@sogni-ai/sogni-intelligence-client 3.14.4 → 3.15.0

package/dist/public-skill-runtime/index.js

@@ -1,6 +1,6 @@
 "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
-exports.STORYBOARD_DEFAULT_MAX_FRAMES = exports.STORYBOARD_DEFAULT_MIN_FRAMES = exports.STORYBOARD_PLANNING_CONTRACT_JSON_SCHEMA = exports.STORYBOARD_PLANNING_CONTRACT_SCHEMA_VERSION = exports.IMAGE_MODEL_ALIASES = exports.IMAGE_MODEL_DEFAULTS = exports.QUALITY_TIERS = exports.VIDEO_MODEL_ALIASES = exports.VIDEO_WORKFLOW_DEFAULT_MODELS = exports.EXPANDED_VIDEO_MODEL_REGISTRY = exports.VIDEO_MODEL_REGISTRY = exports.HAPPYHORSE_WORKFLOW_MODELS = exports.SEEDANCE_WORKFLOW_MODELS = exports.LTX23_DEV_WORKFLOW_MODELS = exports.LTX23_EROS_MODEL_ID = exports.LTX23_WORKFLOW_MODELS = exports.DEFAULT_STORYBOARD_CANVAS_HINT_MARKER = exports.GPT_IMAGE_STORYBOARD_DEFAULTS = exports.SEEDANCE_V2V_REFERENCE_MAX_DURATION_SECONDS = exports.SEEDANCE_STORYBOARD_REFERENCE_PROMPT = exports.storyboardAdapterRegistry = exports.SkillRegistry = exports.StoryboardAdapterUnsupportedStageError = exports.ALL_BUILT_IN_SKILLS = exports.COMPOSITION_PLANNING_SKILL = exports.APP_SETTINGS_SKILL = exports.PERSONA_MANAGEMENT_SKILL = exports.MEDIA_ANALYSIS_SKILL = exports.MUSIC_GENERATION_SKILL = exports.VIDEO_EDITING_SKILL = exports.VIDEO_GENERATION_SKILL = exports.IMAGE_EDITING_SKILL = exports.IMAGE_GENERATION_SKILL = exports.QUALITY_AUDIT_SKILL = exports.ASSET_REFERENCE_MANAGEMENT_SKILL = exports.SESSION_CONTROL_SKILL = exports.TOOL_RESULT_DELIMITERS = exports.PUBLIC_SKILL_DEFAULT_TOOL_DEFINITIONS = exports.PUBLIC_SKILL_DEFAULT_TOOL_NAMES = exports.PUBLIC_SKILL_DEFAULT_REPAIR_RECIPES = exports.PUBLIC_SKILL_DEFAULT_PROMPT_CONTRACTS = exports.PUBLIC_SKILL_DEFAULT_POLICIES = exports.SKILL_RUNTIME_VERSION = void 0;
+exports.STORYBOARD_DEFAULT_MAX_FRAMES = exports.STORYBOARD_DEFAULT_MIN_FRAMES = exports.STORYBOARD_PLANNING_CONTRACT_JSON_SCHEMA = exports.STORYBOARD_PLANNING_CONTRACT_SCHEMA_VERSION = exports.IMAGE_MODEL_ALIASES = exports.IMAGE_MODEL_DEFAULTS = exports.QUALITY_TIERS = exports.VIDEO_MODEL_ALIASES = exports.VIDEO_WORKFLOW_DEFAULT_MODELS = exports.EXPANDED_VIDEO_MODEL_REGISTRY = exports.VIDEO_MODEL_REGISTRY = exports.HAPPYHORSE_WORKFLOW_MODELS = exports.SEEDANCE_WORKFLOW_MODELS = exports.LTX25_DEV_WORKFLOW_MODELS = exports.LTX25_WORKFLOW_MODELS = exports.LTX23_DEV_WORKFLOW_MODELS = exports.LTX23_EROS_MODEL_ID = exports.LTX23_WORKFLOW_MODELS = exports.DEFAULT_STORYBOARD_CANVAS_HINT_MARKER = exports.GPT_IMAGE_STORYBOARD_DEFAULTS = exports.SEEDANCE_V2V_REFERENCE_MAX_DURATION_SECONDS = exports.SEEDANCE_STORYBOARD_REFERENCE_PROMPT = exports.storyboardAdapterRegistry = exports.SkillRegistry = exports.StoryboardAdapterUnsupportedStageError = exports.ALL_BUILT_IN_SKILLS = exports.COMPOSITION_PLANNING_SKILL = exports.APP_SETTINGS_SKILL = exports.PERSONA_MANAGEMENT_SKILL = exports.MEDIA_ANALYSIS_SKILL = exports.MUSIC_GENERATION_SKILL = exports.VIDEO_EDITING_SKILL = exports.VIDEO_GENERATION_SKILL = exports.IMAGE_EDITING_SKILL = exports.IMAGE_GENERATION_SKILL = exports.QUALITY_AUDIT_SKILL = exports.ASSET_REFERENCE_MANAGEMENT_SKILL = exports.SESSION_CONTROL_SKILL = exports.TOOL_RESULT_DELIMITERS = exports.PUBLIC_SKILL_DEFAULT_TOOL_DEFINITIONS = exports.PUBLIC_SKILL_DEFAULT_TOOL_NAMES = exports.PUBLIC_SKILL_DEFAULT_REPAIR_RECIPES = exports.PUBLIC_SKILL_DEFAULT_PROMPT_CONTRACTS = exports.PUBLIC_SKILL_DEFAULT_POLICIES = exports.SKILL_RUNTIME_VERSION = void 0;
 exports.createPublicSkillContractRuntime = createPublicSkillContractRuntime;
 exports.createPublicSkillDefaultContractRuntime = createPublicSkillDefaultContractRuntime;
 exports.classifyPublicSkillTurn = classifyPublicSkillTurn;
@@ -21,6 +21,7 @@ exports.compileForModel = compileForModel;
 exports.normalizeReferenceAudioMimeType = normalizeReferenceAudioMimeType;
 exports.detectReferenceAudioFormat = detectReferenceAudioFormat;
 exports.shouldTrimSeedanceV2VSourceVideo = shouldTrimSeedanceV2VSourceVideo;
+exports.resolveLtx25WorkflowModelForQuality = resolveLtx25WorkflowModelForQuality;
 exports.resolveLtx23WorkflowModelForQuality = resolveLtx23WorkflowModelForQuality;
 exports.isLtx2Model = isLtx2Model;
 exports.isWanModel = isWanModel;
@@ -1037,6 +1038,29 @@ const PUBLIC_GPT_IMAGE_2_ADAPTER = {
         return 'GPT IMAGE 2 ROUTING: When the user asks for a ChatGPT, OpenAI, GPT, GPT-2, GPT Image, or gpt-image-2 image/model, use model="gpt-image-2". Use generate_image for text-to-image requests. If uploaded/reference/persona images must guide identity, likeness, composition, style, or objects, use edit_image with model="gpt-image-2" instead of forcing generate_image.';
     },
 };
+const PUBLIC_LTX25_ADAPTER = {
+    modelId: 'ltx25',
+    name: 'LTX 2.5',
+    supportedStages: ['scene_clip'],
+    compile(storyboard, input) {
+        if (input.stage !== 'scene_clip')
+            throw new StoryboardAdapterUnsupportedStageError('ltx25', input.stage);
+        const scene = requireStoryboardScene('LTX25_ADAPTER', input);
+        const referenceTag = input.primaryReferenceTag ?? 'context_image_0';
+        return {
+            stage: 'scene_clip',
+            prompt: compactSceneVideoPrompt(storyboard, scene, referenceTag),
+            args: {
+                videoModel: 'ltx25',
+                duration: clampSeedanceStoryboardDuration(scene.durationSec ?? 5),
+                aspectRatio: storyboard.targetVideoAspectRatio,
+            },
+        };
+    },
+    getSystemPromptGuidance() {
+        return 'LTX 2.5 VIDEO PROMPTING: For image-to-video, describe motion, action, camera, dialogue, and sound not already obvious in the reference frame. Keep recurring character names and visual anchors stable across scenes. Use LTX 2.3 only for explicit rollback or its ID-LoRA/transition/10Eros-only paths.';
+    },
+};
 const PUBLIC_LTX23_ADAPTER = {
     modelId: 'ltx23',
     name: 'LTX-2.3',
@@ -1088,6 +1112,7 @@ const PUBLIC_WAN_ADAPTER = {
 const PUBLIC_STORYBOARD_ADAPTERS = [
     PUBLIC_SEEDANCE_ADAPTER,
     PUBLIC_GPT_IMAGE_2_ADAPTER,
+    PUBLIC_LTX25_ADAPTER,
     PUBLIC_LTX23_ADAPTER,
     PUBLIC_WAN_ADAPTER,
 ];
@@ -1146,8 +1171,12 @@ exports.storyboardAdapterRegistry = {
             return PUBLIC_SEEDANCE_ADAPTER;
         if (trimmed.startsWith('gpt-image'))
             return PUBLIC_GPT_IMAGE_2_ADAPTER;
-        if (trimmed.startsWith('ltx'))
+        if (trimmed.startsWith('ltx25'))
+            return PUBLIC_LTX25_ADAPTER;
+        if (trimmed.startsWith('ltx23') || trimmed.startsWith('ltx2-'))
             return PUBLIC_LTX23_ADAPTER;
+        if (trimmed.startsWith('ltx'))
+            return PUBLIC_LTX25_ADAPTER;
         if (trimmed.startsWith('wan'))
             return PUBLIC_WAN_ADAPTER;
         return null;
@@ -1240,6 +1269,25 @@ exports.LTX23_DEV_WORKFLOW_MODELS = Object.freeze({
     ia2v: 'ltx23-22b-fp8_ia2v_dev',
     a2v: 'ltx23-22b-fp8_a2v_dev'
 });
+exports.LTX25_WORKFLOW_MODELS = Object.freeze({
+    t2v: 'ltx25-22b-int8_t2v_distilled',
+    i2v: 'ltx25-22b-int8_i2v_distilled',
+    ia2v: 'ltx25-22b-int8_ia2v_distilled',
+    a2v: 'ltx25-22b-int8_a2v_distilled',
+    v2v: 'ltx25-22b-int8_v2v_distilled'
+});
+exports.LTX25_DEV_WORKFLOW_MODELS = Object.freeze({
+    t2v: 'ltx25-22b-int8_t2v_dev',
+    i2v: 'ltx25-22b-int8_i2v_dev',
+    ia2v: 'ltx25-22b-int8_ia2v_dev',
+    a2v: 'ltx25-22b-int8_a2v_dev',
+    v2v: 'ltx25-22b-int8_v2v_dev'
+});
+function resolveLtx25WorkflowModelForQuality(workflow, qualityTier) {
+    return qualityTier === 'pro'
+        ? exports.LTX25_DEV_WORKFLOW_MODELS[workflow]
+        : exports.LTX25_WORKFLOW_MODELS[workflow];
+}
 function resolveLtx23WorkflowModelForQuality(workflow, qualityTier) {
     if (qualityTier === 'pro' && workflow !== 'v2v') {
         return exports.LTX23_DEV_WORKFLOW_MODELS[workflow];
@@ -1507,6 +1555,30 @@ exports.VIDEO_MODEL_REGISTRY = Object.freeze({
 });
 exports.EXPANDED_VIDEO_MODEL_REGISTRY = (() => {
     const registry = { ...exports.VIDEO_MODEL_REGISTRY };
+    for (const workflow of ['t2v', 'i2v', 'ia2v', 'a2v', 'v2v']) {
+        const base = registry[exports.LTX23_WORKFLOW_MODELS[workflow]];
+        if (!base)
+            continue;
+        const ltx25Base = {
+            ...base,
+            family: 'ltx25',
+            maxDimension: 3840,
+            fps: 24,
+            sampler: 'euler_ancestral',
+            scheduler: 'manual_sigmas',
+            supportsNativeAudio: true
+        };
+        registry[exports.LTX25_WORKFLOW_MODELS[workflow]] = {
+            ...ltx25Base,
+            steps: 8,
+            guidance: 1.0
+        };
+        registry[exports.LTX25_DEV_WORKFLOW_MODELS[workflow]] = {
+            ...ltx25Base,
+            steps: 30,
+            guidance: 3.0
+        };
+    }
     for (const workflow of ['t2v', 'i2v', 'ia2v', 'a2v']) {
         const base = registry[exports.LTX23_WORKFLOW_MODELS[workflow]];
         if (!base)
@@ -1540,16 +1612,22 @@ exports.EXPANDED_VIDEO_MODEL_REGISTRY = (() => {
     return Object.freeze(registry);
 })();
 exports.VIDEO_WORKFLOW_DEFAULT_MODELS = Object.freeze({
-    t2v: exports.LTX23_WORKFLOW_MODELS.t2v,
-    i2v: 'wan_v2.2-14b-fp8_i2v_lightx2v',
+    t2v: exports.LTX25_WORKFLOW_MODELS.t2v,
+    i2v: exports.LTX25_WORKFLOW_MODELS.i2v,
     s2v: 'wan_v2.2-14b-fp8_s2v_lightx2v',
-    ia2v: exports.LTX23_WORKFLOW_MODELS.ia2v,
-    a2v: exports.LTX23_WORKFLOW_MODELS.a2v,
+    ia2v: exports.LTX25_WORKFLOW_MODELS.ia2v,
+    a2v: exports.LTX25_WORKFLOW_MODELS.a2v,
     'animate-move': 'wan_v2.2-14b-fp8_animate-move_lightx2v',
     'animate-replace': 'wan_v2.2-14b-fp8_animate-replace_lightx2v',
-    v2v: exports.LTX23_WORKFLOW_MODELS.v2v
+    v2v: exports.LTX25_WORKFLOW_MODELS.v2v
 });
 exports.VIDEO_MODEL_ALIASES = Object.freeze({
+    ltx25: exports.LTX25_WORKFLOW_MODELS.t2v,
+    'ltx25-t2v': exports.LTX25_WORKFLOW_MODELS.t2v,
+    'ltx25-i2v': exports.LTX25_WORKFLOW_MODELS.i2v,
+    'ltx25-ia2v': exports.LTX25_WORKFLOW_MODELS.ia2v,
+    'ltx25-a2v': exports.LTX25_WORKFLOW_MODELS.a2v,
+    'ltx25-v2v': exports.LTX25_WORKFLOW_MODELS.v2v,
     ltx23: exports.LTX23_WORKFLOW_MODELS.t2v,
     'ltx23-t2v': exports.LTX23_WORKFLOW_MODELS.t2v,
     'ltx23-i2v': exports.LTX23_WORKFLOW_MODELS.i2v,
@@ -1639,7 +1717,7 @@ exports.IMAGE_MODEL_ALIASES = Object.freeze({
     'dark-beast-krea-2-identity-edit-lora-v1-2': 'dark_beast_krea2_identity_edit_v1_2'
 });
 function isLtx2Model(modelId) {
-    return modelId?.startsWith('ltx2-') || modelId?.startsWith('ltx23-') || false;
+    return modelId?.startsWith('ltx2-') || modelId?.startsWith('ltx23-') || modelId?.startsWith('ltx25-') || false;
 }
 function isWanModel(modelId) {
     return modelId?.startsWith('wan_') || false;
@@ -1659,6 +1737,11 @@ function resolveVideoModelAlias(modelId, workflow) {
     if (!modelId)
         return modelId;
     const key = String(modelId).trim().toLowerCase();
+    if (key === 'ltx25') {
+        if (workflow && isLtxWorkflow(workflow)) {
+            return exports.LTX25_WORKFLOW_MODELS[workflow];
+        }
+    }
     if (key === 'ltx23' && isLtxWorkflow(workflow)) {
         return exports.LTX23_WORKFLOW_MODELS[workflow];
     }
@@ -1694,6 +1777,9 @@ function getBuiltinVideoModelConfig(modelId) {
     if (id.startsWith('ltx23-') && isLtxWorkflow(workflow)) {
         return exports.EXPANDED_VIDEO_MODEL_REGISTRY[exports.LTX23_WORKFLOW_MODELS[workflow]] || null;
     }
+    if (id.startsWith('ltx25-') && isLtxWorkflow(workflow)) {
+        return exports.EXPANDED_VIDEO_MODEL_REGISTRY[exports.LTX25_WORKFLOW_MODELS[workflow]] || null;
+    }
     if (id.startsWith('ltx2-')) {
         return {
             workflow,
@@ -1959,15 +2045,15 @@ function selectDefaultVideoModel(workflow, opts = {}, config) {
     if (configured)
         return resolveVideoModelAlias(configured, workflow) || null;
     if (workflow === 'ia2v')
-        return resolveLtx23WorkflowModelForQuality('ia2v', opts.quality);
+        return resolveLtx25WorkflowModelForQuality('ia2v', opts.quality);
     if (workflow === 'a2v')
-        return resolveLtx23WorkflowModelForQuality('a2v', opts.quality);
+        return resolveLtx25WorkflowModelForQuality('a2v', opts.quality);
     if (workflow === 'v2v')
-        return exports.LTX23_WORKFLOW_MODELS.v2v;
+        return resolveLtx25WorkflowModelForQuality('v2v', opts.quality);
     if (workflow === 't2v')
-        return resolveLtx23WorkflowModelForQuality('t2v', opts.quality);
+        return resolveLtx25WorkflowModelForQuality('t2v', opts.quality);
     if (workflow === 'i2v' && (opts.referenceAudioIdentity || promptNeedsLtxNativeAudio(opts.prompt) || opts.quality === 'hq' || opts.quality === 'pro')) {
-        return resolveLtx23WorkflowModelForQuality('i2v', opts.quality);
+        return resolveLtx25WorkflowModelForQuality('i2v', opts.quality);
     }
     return exports.VIDEO_WORKFLOW_DEFAULT_MODELS[workflow] || null;
 }
@@ -4119,7 +4205,7 @@ function extractStoryboardRequiredText(text) {
     }
     const labeledTextPattern = /\b(?:on[-\s]?screen\s+text|visible\s+text|text\s+only|text|end\s+card\s+text|final\s+text|tagline|cta|headline|title\s+card|copy)\s*:\s*([^\n]{1,260})/gi;
     for (const match of text.matchAll(labeledTextPattern)) {
-        const value = match[1] || '';
+        const value = (match[1] || '').split('|')[0];
         const matchStartInText = match.index ?? 0;
         const lineStart = text.lastIndexOf('\n', matchStartInText) + 1;
         const valueStartInMatch = match[0].indexOf(value);
@@ -4822,7 +4908,7 @@ function splitStoryboardTableSections(text) {
     return sections;
 }
 function splitStoryboardSceneSections(text) {
-    const matches = Array.from(text.matchAll(/^\s*(?:[-*+>#_]{1,6}\s*)*(?:[*_]{1,3})?\s*(?:Scene|Shot|Beat|Panel|Frame)\s*(\d{1,2})\b\s*(?:[-:.)|]\s*)?([^\n]*)/gim));
+    const matches = Array.from(text.matchAll(/^\s*(?:[-*+>#_]{1,6}\s*)*(?:[*_]{1,3})?\s*(?:Scene|Shot|Beat|Panel|Frame)\s*_?\s*(\d{1,2})\b\s*(?:[-:.)|]\s*)?([^\n]*)/gim));
     if (matches.length === 0)
         return [];
     return matches.map((match, index) => {
@@ -4888,7 +4974,7 @@ function splitStoryboardNumberedListSceneSections(text) {
     return [];
 }
 function splitStoryboardInlineSceneSections(text) {
-    const markerPattern = /(^|[\s.;])((?:Scene|Shot|Beat|Panel|Frame)\s*(\d{1,2})(?:\s*,\s*(?:Scene|Shot|Beat|Panel|Frame)\s*\d{1,2})?\s*(?:\([^)]{0,120}\))?\s*(?:[:\-–—]\s*)?)/gi;
+    const markerPattern = /(^|[\s.;])((?:Scene|Shot|Beat|Panel|Frame)\s*_?\s*(\d{1,2})(?:\s*,\s*(?:Scene|Shot|Beat|Panel|Frame)\s*_?\s*\d{1,2})?\s*(?:\([^)]{0,120}\))?\s*(?:[:\-–—]\s*)?)/gi;
     const matches = Array.from(text.matchAll(markerPattern))
         .map(match => {
         const leading = match[1] || '';
@@ -5946,7 +6032,7 @@ function buildStoryboardProject(options) {
             if (Number.isInteger(value) && value >= 1 && value <= 64)
                 numbersSeen.add(value);
         }
-        for (const match of text.matchAll(/^\s*(?:[-*+>#_]{1,6}\s*)*(?:[*_]{1,3})?\s*(?:Scene|Shot|Beat|Panel|Frame)\s*(\d{1,2})\b/gim)) {
+        for (const match of text.matchAll(/^\s*(?:[-*+>#_]{1,6}\s*)*(?:[*_]{1,3})?\s*(?:Scene|Shot|Beat|Panel|Frame)\s*_?\s*(\d{1,2})\b/gim)) {
             const value = Number(match[1]);
             if (Number.isInteger(value) && value >= 1 && value <= 64)
                 numbersSeen.add(value);