@kolbo/mcp 1.86.4 → 1.87.1

package/skill/references/models/seedance.md

@@ -23,48 +23,20 @@ Load this file when the user wants a **Seedance 2 / Seedance 2.0** (ByteDance) v
   Then Locked Intro, then `SHOT N — 0:00–0:02 — Size / camera` beats whose ranges **sum exactly to Xs**. Last line repeats `Total: Xs / N shots / AR`.
   - Example (15s / 6 shots): `6 connected cinematic shots, 15 seconds total, 16:9, Multishot ON` + `Total: 15s / 6 shots / 16:9`
   - UGC / phone vertical: `N connected phone shots, Xs total, 9:16, Multishot ON` (never the word "cinematic").
-  - A MULTI-shot prompt with only shot body and no Total / Multishot header is a **failed turn** — rewrite before calling `generate_*`. A single-shot prompt uses the single-shot header instead and carries no `Multishot ON`; see "Shot count comes from the USER".
+  - A prompt with only shot body and no Total / Multishot header is a **failed turn** — rewrite before calling `generate_*`.
 - **MCP `duration` must match the Total line.** Pass `duration: X` (whole seconds) on `generate_video` / `generate_elements` / `generate_video_from_image` equal to the `Xs` in `Total: Xs / …`. Mismatch = wrong-length clip.
 - **Then the Locked Intro** — `[GLOBAL LOOK]` / `[CAST]` / `[LOCATION]` (+ LOCATION MAP / CONTINUITY / PHYSICS for multi-shot) — before any shot. A one-liner `same character throughout` is not a character lock.
 - **Order inside each shot**: Subject → Action → Camera → Constraints → (Audio/SFX if relevant). Do NOT restack GLOBAL LOOK style inside the shot.
 - **Prompt length**: simple single-idea pieces ~120–280 words. Locked-intro cinematic typically 400–900 words. Shorter than ~120 words = random output. The 10,000-char cap below always wins.
-- **Shot count is user-directed.** If the user asks for N shots, deliver exactly N in one prompt unless they ask to split — and if they ask for ONE shot, deliver one shot with no `Multishot ON`.
+- **Shot count is user-directed.** If the user asks for N shots, deliver exactly N in one prompt unless they ask to split.
 - **Always describe at least one camera movement per shot.**
 - **Tell Seedance what the camera is NOT doing** (e.g. `no cuts, no zoom, natural head movement`) — this is what locks POV.
 - **Final prompt is always English**, wrapped in a copy-ready code block. Detect intent in any language and reply in the user's language, but the prompt itself is English.
 - **HARD CAP: 10,000 characters TOTAL for the ENTIRE prompt** — measured as one single string including all shots, boilerplate, SFX lines, and the Total lines. It is per PROMPT, not per shot. **Never** split into multiple prompts, code blocks, or "part 1 / part 2" to evade the cap. Count the final prompt before output; if over, trim (cut adjectives, collapse boilerplate, shorten SFX lists, merge or drop shots) and re-count until it fits.
 
+## Locked Intro (DEFAULT for any multi-shot cinematic — including Elements)
 
-## Shot count comes from the USER — decide this FIRST
-
-A SHOT is one uninterrupted camera take. A CUT is what separates two shots. Count
-what the user asked for before choosing the output shape.
-
-**ONE shot requested → write ONE shot.**
-- No `SHOT N` labels, no per-shot beats, and **no `Multishot ON`**. That flag declares
-  "this clip contains hard cuts" — on a single take it is false, and it additionally
-  forces prompt enhancement on the wire, rewriting the prompt the user just approved.
-- Header: `Single continuous shot, Xs total, AR`. Closing line: `Total: Xs / 1 shot / AR`.
-- Describe the take as ONE unbroken movement; internal beats are timestamps inside it
-  (`0:00–0:05 — the camera pushes in past the doorway…`), never numbered shots.
-- **Keep the Locked Intro blocks.** GLOBAL LOOK / CAST / LOCATION / PHYSICS are the
-  consistency stack, not the multi-shot part — a long continuous move needs them most.
-
-These all mean ONE shot, however long it runs and however far the camera travels:
-"one shot", "single shot", "one continuous take", "a oner", "no cuts", "unbroken",
-"one continuous camera movement". **Writing "N connected shots … no visible cut" is a
-contradiction** — no cut means one shot. That exact output is what this rule exists to
-stop; never emit it.
-
-**N shots requested → deliver exactly N.** Never round up to a nicer-sounding number,
-never add shots the user did not ask for, and never invent a maximum.
-
-**No count given → pick the SIMPLEST structure the idea needs.** One continuous take is
-very often right. A montage is a deliberate choice, never a default.
-
-## Locked Intro (DEFAULT for any cinematic piece — single-shot and Elements included)
-
-After the Total lines, every prompt with recurring people, a recurring place, or more than one shot opens with the locked blocks. A single continuous take keeps ALL of them — only the `SHOT N` beats and `Multishot ON` are multi-shot-only. Skip entirely for: a bare POV/orb with no cast, 3×3 grid-panel mode, or video-edit tasks.
+After the Total lines, every multi-shot prompt — and any piece with recurring people or a recurring place — opens with the locked blocks. Skip only for: true single-shot POV/orb, 3×3 grid-panel mode, or video-edit tasks.
 
 ```
 N connected cinematic shots, Xs total, AR, Multishot ON

package/skill/references/workflows/cost-and-validation.md

@@ -17,7 +17,7 @@ Creative generations bill against the user's Kolbo credit balance. **Billing uni
 | **Music** | per generation (flat) | 15–60 cr | Suno v5 = 15 cr; ElevenLabs Music = 60 cr |
 | **Speech (TTS)** | per 100 characters | 2–5 cr/100 chars | ElevenLabs (5) × 500 chars = 25 cr |
 | **Sound effects** | per generation (flat) | 4–7 cr | |
-| **3D model** | per model (flat, × toggle multipliers on Meshy V7) | 5–300 cr | Trellis = 5 cr; Trellis 2 = 60 cr; Meshy v5/v6 = 150 cr; Meshy V7 = 186 cr base (no-texture ~124; +rigging ~217; +rigging+animation ~236); Marble 1.1 = 300 cr |
+| **3D model** | per model (flat) | 5–300 cr | Trellis = 5 cr; Meshy v6 = 150 cr; Marble 1.1 = 300 cr |
 | **Transcription (stt)** | per minute of audio | `model.credit × duration_minutes` | |
 
 ## Calculation Formulas

package/skill/references/workflows/visual-dna.md

@@ -130,18 +130,18 @@ visual_dna_ids: ["vdna_abc",  // dana
 
 The match is **literal and case-insensitive**, so:
 - The `@name` must equal the stored `name` field (e.g. if `name: "esther_model"` → write `@esther_model`, not `@Esther`, not `@אסתר`, not `@the model`).
-- Any-language characters are supported — if the DNA was created with `name: "אסתר"` you write `@אסתר`. Use the EXACT stored string.
+- Any-language characters are supported — if the DNA was created with `name: "אסתר"` you write `@אסתר`. Use the EXACT stored string. Asset tags are identifiers and are exempt from English-only prompt/dialogue rules: never translate, transliterate, lowercase, strip diacritics, replace spaces, or slugify an existing name. Resolve the selected id with `get_visual_dna` if the current name is unknown; never guess from the user's language or a production-log alias.
 - Mentions terminate at punctuation (`.,!?`), double-spaces, another `@`, or end of string. So `@maya, wearing...` matches `maya`.
 
 This composes with `@image1` / `@image2` positional tags for plain reference/source images — see "Reference Tagging" below.
 
 ### ⚠️ Naming rule for `create_visual_dna` — NO SPACES (MANDATORY)
 
-The `name` you set MUST be a **single token, lowercase, no spaces, ASCII-safe** — `esther_model`, `dana`, `tokyo_neon`, `brand_red`. Never `Sarah Johnson`, never `the red dress`.
+For a NEW DNA, prefer a short single token in the user's chosen language, such as `אסתר`, `ليلى`, `小雨`, or `esther_model`. ASCII and lowercase are not required. This naming recommendation never permits rewriting an EXISTING stored name or its prompt tag.
 
 Reason: the prompt parser stops the `@<token>` match at the first space (and at `.,!?` punctuation). So `@Sarah Johnson` matches *only* `Sarah` — if no DNA named `Sarah` exists, the mention is silently dropped and the DNA never binds. A single-token name is the only way to guarantee inline `@name` works in any sentence, in any language, without forcing the user to write awkward punctuation around it.
 
-Use underscores for multi-word concepts (`old_town`, not `Old Town`). When the user proposes a name with spaces, accept the intent but collapse it into a single token before storing (`"Sarah Johnson"` → `sarah_johnson`) and tell them once how you'll refer to it. Source of truth: [kolbo-docs / Visual DNA & @ References](https://docs.kolbo.ai/kolbo-code/visual-dna).
+For new multi-word names, suggest underscores in the same language. Do not silently translate or rename a user-specified name. For existing names, preserve the full stored string and selected id, including spaces; if binding fails, report the limitation instead of inventing an alias or renaming the user's DNA.
 
 ## Reference Tagging — `@image1` / `@video1` / `@Audio1`
 

package/src/tools/chat.js

@@ -19,11 +19,12 @@ function registerChatTools(server, client) {
       system_prompt: z.string().optional().describe('System prompt for the conversation. Only applied when creating a new session.'),
       web_search: z.boolean().optional().describe('Enable web search for this message. Default: false'),
       deep_think: z.boolean().optional().describe('Enable deep think (extended reasoning). Default: false'),
+      thinking_level: z.string().optional().describe('Thinking effort ID from list_models type="text" thinkingLevels. The server uses the model catalog default when omitted or invalid. Separate from legacy deep_think; safeguards take precedence.'),
       enhance_prompt: z.boolean().optional().describe('Enhance the prompt. Default: false — only pass true if the user explicitly asks to enhance/improve the prompt.'),
       media_urls: z.array(z.string()).optional().describe('Public URLs of images, videos, or audio files to analyze. The model auto-routes to a vision-capable model when media is present. For a local file, get a URL first via the LOCAL FILE route in this tool\'s description.'),
       project_id: projectIdField
     },
-    async ({ message, model, session_id, system_prompt, web_search, deep_think, enhance_prompt = false, media_urls, project_id }) => {
+    async ({ message, model, session_id, system_prompt, web_search, deep_think, thinking_level, enhance_prompt = false, media_urls, project_id }) => {
       // Every generate_* tool resolves its model this way; chat was the one
       // `model` arg that went straight to the API, which has no fuzzy matching.
       // So the display names list_models hands back ("Claude Fable 5") came
@@ -38,6 +39,7 @@ function registerChatTools(server, client) {
         system_prompt,
         web_search,
         deep_think,
+        ...(thinking_level !== undefined ? { thinking_level } : {}),
         enhance_prompt,
         media_urls,
         project_id

package/src/tools/models.js

@@ -192,6 +192,9 @@ function registerModelTools(server, client, options = {}) {
       // model says no, and absence means the API doesn't expose the field.
       const formatSpecs = m => {
         const parts = [];
+        if (m.haveThinking && Array.isArray(m.thinkingLevels) && m.thinkingLevels.length) {
+          parts.push(`thinking_level: ${m.thinkingLevels.map(level => level.id).join('/')} (default ${m.thinkingDefault})`);
+        }
         const types = Array.isArray(m.types) ? m.types : [];
         const isVideoType = types.some(t =>
           ['text_to_video', 'img_to_video', 'video_to_video', 'elements',

package/skill/references/models/3d.md

@@ -1,51 +0,0 @@
-# 3D Generation (`generate_3d`)
-
-Three model families, discoverable via `list_models` with types `3d_text_to_model`,
-`3d_image_to_model`, `3d_multi_image_to_model` (plus `3d_world` for world/splat generation).
-Settings are **family-scoped** — params for a different family than the selected model are
-silently ignored, so match the params to the model you pass.
-
-## Families & when to pick each
-
-| Family | Identifiers | Pick when | Base credits |
-|---|---|---|---|
-| **Meshy V7** | `fal-ai/meshy/v7/image-to-3d`, `fal-ai/meshy/v7/multi-image-to-3d` | Game-ready assets, characters (rigging!), highest fidelity, PBR | 186 |
-| Meshy v5/v6 | `meshy/v5/multi-image-to-3d`, `meshy/v6-preview/*` (incl. the only **text**-to-3D) | Text mode, or legacy compatibility | 150 |
-| Trellis v1 | `trellis-image-to-3d`, `trellis-multi-image-to-3d` | Fast + cheap drafts | 5 |
-| Trellis 2 | `trellis-2-image-to-3d` | Better quality than v1, 4K textures, polygon control | 60 |
-
-Multi-image mode: 2–4 images of the SAME object from different angles (Trellis: up to 6).
-More angles = better reconstruction. Output formats: GLB (always, in-app preview), FBX/OBJ/USDZ
-and textures via the full package.
-
-## Meshy V7 settings (the full-control family)
-
-- `topology` `"triangle"|"quad"`, `target_polycount` 100–300000 (default 30000), `symmetry_mode` `"off"|"auto"|"on"`.
-- `should_remesh` (default true) — false keeps the raw reconstructed mesh, ignores topology/polycount.
-- `should_texture` (default true) — **false is cheaper** (~0.67× base). Disables `enable_pbr`,
-  `texture_prompt`, `texture_image_url`.
-- `texture_prompt` (≤600 chars) and/or `texture_image_url` — guide texturing.
-- `enable_tpose` — output an A/T-pose character.
-- **Rigging**: `enable_rigging` auto-rigs a humanoid (best with clear limbs) + basic walk/run
-  animations; `rigging_height_meters` (default 1.7). **~1.17× credit surcharge.**
-- **Animation**: `enable_animation` (requires `enable_rigging`) applies one preset from Meshy's
-  ~697-action library via `animation_action_id` 0–696 (default 92 "Idle"; 0=Idle, 1=Walk, 14=Run,
-  4=Attack, 22=Dance, 290=Wave — full catalog at docs.meshy.ai/en/api/animation-library).
-  **Additional ~1.09× surcharge on top of rigging.** Returns `animation_glb`/`animation_fbx` plus
-  the rigged character files.
-
-Credit math is toggle-multiplied off the base price (server-computed; the exact quote comes back
-in the generation response) — e.g. base 186 → +rigging 217 → +animation 236; no-texture 124.
-
-## Trellis settings
-
-- v1: `texture_size` `"512"|"1024"|"2048"`, `ss_guidance_strength`/`ss_sampling_steps`,
-  `slat_guidance_strength`/`slat_sampling_steps` (more steps = higher quality, slower),
-  `mesh_simplify`, `multiimage_algo` `"stochastic"|"multidiffusion"` (multi mode), `seed`.
-- Trellis 2: `resolution` `"512"|"1024"|"1536"`, `t2_texture_size` `"1024"|"2048"|"4096"`,
-  `decimation_target` (polygons), `remesh` (default true), `tex_sampling_steps`, `seed`.
-
-## Text mode (Meshy v6-preview only)
-
-`prompt` required; `art_style` `"realistic"|"sculpture"` (sculpture disables PBR),
-`enable_prompt_expansion` for AI prompt enrichment.