@kolbo/mcp 1.88.2 → 1.88.4

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@kolbo/mcp",
-  "version": "1.88.2",
+  "version": "1.88.4",
   "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@0ac4fba, the single source of truth.
+This tree is mirrored from kolbo-code@1dc58c1, 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: 0.9.15
+version: 0.9.16
 name: kolbo
 description: |
   Generate, edit, analyze, and direct creative media through Kolbo AI: images,
@@ -72,7 +72,8 @@ For multi-scene / batch work this pairs with `generate_creative_director` (see b
 | Build, inspect, animate, light, render, or edit a **Blender scene through Kolbo Blender MCP** | `references/workflows/blender-mcp.md` |
 | Generate a **Seedance 2.5** video | `skill` `elements-prompting` + `references/models/seedance25.md` + Locked Intro in `references/models/seedance.md`. For narrative/continuity also load `references/workflows/filmmaking.md` — but compile the prompt as Locked Intro, NOT the SCENE CONTEXT / OPTICS / ACTION pack |
 | Generate a **Seedance 2 / WAN / MiniMax H3 / Gemini / Elements** video (`generate_elements` or Visual DNA) | `skill` `elements-prompting` + `references/models/seedance.md` — same Locked Intro. Elements is NOT a different prompt language |
-| Generate a **GPT Image 2** image | `references/models/gpt-image.md` |
+| Generate a **GPT Image 2 / 2.5** image | `references/models/gpt-image.md` |
+| **No background / transparent PNG / cutout** on any image | `references/models/gpt-image.md` |
 | Generate a **Nano Banana / Gemini** image | `references/models/nano-banana.md` |
 | Generate a **Veo 3 / 3.1** video | `references/models/veo.md` |
 | Build a **multi-scene set** (Creative Director, storyboard, campaign batch, 4+ angles) | `references/models/creative-director.md` |
@@ -113,8 +114,8 @@ Font tools (when exposed by the installed MCP): `list_fonts`, `get_font`, `uploa
 ### Generation
 | Tool | Description |
 |------|-------------|
-| `generate_image` | Single image from a text prompt. Supports Visual DNA, moodboards, image presets (custom instructions live here), reference images, web-search grounding. Named sheets/styles: `list_presets({ type: "image", search: "headless" })` then `preset_id`. |
-| `generate_image_edit` | Edit/transform an existing image. Pass `source_images` + edit prompt. Image-editing presets are supported through `preset_id` from `list_presets({ type: "image_edit" })`. |
+| `generate_image` | Single image from a text prompt. Supports Visual DNA, moodboards, image presets (custom instructions live here), reference images, web-search grounding. Named sheets/styles: `list_presets({ type: "image", search: "headless" })` then `preset_id`.  NO BACKGROUND: require `supports_transparent_background: true` from `list_models`, then pass `background: "transparent"` with PNG/WebP — Kolbo appends the phrase `no background` once; prompt wording alone does not replace the setting. |
+| `generate_image_edit` | Edit/transform an existing image. Pass `source_images` + edit prompt. Image-editing presets are supported through `preset_id` from `list_presets({ type: "image_edit" })`.  Native no-background output uses the same capability gate and `background: "transparent"` contract as `generate_image`. |
 | `generate_creative_director` | **2–8 related images or videos as one coherent set.** Use INSTEAD of multiple `generate_image` calls for any related multi-output. |
 | `generate_video` | Text-to-video. Accepts `visual_dna_ids` and `sound_enabled`; `generate_elements` is still the primary reference-driven route for a DNA-anchored film. |
 | `generate_video_from_image` | Animate a still. Prompt describes motion, not subject. |
@@ -274,6 +275,7 @@ A user-named tool — in any language — overrides every other rule. Recognized
 | "Modify THIS one image" — change bg, remove object, recolor | `generate_image_edit` | ❌ Not for multi-output |
 | "4 angles / poses / views of this character" / "variations of this character" | `generate_creative_director` with `visual_dna_ids` | ❌ Don't loop `generate_image_edit` |
 | "4 variations of THIS exact image" (same prompt, different seeds) | `generate_image` with `num_images=4` | ❌ Not `generate_image_edit` |
+| "No background" / transparent PNG / cutout, **בלי רקע** / **רקע שקוף** — new image or existing one | `generate_image` / `generate_image_edit` with `background: "transparent"` on a GPT Image 2 / 2.5 model (gate on `supports_transparent_background: true`) | ❌ Prompt wording alone — it returns an opaque image. ❌ `edit_image` `removebg` unless they want a purely mechanical cutout of an existing photo |
 
 ## Core Workflow
 

package/skill/VERSION

@@ -1 +1 @@
-0.9.15
+0.9.16

package/skill/references/models/gpt-image.md

@@ -2,16 +2,17 @@
      kolbo-api/src/config/systemPrompt.js (lines ~858–965).
      When that function changes, update this file in the same session. -->
 
-# GPT Image 2 — Prompt Rules
+# GPT Image 2 / 2.5 — Prompt Rules
 
-Load this file when the user wants a **GPT Image 2 / gpt-image-2** image (OpenAI). For other image models see `models/nano-banana.md`, `models/creative-director.md`, or `models/prompt-copilot.md`.
+Load this file when the user wants a **GPT Image 2 or GPT Image 2.5** image (OpenAI). The live family is `gpt-image-2`, `gpt-image-2.5-sunburst` and `gpt-image-2.5-flare` — all three are the ONLY Kolbo image models that can output a native transparent background. (`gpt-image/1.5-text-to-image` is the older row and cannot.) For other image models see `models/nano-banana.md`, `models/creative-director.md`, or `models/prompt-copilot.md`.
 
-**Kolbo MCP routing:** call `generate_image` (text-to-image) or `generate_image_edit` (edits with `source_images`). Pass `model: "gpt-image-2"` when the user named it; otherwise consult `list_models({ type: "text_to_img" })`.
+**Kolbo MCP routing:** call `generate_image` (text-to-image) or `generate_image_edit` (edits with `source_images`). Pass the exact identifier the user named (`gpt-image-2`, `gpt-image-2.5-sunburst`, `gpt-image-2.5-flare`); otherwise consult `list_models({ type: "text_to_img" })`. On `generate_image_edit` the same families appear as their `/edit` rows under `list_models({ type: "image_editing" })`.
 
 ## CRITICAL Kolbo Platform Rules
 
 - **Resolution and aspect ratio are MCP-tool params** (`aspect_ratio`, `resolution`) — NEVER include `size=`, `1024x1536`, aspect-ratio tags, or any resolution syntax inside the `prompt` field.
 - Pass aspect / resolution as separate tool parameters. Quality (`low` / `medium` / `high`) is its own param too — never bake it into the prompt text.
+- **No background uses both a tool parameter and a prompt cue.** If the user asks for no background / a transparent cutout, call `list_models` for the chosen image type and require `supports_transparent_background: true`. Then pass `background: "transparent"` and `output_format: "png"` (or `"webp"`). Kolbo appends the exact phrase `no background` once to the effective prompt automatically. The phrase alone is not sufficient; if the model does not advertise the capability, choose a supported model with the user.
 - Do not write Python, `client.images.generate`, OpenAI SDK code, or `size=` keyword arguments. The user is generating through Kolbo's MCP tools.
 
 ## Universal Prompting Rules (apply to EVERY prompt)

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

@@ -119,3 +119,7 @@ Normal cost formula: `final_cost = credit × output_seconds × resolution_multip
 ## Log Approved Resolution / Duration / Sound Choices
 
 After the user approves the actual result, log its `credits_used`, resolution, duration, and sound state. Pending and rejected outputs stay out; use the format in `production-log.md`.
+
+## Reference evidence in generation status
+
+For a reference audit, inspect the persisted status result's `visual_dna` and reference image fields. The widget's `visual_dnas` is display metadata, not the original submitted request. Missing widget metadata means unknown, not that no reference was used. Server-side @mention resolution can attach references beyond the caller’s explicit fields. Keep submitted inputs, persisted references, and observed visual fidelity distinct; attaching a DNA does not prove identity fidelity.

package/skill/references/workflows/media-library.md

@@ -4,6 +4,10 @@ Load this file when the user wants to browse, list, organize, delete, restore, m
 
 The library covers both **uploaded files** and **AI-generated outputs the user has saved**. Tools fall into five groups: ingest, browse, lifecycle (delete/restore/move), folders, and favorites.
 
+`list_media` returns the complete requested page in both text and widget results. Follow `pagination.has_next` using the returned page size; use smaller pages (for example 20) when reviewing detailed assets to keep context manageable. `project_id` restricts SDK results to that project's recorded membership, not the project-looking segment in a CDN URL (files can be moved). A folder filter takes precedence over the project filter.
+
+For production research, `get_visual_dna` provides the full stored description and references; list descriptions are compact previews. Internal extraction system prompts are intentionally private. `get_project_profile` reads the stored synthesized brief; it is not a substitute for the original project documents or session history.
+
 ## ⚠️ Already-hosted URLs — never re-upload
 
 `generate_*` / `list_media` / `get_media` / a prior `upload_media` already return a Kolbo CDN URL (`media.kolbo.ai`, `*.kolbo.ai`, Spaces). Pass that exact URL into the next generation tool. Calling `upload_media` on it duplicates the file.

package/src/apps/widgets/generation.js

@@ -1330,6 +1330,64 @@ function openPromptRow(placeholder, onSend) {
    The host mounts this iframe as soon as the tool is CALLED; the result can
    take many seconds (model resolution, file upload, submit). Show a live
    shell immediately instead of a blank card. */
+// Which tool args carry a reference the browser can actually load. The kind
+// here is only a FALLBACK: refKind() still lets the file extension win, exactly
+// like the result path, so a .mp4 handed to the files arg renders as video.
+var PRE_IMAGE_KEYS = ['source_images', 'reference_images', 'image_url', 'mask_image_url',
+  'additional_images', 'first_frame', 'last_frame', 'seed_reference_image_url',
+  'elements', 'files', 'keyframes', 'source'];
+var PRE_VIDEO_KEYS = ['source_video', 'reference_videos'];
+var PRE_AUDIO_KEYS = ['audio', 'audio_url', 'reference_audio_urls', 'seed_reference_audio_urls'];
+
+// The card mounts the moment the tool is CALLED, so the only thing it knows is
+// the raw tool input - and for an edit/elements call the submit that follows is
+// the LONGEST wait on the card (local files are re-hosted first). Map the input
+// onto the same shape renderChips already reads for a server payload so the
+// references, DNA count and settings are on screen immediately instead of after
+// a minute of blank skeleton.
+// Deliberately NOT shown here: model, voice, DNA and moodboard NAMES. Those are
+// resolved server-side and arrive with the result (which overwrites all of
+// this) - rendering the raw identifier the caller passed would put an id on the
+// card, which is never allowed.
+function preRefSc(toolName, a) {
+  var img = [], vid = [], aud = [];
+  var take = function (v, bucket) {
+    if (typeof v === 'string') {
+      // http(s) only - an absolute local path is not loadable from the iframe.
+      if (/^https?:/i.test(v) && bucket.indexOf(v) < 0) bucket.push(v);
+      return;
+    }
+    if (Array.isArray(v)) {
+      v.forEach(function (item) {
+        take(typeof item === 'string' ? item : (item && (item.image_url || item.url)), bucket);
+      });
+    }
+  };
+  PRE_IMAGE_KEYS.forEach(function (k) { take(a[k], img); });
+  PRE_VIDEO_KEYS.forEach(function (k) { take(a[k], vid); });
+  PRE_AUDIO_KEYS.forEach(function (k) { take(a[k], aud); });
+  return {
+    tool: toolName,
+    kind: kindFromTool(toolName, null),
+    count: a.num_images || (Array.isArray(a.prompts) ? a.prompts.length : 1),
+    reference_images: img,
+    reference_videos: vid,
+    reference_audio: aud,
+    settings: {
+      duration: a.duration,
+      resolution: a.resolution,
+      aspect_ratio: a.aspect_ratio,
+      quality: a.quality,
+      mode: a.mode,
+      cinematic: a.cinematic,
+      visual_dna_ids: a.visual_dna_ids,
+      moodboard_ids: a.moodboard_ids,
+      moodboard_id: a.moodboard_id,
+      preset_id: a.preset_id
+    }
+  };
+}
+
 function bootPre(toolName, args) {
   if (toolName) originTool = toolName;
   if (args) originArgs = args;
@@ -1347,6 +1405,7 @@ function bootPre(toolName, args) {
     setPrompt(promptHTML(raw), raw);
   }
   setPhaseChip('Preparing', true);
+  renderChips(preRefSc(toolName, args || {}));
   if (!el('stage').innerHTML) {
     el('stage').innerHTML = '<div class="k-gen-grid n1"><div class="k-skel video" style="min-height:100px;max-height:140px"></div></div>';
   }