (b.id.length < a.id.length ? b : a)).id;
 }
 
+// Strip modality tokens so a t2v id and its i2v sibling share one family key
+// (grok-imagine-text-to-video ↔ grok-imagine-image-to-video; kling …/text-to-video
+// ↔ …/image-to-video). Version tokens stay (1.5 ≠ 1.0).
+function fam(s) {
+  return normId(s).replace(
+    /texttovideo|imagetovideo|imgtovideo|texttoimage|imagetoimage|imageediting|imageedit|referencetovideo|videotovideo|videoedit|editvideo|firstlastframe|firstlast/g,
+    '',
+  );
+}
+
+function sibling(models, hit, types) {
+  if (!hit || !types.length) return hit;
+  const row = models.find((i) => i.id === hit);
+  if (row && row.types.some((t) => types.includes(t))) return hit;
+  const key = fam(hit);
+  const sibs = models.filter((i) => fam(i.id) === key && i.types.some((t) => types.includes(t)));
+  if (!sibs.length) return hit;
+  return sibs.reduce((a, b) => (b.id.length < a.id.length ? b : a)).id;
+}
+
 /**
  * Lenient model-identifier resolution for LLM-supplied model args.
  * Users say "z-image"; the real identifier is "z-image/turbo" — the backend
@@ -334,6 +354,9 @@ function pickForType(candidates, types) {
  * without it, "Kling 2.6 Pro" from generate_video_from_image resolved to
  * kling-video/v2.6/pro/text-to-video (2026-08-10), so the image-to-video
  * pipeline submitted the TEXT-to-video endpoint and billed against it.
+ * An explicit t2v identifier on an i2v tool remaps to the unique same-family
+ * sibling (grok-imagine-text-to-video → grok-imagine-image-to-video). No
+ * sibling → the id is passed through unchanged (MiniMax H3).
  *
  * Still unresolved: throw with the near misses named. The API answers a bad
  * identifier with a bare INVALID_*_MODEL and no hint, which on 2026-08-09 sent
@@ -362,13 +385,13 @@ async function canonicalModelId(client, input, type) {
   const dashed = key.replace(/\s+/g, '-');
 
   // 1. exact name / identifier hit
-  const exact = pickForType(models.filter((i) => [i.id, i.name].some(
+  const exact = sibling(models, pickForType(models.filter((i) => [i.id, i.name].some(
     (k) => k && (k.toLowerCase() === key || k.toLowerCase() === dashed)
-  )), types);
+  )), types), types);
   if (exact) return exact;
 
   // 2. separator-insensitive exact ("flux-2-flash" → "flux-2/flash")
-  const loose = pickForType(models.filter((i) => normId(i.id) === want || normId(i.name) === want), types);
+  const loose = sibling(models, pickForType(models.filter((i) => normId(i.id) === want || normId(i.name) === want), types), types);
   if (loose) return loose;
 
   // 3. unique prefix ("z-image" → "z-image/turbo") — the modality filter runs

package/src/index.js

@@ -76,8 +76,8 @@ const { registerReviewTools } = require('./tools/review');
 const { registerVoiceTools } = require('./tools/voices');
 const { registerMusicLibraryTools } = require('./tools/music_library');
 const { registerStockLibraryTools } = require('./tools/stock_library');
-const { registerApps, attachToolWidgetMeta } = require('./apps');
-const { attachToolAnnotations } = require('./toolAnnotations');
+const { registerApps, attachToolWidgetMeta } = require('./apps');
+const { attachToolAnnotations } = require('./toolAnnotations');
 
 /**
  * Build a fully-configured Kolbo MCP server (all tool groups registered)
@@ -126,9 +126,9 @@ function createServer(opts = {}) {
       '6. TIMEOUT HANDLING (applies to EVERY generate_* / edit_* / chat_send_message / transcribe_audio tool): each tool blocks and polls internally, then gives up after its own window if the job is not yet terminal. A timeout is NOT a failure — it returns `_timed_out:true` with the `generation_id` (not an error), because the job is almost always STILL RUNNING on the server (or already finished). Call `get_generation_status` with that `generation_id` and `wait=true` to keep checking until state="completed". NEVER conclude the generation failed and re-run the same tool from scratch after a `_timed_out:true` result — that wastes the user\'s credits by paying twice. DIRECTOR / BATCH JOBS follow the same convention through a dedicated tool: generate_creative_director runs its scenes (image OR video) in parallel and only reports state="completed" once EVERY scene is terminal; on `_timed_out:true` call `get_creative_director_status` (not get_generation_status) with the returned generation_id and keep checking. If scenes already carry image_urls/video_urls, they are done; do not regenerate.',
       '7. SESSION CONTINUITY — one task, one session, always: every generation tool returns a `session_id`. For ANY follow-up, refinement, retry, or next step on the SAME task, pass that session_id back — never start fresh. BATCH RULE (critical): when a single user request produces multiple parallel generations (e.g. "animate these 5 images", "generate 3 variants"), do NOT launch them all at once without a session_id. Instead: (1) run the FIRST generation without session_id to create the session, (2) capture the session_id from its response, (3) pass that session_id to ALL remaining generations in the batch. This keeps the entire batch in one session. Exception: only omit session_id and start fresh when the user explicitly starts an unrelated new task.',
       '8. LOCAL FILES / REFERENCE MEDIA — HOW TO HANDLE EVERY CASE. (A) User has a LOCAL file (audio, video, image, document) on their machine. What matters is WHERE THIS SERVER RUNS, not what your client can do — your own filesystem access is irrelevant if the server is somewhere else. On a LOCAL stdio install (server and client share a machine) → call `upload_media` with the absolute path, or pass the path straight to tools like `transcribe_audio` that accept local paths. Over a REMOTE connector the server cannot see that path no matter how capable you are, so a local path will always fail: if you can run shell commands or issue HTTP requests → call `create_upload_ticket` and POST the file to the returned upload_url yourself (fastest, no user interaction); if you cannot → call `media_upload_widget` IMMEDIATELY, the user uploads, and a `media.kolbo.ai` CDN URL comes back for any follow-up call. (B) You already have a public URL (media.kolbo.ai, any CDN, any direct link) → pass it directly; all Kolbo tools accept public URLs. NEVER search for DO Spaces keys, DigitalOcean credentials, or server-side upload credentials. NEVER ask the user to put the file on Google Drive, Dropbox, or Loom. NEVER invent or guess a URL. NEVER base64 anything but a tiny file — it costs context in proportion to file size; use the ticket or the widget instead.',
-      '9. MODEL SELECTION — ROUTE BY THE STRENGTHS SUMMARY, NEVER BY THE BADGE OR THE PRICE TAG: ALWAYS pass a specific `model` on every generation tool — do NOT omit it (omitting falls back to "Smart Select" auto-routing, which hides the choice from the user; use it ONLY if the user explicitly asks you to auto-pick). To choose: call `list_models` with the matching `type` and read each model\'s STRENGTHS SUMMARY — the "— …" clause printed after the credit cost. That summary IS the routing instruction: match it against what the user actually asked for (subject, style, motion, length, quality bar, speed), then pick the CHEAPEST model whose summary covers the task. `[NEW]` and `[RECOMMENDED]` badges, a high credit number, and "flagship"/"most intelligent" wording are NOT selection signals — never pick a model because it is newest, biggest or most expensive. Escalate to a premium/frontier model only when the user explicitly asks for maximum quality, or when no cheaper summary covers the requirement. Models printed under "Named-only" (no summary) are opt-in: use them only when the user names them. TEXT/CHAT: `chat_send_message` bills PER TOKEN, so the listed credit number is not the cost — a frontier text model (Claude Fable 5, GPT-5.6 Sol, Pro-class) costs 5-30x a mid-tier one per reply. Default ordinary chat (writing, brainstorming, Q&A, summarising) to a balanced mid-tier model and reserve the frontier tier for hard reasoning or long-form code the user asked for.',
-      '10. IMAGE EDITING: for ANY prompt-driven / content edit of an existing image — "make it night", changing scene/lighting/colors, adding/removing/replacing objects, restyling — use `generate_image_edit`. Auto-pick only Nano Banana 2 (`nano-banana-2` / `nano-banana-2-image-editing`) or GPT Image 2 (`gpt-image-2` / `gpt-image-2/edit`) for photoreal photo edits, object removal, keep-subject/remove-others, or crowd cleanup. Do NOT auto-pick Flux 2 / flux-2/edit / Flux Klein — those are generate-from-scratch / style, named-only for editing. Do NOT use `edit_image` for content edits — `edit_image` is ONLY for mechanical enhancements (upscale, expand/outpaint, remove-background, skin retouch). Its `magic_edit` operation is deprecated in favor of `generate_image_edit`. EXPANDING AN IMAGE: to widen/extend/uncrop an image or fit it into a wider frame while KEEPING the existing artwork, use `edit_image` with operation="zoom_out" (outpainting — original pixels preserved; size it with `zoom_out_percentage` or the `expand_left/right/top/bottom` pixel args). The "reframe" operation is NOT this: it re-generates the whole picture at a new aspect ratio and the subject comes back re-imagined. Only pick "reframe" when the user wants the shot re-taken, never when they want their image extended.',
-      '11. PRESET CONTRACT: if the user asks for a preset, names a preset, or says to use one of their/Kolbo presets, you MUST call `list_presets` with the matching type before generation, resolve the named or closest matching preset, and pass its exact returned `id` as `preset_id`. Use type="image" for generate_image and type="image_edit" for generate_image_edit. Never silently ignore a preset request, never invent an id, and never claim a preset was applied unless `preset_id` was present in the generation call.'
+      '9. MODEL SELECTION — NAMED MODEL WINS, THEN STRENGTHS SUMMARY: ALWAYS pass a specific `model` on every generation tool — do NOT omit it (omitting falls back to "Smart Select" auto-routing, which hides the choice from the user; use it ONLY if the user explicitly asks you to auto-pick). If the user named a model this turn OR earlier in the conversation (including a compaction "Locked choices" / summary), that name is a FAMILY LOCK: pass it (or its display name) on every follow-up, including when the tool changes (text-to-video → image-to-video). Identifier resolution remaps a t2v id to the family\'s i2v sibling automatically. NEVER substitute a different brand because it is cheaper, faster, or "best balance" (Grok Imagine named → do not fire Seedance). If the named family has no variant for this modality, ASK — do not silently switch. Cheapest-summary routing applies ONLY when no model was named on this task: call `list_models` with the matching `type` and read each model\'s STRENGTHS SUMMARY — the "— …" clause printed after the credit cost — then pick the CHEAPEST model whose summary covers the task. `[NEW]` and `[RECOMMENDED]` badges, a high credit number, and "flagship"/"most intelligent" wording are NOT selection signals — never pick a model because it is newest, biggest or most expensive. Escalate to a premium/frontier model only when the user explicitly asks for maximum quality, or when no cheaper summary covers the requirement. Models printed under "Named-only" (no summary) are opt-in: use them only when the user names them. TEXT/CHAT: `chat_send_message` bills PER TOKEN, so the listed credit number is not the cost — a frontier text model (Claude Fable 5, GPT-5.6 Sol, Pro-class) costs 5-30x a mid-tier one per reply. Default ordinary chat (writing, brainstorming, Q&A, summarising) to a balanced mid-tier model and reserve the frontier tier for hard reasoning or long-form code the user asked for.',
+      '10. IMAGE EDITING: for ANY prompt-driven / content edit of an existing image — "make it night", changing scene/lighting/colors, adding/removing/replacing objects, restyling — use `generate_image_edit`. Auto-pick only Nano Banana 2 (`nano-banana-2` / `nano-banana-2-image-editing`) or GPT Image 2 (`gpt-image-2` / `gpt-image-2/edit`) for photoreal photo edits, object removal, keep-subject/remove-others, or crowd cleanup. Do NOT auto-pick Flux 2 / flux-2/edit / Flux Klein — those are generate-from-scratch / style, named-only for editing. Do NOT use `edit_image` for content edits — `edit_image` is ONLY for mechanical enhancements (upscale, expand/outpaint, remove-background, skin retouch). Its `magic_edit` operation is deprecated in favor of `generate_image_edit`. EXPANDING AN IMAGE: to widen/extend/uncrop an image or fit it into a wider frame while KEEPING the existing artwork, use `edit_image` with operation="zoom_out" (outpainting — original pixels preserved; size it with `zoom_out_percentage` or the `expand_left/right/top/bottom` pixel args). The "reframe" operation is NOT this: it re-generates the whole picture at a new aspect ratio and the subject comes back re-imagined. Only pick "reframe" when the user wants the shot re-taken, never when they want their image extended.',
+      '11. PRESET CONTRACT: if the user asks for a preset, names a preset, or says to use one of their/Kolbo presets, you MUST call `list_presets` with the matching type before generation, resolve the named or closest matching preset, and pass its exact returned `id` as `preset_id`. Use type="image" for generate_image and type="image_edit" for generate_image_edit. Never silently ignore a preset request, never invent an id, and never claim a preset was applied unless `preset_id` was present in the generation call.'
     ].join('\n')
   });
   const progress = require('./progress');
@@ -173,12 +173,12 @@ function createServer(opts = {}) {
   // transport. Without it, a remote-connector model reads "absolute local path",
   // sees no filesystem, and tells the user Kolbo cannot accept their file —
   // the single most-reported failure, despite the upload tools existing.
-  attachFileInputHints(server, toolOptions);
-  // OpenAI public-app review requires every exposed tool to declare the three
-  // safety hints explicitly. The exact contract also fails closed when a tool
-  // is added or removed without a classification.
-  attachToolAnnotations(server);
-  // Declaration-level `_meta['ui/resourceUri']` on every widget-carrying tool —
+  attachFileInputHints(server, toolOptions);
+  // OpenAI public-app review requires every exposed tool to declare the three
+  // safety hints explicitly. The exact contract also fails closed when a tool
+  // is added or removed without a classification.
+  attachToolAnnotations(server);
+  // Declaration-level `_meta['ui/resourceUri']` on every widget-carrying tool —
   // claude.ai prepares the widget iframe from tools/list, not from the result.
   attachToolWidgetMeta(server);
 

package/src/tools/generate.js

@@ -438,7 +438,7 @@ function registerGenerateTools(server, client, options = {}) {
     {
       prompt: z.string().optional().describe('Text description of the video to generate. Required unless `prompts` is provided.'),
       prompts: promptsField('videos'),
-      model: z.string().optional().describe('Model identifier — pick a SPECIFIC model, do NOT omit (omitting = Smart Select auto-pick, which we avoid). Strong current defaults: "seedance-2" (versatile) or "veo3" (Veo 3.1, cinematic + native audio); the Kling family (call list_models for exact ids like kling-video/v3/pro/text-to-video) is strongest for motion. Call list_models type="text_to_video" to see all options + check supported_durations / supported_aspect_ratios, and choose per the user\'s intent.'),
+      model: z.string().optional().describe('Model identifier — pick a SPECIFIC model, do NOT omit (omitting = Smart Select auto-pick, which we avoid). If the user already named a model/family (this turn or earlier), pass that name — do not substitute a cheaper default. Only when no model was named: "seedance-2" (versatile) or "veo3" (cinematic + native audio) are reasonable auto-picks; Kling is strongest for motion (list_models type="text_to_video" for exact ids). Call list_models for supported_durations / supported_aspect_ratios.'),
       aspect_ratio: z.string().optional().describe('Aspect ratio (e.g., "16:9", "9:16", "1:1"). Must be in the chosen model\'s `supported_aspect_ratios` from list_models. Default: "16:9"'),
       duration: z.number().optional().describe('Duration in seconds. Must be a value in `supported_durations` from list_models, OR within `min_output_duration`-`max_output_duration` (whichever the model exposes). Default: 5'),
       enhance_prompt: z.boolean().optional().describe('Enhance the prompt. Default: false — only pass true if the user explicitly asks to enhance/improve the prompt.'),
@@ -520,7 +520,7 @@ function registerGenerateTools(server, client, options = {}) {
       })).max(MAX_BATCH_PROMPTS).optional().describe(
         `BATCH MODE — several DIFFERENT stills (2–${MAX_BATCH_PROMPTS}) animated concurrently in ONE call and rendered together in ONE combined widget. Unlike the \`prompts\` array on generate_image / generate_video, each entry pairs its OWN \`image_url\` with its OWN motion \`prompt\` — the image is what varies, and that is the point. **Hard cap: ${MAX_BATCH_PROMPTS} items per call — more than that is REJECTED with an error (never silently truncated), so split a longer sequence across several calls of at most ${MAX_BATCH_PROMPTS}.** Whenever the user wants several stills animated (a shot sequence, a storyboard, an animatic), ALWAYS pass them all here instead of making several separate calls — separate calls clutter the chat with stacked widgets. All items share the same model / duration / resolution / aspect_ratio / sound_enabled / project_id / session_id. When set, \`image_url\` and \`prompt\` are ignored.`
       ),
-      model: z.string().optional().describe('Model identifier — pick a SPECIFIC model, do NOT omit (omitting = Smart Select auto-pick, which we avoid). Strong current defaults: "seedance-2" (versatile) or "veo3" (Veo 3.1, cinematic + native audio); the Kling family (call list_models for exact ids like kling-video/v3/pro/image-to-video) is strongest for motion. Call list_models type="img_to_video" to see all options and choose per the user\'s intent.'),
+      model: z.string().optional().describe('Model identifier — pick a SPECIFIC model, do NOT omit (omitting = Smart Select auto-pick, which we avoid). If the user already named a model/family (this turn or earlier), pass that name — a text-to-video id remaps to the family\'s image-to-video sibling. Do not substitute a cheaper default (named Grok Imagine → not Seedance). Only when no model was named: "seedance-2" or "veo3" are reasonable auto-picks; Kling is strongest for motion (list_models type="img_to_video" for exact ids).'),
       aspect_ratio: z.string().optional().describe('Output aspect ratio (e.g., "16:9", "9:16", "1:1"). Must be in the chosen model\'s `supported_aspect_ratios` from list_models. Default: "16:9"'),
       duration: z.number().optional().describe('Duration in seconds. Must be in `supported_durations` from list_models, OR within `min_output_duration`-`max_output_duration`. Default: 5'),
       enhance_prompt: z.boolean().optional().describe('Enhance the motion prompt. Default: false — only pass true if the user explicitly asks to enhance/improve the prompt.'),

package/src/tools/models.js

@@ -363,7 +363,7 @@ function registerModelTools(server, client, options = {}) {
       }
 
       if (withSummary.length > 0) {
-        sections.push(`Auto-selectable models (${withSummary.length}) — CHOOSE BY THE SUMMARY after each "—": match it to what the user asked for, then take the CHEAPEST model that fits. Credit cost, [NEW] and [RECOMMENDED] are not reasons to pick a model:\n${withSummary.map(formatModel).join('\n')}`);
+        sections.push(`Auto-selectable models (${withSummary.length}) — If the user already named a model/family (this turn or earlier), use that family — do not cheapest-swap. Otherwise CHOOSE BY THE SUMMARY after each "—": match it to what the user asked for, then take the CHEAPEST model that fits. Credit cost, [NEW] and [RECOMMENDED] are not reasons to pick a model:\n${withSummary.map(formatModel).join('\n')}`);
       }
       if (withoutSummary.length > 0) {
         sections.push(`Named-only models (${withoutSummary.length}) — only use if the user explicitly requests by name:\n${withoutSummary.map(formatModel).join('\n')}`);

package/src/tools/visual_dna.js

@@ -197,7 +197,7 @@ function registerVisualDnaTools(server, client, options = {}) {
     'Generate a reference sheet for a Visual DNA from 1+ reference image URLs — the same step the in-app Visual DNA wizard offers, for EVERY DNA type via `sheet_type`: character = multi-angle turnaround, product = angles + branding/material/construction close-ups, environment = location angles + one signature detail, style = a style board (the same look applied to six varied subjects). The sheet is the single strongest consistency booster for a DNA, and it always preserves the reference\'s original art style (2D stays 2D, photo stays photo). CHARGES CREDITS, so when the user is about to create a DNA, OFFER this first ("want me to generate a reference sheet for stronger consistency? it costs a few credits") and only run it on a yes. Returns `character_sheet_url` — pass it as `character_sheet_url` to `create_visual_dna` with the matching `dna_type`.',
     {
       image_urls: z.array(z.string()).min(1).describe('Reference image URLs of the subject (for characters: front/side/varied angles work best). Use generated-image URLs or upload_media output.'),
-      sheet_type: z.enum(['character', 'character_headless', 'product', 'environment', 'style']).optional().describe('Sheet layout. character = front/back/face turnaround. character_headless = wardrobe/body refs with a headless front panel (use when clothing must change without fighting the face sheet). product / environment / style = matching DNA types. Defaults to character.')
+      sheet_type: z.enum(['character', 'character_headless', 'character_bible', 'product', 'environment', 'style']).optional().describe('Sheet layout. character = front/back/face turnaround. character_headless = wardrobe/body refs with a headless front panel (use when clothing must change without fighting the face sheet). character_bible = denser production model-sheet (turnaround + faces + wardrobe + color swatches). product / environment / style = matching DNA types. Defaults to character.')
     },
     async ({ image_urls, sheet_type }) => {
       const result = await client.post('/v1/visual-dna/character-sheet', { image_urls, ...(sheet_type ? { sheet_type } : {}) });