@slatesvideo/shared 0.6.10 → 0.6.11

package/dist/prompts/model-facts.js

@@ -69,14 +69,14 @@ export function multimodalRefSummary(id) {
         return '';
     const parts = [];
     if (v > 0)
-        parts.push(`${v} reference video${v === 1 ? '' : 's'} (${f.maxReferenceVideoSeconds}s combined)`);
+        parts.push(`${v} video${v === 1 ? '' : 's'} (${f.maxReferenceVideoSeconds}s total)`);
     if (a > 0)
-        parts.push(`${a} reference audio clip${a === 1 ? '' : 's'} (${f.maxReferenceAudioSeconds}s combined)`);
-    const total = f.maxReferenceFilesTotal ? `, ${f.maxReferenceFilesTotal} files max across all modalities` : '';
+        parts.push(`${a} audio clip${a === 1 ? '' : 's'} (${f.maxReferenceAudioSeconds}s total)`);
+    const total = f.maxReferenceFilesTotal ? `, ${f.maxReferenceFilesTotal} files max` : '';
     const companion = f.audioRefNeedsCompanion
-        ? ' Audio needs at least one image or video reference alongside it.'
-        : ' Audio-only references are allowed.';
-    return `${f.label}: up to ${parts.join(' and ')}${total}.${companion}`;
+        ? '; audio needs an image or video alongside.'
+        : '; audio-only is allowed.';
+    return `${f.label}: up to ${parts.join(' + ')}${total}${companion}`;
 }
 /**
  * The prompt words that make Seedance 2.5 reclassify a reference-carrying
@@ -120,6 +120,7 @@ export const MODEL_FACTS = [
     {
         id: 'nano-banana-2',
         route: 'generate',
+        tier: 'default',
         // Gemini 3.1 FLASH Image — verified against the runtime slug map in
         // slate/src/main/api/google.ts. Nano Banana PRO is a different model
         // (gemini-3-pro-image-preview); do not conflate them.
@@ -132,6 +133,7 @@ export const MODEL_FACTS = [
     {
         id: 'nano-banana-2-lite',
         route: 'generate',
+        tier: 'specialist',
         label: 'Nano Banana 2 Lite',
         kind: 'image',
         ...caps('nano-banana-2-lite'),
@@ -140,6 +142,7 @@ export const MODEL_FACTS = [
     {
         id: 'nano-banana-pro',
         route: 'generate',
+        tier: 'specialist',
         label: 'Nano Banana Pro',
         kind: 'image',
         ...caps('nano-banana-pro'),
@@ -148,6 +151,7 @@ export const MODEL_FACTS = [
     {
         id: 'gpt-image-2-5-flare',
         route: 'generate',
+        tier: 'specialist',
         label: 'GPT Image 2.5 Flare',
         kind: 'image',
         ...caps('gpt-image-2-5-flare'),
@@ -156,6 +160,7 @@ export const MODEL_FACTS = [
     {
         id: 'gpt-image-2-5-sunburst',
         route: 'generate',
+        tier: 'specialist',
         label: 'GPT Image 2.5 Sunburst',
         kind: 'image',
         ...caps('gpt-image-2-5-sunburst'),
@@ -164,6 +169,7 @@ export const MODEL_FACTS = [
     {
         id: 'flux-2-max',
         route: 'generate',
+        tier: 'specialist',
         label: 'FLUX.2 Max',
         kind: 'image',
         ...caps('flux-2-max'),
@@ -172,6 +178,7 @@ export const MODEL_FACTS = [
     {
         id: 'seedream-5-lite',
         route: 'generate',
+        tier: 'specialist',
         label: 'Seedream 5 Lite',
         kind: 'image',
         ...caps('seedream-5-lite'),
@@ -180,25 +187,28 @@ export const MODEL_FACTS = [
     {
         id: 'seedance-2',
         route: 'generate',
+        tier: 'specialist',
         label: 'Seedance 2.0',
         kind: 'video',
         ...caps('seedance-2'),
         audioRefNeedsCompanion: true,
-        notes: 'PREMIUM video tier and the DEFAULT video model — route here the moment physics, effects, destruction or scale matter, and for hero shots. VIDEO-ONLY. Strong image-to-video and own-footage restyle. 4K is Pro-gated (base accounts get PRO_REQUIRED). Stays the default over 2.5: it is the only Seedance with native 4K and it is cheaper at every tier the two share.',
+        notes: 'THE 4K AND VALUE SEAT beside the 2.5 default — the only Seedance with native 4K (Pro-gated; base accounts get PRO_REQUIRED) and cheaper than 2.5 at every resolution they share, with the same physics, effects and scale strengths; shorter takes, fewer references, no timestamps. VIDEO-ONLY. A bare "seedance" still resolves here for older CLIs that expect 4K.',
     },
     {
         id: 'seedance-2.5',
         route: 'generate',
+        tier: 'default',
         label: 'Seedance 2.5',
         kind: 'video',
         ...caps('seedance-2.5'),
         // No companion requirement — audio-only references are one of the things
         // the second seat actually buys.
-        notes: 'A SECOND SEAT NEXT TO 2.0, NOT AN UPGRADE — and the dearer one at every tier they share. Pick 2.5 when the shot needs LENGTH, MANY references, an AUDIO-ONLY reference, TIMED BEATS, or tighter prompt adherence; pick 2.0 for 4K and for the same resolution cheaper. VIDEO-ONLY. Timestamp grammar, and the edit/extend words that make the provider reclassify a fresh generation and fail it, are in slates-prompting-seedance-2-5.',
+        notes: 'DEFAULT VIDEO MODEL — the strongest seat for physics, effects, scale and hero shots, and the only Seedance that takes long single takes, many references, audio-only references and integer-second timestamps. No 4K, and dearer than 2.0 at every shared resolution: go to 2.0 for 4K or the same resolution cheaper. LENGTH is the price dial — quote long takes first. VIDEO-ONLY. Timestamp grammar and the edit/extend words that make the provider reclassify and fail a generation are in slates-prompting-seedance-2-5.',
     },
     {
         id: 'seedance-2.5-edit',
         route: 'edit',
+        tier: 'specialist',
         label: 'Seedance 2.5 Edit',
         kind: 'video',
         // 0 ingredients: prompt + source clip only on slates_edit_video.
@@ -208,15 +218,17 @@ export const MODEL_FACTS = [
     {
         id: 'kling-v3',
         route: 'generate',
+        tier: 'specialist',
         label: 'Kling 3.0',
         kind: 'video',
         // Family-level fact — caps are identical across std/pro/omni/omni-pro.
         ...caps('kling-v3.0-std'),
-        notes: 'DEFAULT general-purpose video model — cost-effective, strong start-frame adherence (identity, layout, text), acting, dialogue, lip-sync, and the widest aspect-ratio set. Escalate to Seedance for physics. Kling is also the ONLY engine behind the Motion Transfer and Lip Sync tools.',
+        notes: 'THE COST-EFFECTIVE SEAT — strong start-frame adherence (identity, layout, text), acting, dialogue, lip-sync and the widest aspect-ratio set; pick it when the budget matters and the shot is a performance or a start-frame animation. Kling is also the ONLY engine behind the Motion Transfer and Lip Sync tools.',
     },
     {
         id: 'kling-v3-edit',
         route: 'edit',
+        tier: 'specialist',
         label: 'Kling O3 Video Edit',
         kind: 'video',
         // Family-level fact; 4 = combined subject elements + style refs per edit.
@@ -226,15 +238,17 @@ export const MODEL_FACTS = [
     {
         id: 'veo-3.1',
         route: 'generate',
+        tier: 'niche',
         label: 'Veo 3.1',
         kind: 'video',
         // Family-level fact — fast and standard declare the same caps.
         ...caps('veo-3.1-fast'),
-        notes: 'NICHE, never the default — pick only when native synchronized audio must generate WITH the video in one pass, and the narrowest aspect-ratio and duration sets in the catalogue are acceptable. Otherwise Kling (default) or Seedance (physics/premium) win.',
+        notes: 'NICHE, never the default — pick only when native synchronized audio must generate WITH the video in one pass, and the narrowest aspect-ratio and duration sets in the catalogue are acceptable. Otherwise Seedance 2.5 (the default) or Kling (cost-effective performance) win.',
     },
     {
         id: 'omni-flash',
         route: 'generate',
+        tier: 'specialist',
         label: 'Gemini Omni Flash',
         kind: 'video',
         // 7 ref2v image_urls — mirrors Google's own reference limit.
@@ -244,6 +258,7 @@ export const MODEL_FACTS = [
     {
         id: 'omni-flash-edit',
         route: 'edit',
+        tier: 'default',
         label: 'Omni Flash Edit',
         kind: 'video',
         // 0: prompt + source clip ONLY — no element/style refs on this endpoint.
@@ -253,6 +268,7 @@ export const MODEL_FACTS = [
     {
         id: 'minimax-h3',
         route: 'generate',
+        tier: 'specialist',
         label: 'MiniMax H3',
         kind: 'video',
         ...caps('minimax-h3'),
@@ -265,16 +281,24 @@ export const MODEL_FACTS = [
     {
         id: 'minimax-h3-max',
         route: 'generate',
+        tier: 'specialist',
         label: 'MiniMax H3 Max',
         kind: 'video',
-        // No reference caps: fal publishes no reference-to-video endpoint for this
-        // row, so `caps()` returns nulls and the composer refuses references.
+        // References landed 2026-09-09 when the "h3-max/reference-to-video returns
+        // 404" claim was retired against the fetched schema (caps come from
+        // model-capabilities.ts; the rip is second-brain/business/projects/slates/
+        // provider-docs/fal-minimax-h3-openapi-schemas.md, section 6).
         ...caps('minimax-h3-max'),
-        notes: 'THE SPEED SEAT, and the DEARER one at the tier they share — never the cheap H3 and never the default. fal\'s post-train of the H3 weights: MEASURED 2026-08-27 at about 12x faster than base H3 on the same prompt and params, queue to finished file, plus a thin vendor-reported quality edge. It gives up the upper resolution tiers and the REFERENCE endpoint, so the omni-reference set is base-H3 only — but it still animates start and end frames, which is one of the two things it is FOR. Never describe this row as taking no image input. Route here when a fast turnaround on text-to-video or a start-frame shot is worth the premium.',
+        // Quoted off THAT endpoint's reference_audio_urls description, not copied
+        // from the base row: "Audio cannot be the only reference input; provide at
+        // least one reference image or video with it."
+        audioRefNeedsCompanion: true,
+        notes: 'THE SPEED SEAT, and the DEARER one at the tier they share — never the cheap H3 and never the default. fal\'s post-train of the H3 weights: MEASURED 2026-08-27 at about 12x faster than base H3 on the same prompt and params, queue to finished file, plus a thin vendor-reported quality edge. It gives up the upper resolution tiers. It takes the same omni-reference set as base H3 and animates start and end frames — but not both in one call: its reference endpoint has no start/end-frame fields, where base H3\'s does. Never describe this row as taking no image or reference input. Route here when a fast turnaround on text-to-video or a start-frame shot is worth the premium.',
     },
     {
         id: 'ltx-2-5',
         route: 'generate',
+        tier: 'specialist',
         label: 'LTX-2.5',
         kind: 'video',
         // No reference caps: fal publishes text-to-video and image-to-video for LTX
@@ -286,6 +310,7 @@ export const MODEL_FACTS = [
     {
         id: 'ltx-2-5-pro',
         route: 'generate',
+        tier: 'specialist',
         label: 'LTX-2.5 Pro',
         kind: 'video',
         ...caps('ltx-2-5-pro'),
@@ -294,6 +319,7 @@ export const MODEL_FACTS = [
     {
         id: 'seed-audio',
         route: 'generate',
+        tier: 'default',
         label: 'Seed Audio 1.0',
         kind: 'audio',
         // ONE image XOR up to 3 audio clips — the two inputs are mutually exclusive.
@@ -303,6 +329,7 @@ export const MODEL_FACTS = [
     {
         id: 'eleven-sfx',
         route: 'generate',
+        tier: 'specialist',
         label: 'ElevenLabs Sound Effects v2',
         kind: 'audio',
         ...caps('eleven-sfx'),
@@ -311,12 +338,30 @@ export const MODEL_FACTS = [
     {
         id: 'inworld-tts-2',
         route: 'generate',
+        tier: 'specialist',
         label: 'Inworld Realtime TTS-2',
         kind: 'audio',
         ...caps('inworld-tts-2'),
         notes: 'THE VOICE SEAT — one named voice saying one line, billed per CHARACTER not per second. Route here when WHO is speaking matters. NOT scene audio — that is seed-audio; a single effect is eleven-sfx.',
     },
 ];
+/**
+ * One default per lane, asserted where the data is declared. Two rows both
+ * reading "DEFAULT" is exactly the drift `tier` exists to make impossible, and
+ * a lane with no default leaves an agent (and slates-web) nothing to lead with.
+ */
+for (const kind of ['image', 'video', 'audio']) {
+    for (const route of ['generate', 'edit']) {
+        const lane = MODEL_FACTS.filter((f) => f.kind === kind && f.route === route);
+        if (lane.length === 0)
+            continue;
+        const defaults = lane.filter((f) => f.tier === 'default').map((f) => f.id);
+        if (defaults.length !== 1) {
+            throw new Error(`MODEL_FACTS: ${kind}/${route} must have exactly one tier: 'default' row, found ${defaults.length}` +
+                (defaults.length ? ` (${defaults.join(', ')})` : ''));
+        }
+    }
+}
 const FACT_BY_ID = new Map(MODEL_FACTS.map((m) => [m.id, m]));
 /**
  * Routing prose for one lane, generated from the SSOT.

package/dist/prompts/prompting-tips.js

@@ -115,7 +115,7 @@ const SEEDANCE_25 = {
     ...SEEDANCE,
     label: 'Seedance 2.5',
     intro: [
-        'Seedance 2.5 is a SECOND SEAT next to 2.0, not an upgrade of it. It buys one 30-second take instead of 15, up to 30 image references (plus 10 video and 10 audio), audio-only references and integer-second timestamps — and it gives up 4K. It runs at 480p, 720p or 1080p, and it costs more than 2.0 at every tier they share, so pick 2.0 when you want the same resolution cheaper or you want 4K.',
+        'Seedance 2.5 is the default video model. Against 2.0 it buys one 30-second take instead of 15, up to 30 image references (plus 10 video and 10 audio), audio-only references and integer-second timestamps — and it gives up 4K. It runs at 480p, 720p or 1080p, and it costs more than 2.0 at every tier they share, so pick 2.0 when you want the same resolution cheaper or you want 4K.',
         "ByteDance's official advanced formula has 8 slots: precise subject + action details + scene/environment + lighting & color tone + camera movement + visual style + image quality + constraints. Sweet spot 60-150 words for a single shot, longer for multi-shot.",
     ],
     columns: [
@@ -348,7 +348,7 @@ const OMNI_FLASH = {
             },
             {
                 heading: 'Know its seat',
-                note: 'Cheap drafts, iteration volume, and audio-in-one-gen at low cost. For hero shots, Kling 3.0 (general default) or Seedance 2.0 (premium/physics) still win.',
+                note: 'Cheap drafts, iteration volume, and audio-in-one-gen at low cost. For hero shots, Seedance 2.5 (the default) or Seedance 2.0 (4K, cheaper) still win.',
             },
         ],
     ],