fix(wizard): no encolar generación ante preguntas how-to

ui/src/features/agent/agentActions.ts

@@ -1870,13 +1870,13 @@ export function isExplicitRetryRequest(request: string): boolean {
 
 export function isExplicitVideoGenerationRequest(request: string): boolean {
   const text = request.trim()
-  if (!text || NEGATED_VIDEO_REQUEST.test(text)) return false
+  if (!text || NEGATED_VIDEO_REQUEST.test(text) || isHowToGenerateQuestion(text)) return false
   return EXPLICIT_VIDEO_REQUESTS.some(pattern => pattern.test(text))
 }
 
 export function isResumePreparedStudioVideoRequest(request: string): boolean {
   const text = request.trim()
-  if (!text || NEGATED_VIDEO_REQUEST.test(text)) return false
+  if (!text || NEGATED_VIDEO_REQUEST.test(text) || isHowToGenerateQuestion(text)) return false
   const match = text.match(
     /\b(?:genera|generad|lanza|lanzad|encola|encolad|env[ií]a|enviad|start|queue|launch)\b[^.!?\n]{0,24}\b(?:el|la|este|esta|the)\s+(?:video|v[ií]deo|clip)\b(.*)$/i,
   )
@@ -1900,7 +1900,7 @@ const EXPLICIT_IMAGE_REQUESTS = [
 
 export function isExplicitImageGenerationRequest(request: string): boolean {
   const text = request.trim()
-  if (!text || NEGATED_VIDEO_REQUEST.test(text)) return false
+  if (!text || NEGATED_VIDEO_REQUEST.test(text) || isHowToGenerateQuestion(text)) return false
   if (isExplicitVideoGenerationRequest(text)) return false
   return EXPLICIT_IMAGE_REQUESTS.some(pattern => pattern.test(text))
 }
@@ -1920,7 +1920,7 @@ const STUDIO_AUDIO_CONTEXT = [
 export function isExplicitAudioGenerationRequest(request: string): boolean {
   const text = request.trim()
   if (!text || NEGATED_VIDEO_REQUEST.test(text) || MUSIC_VIDEO_CONTEXT.test(text)) return false
-  if (isExplicitSfxGenerationRequest(text)) return false
+  if (isHowToGenerateQuestion(text) || isExplicitSfxGenerationRequest(text)) return false
   if (!STUDIO_AUDIO_CONTEXT.some(pattern => pattern.test(text))) return false
   return EXPLICIT_AUDIO_GENERATION_REQUESTS.some(pattern => pattern.test(text))
 }
@@ -1941,7 +1941,7 @@ const EXPLICIT_3D_REQUESTS = [
 
 export function isExplicit3dGenerationRequest(request: string): boolean {
   const text = request.trim()
-  if (!text || NEGATED_VIDEO_REQUEST.test(text)) return false
+  if (!text || NEGATED_VIDEO_REQUEST.test(text) || isHowToGenerateQuestion(text)) return false
   if (isExplicitVideoGenerationRequest(text) || isExplicitImageGenerationRequest(text) || isExplicitSfxGenerationRequest(text)) return false
   return EXPLICIT_3D_REQUESTS.some(pattern => pattern.test(text))
 }
@@ -2015,8 +2015,11 @@ const HOW_TO_GENERATE = [
 
 export function isHowToGenerateQuestion(request: string): boolean {
   const text = request.trim()
-  if (!text || text.length > 240) return false
-  return HOW_TO_GENERATE.some(pattern => pattern.test(text))
+  if (!text) return false
+  // Classify from the opening window so a long explanation after
+  // "how do I generate a video?" stays educational. A later command
+  // after 240 characters is treated as a separate request.
+  return HOW_TO_GENERATE.some(pattern => pattern.test(text.slice(0, 240)))
 }
 
 const LABS_INVENTORY = /(?:¿\s*)?(?:qu[eé]\s+puedes\s+hacer|what\s+can\s+you\s+do)(?:\s+(?:en|in|con|with))?\s+(?:el\s+)?(?:series\s+lab|story\s+lab)/i
@@ -2145,6 +2148,26 @@ export async function reconcileAgentTurnWithRequest(
       }
     : candidate
   turn = preserveExactEpisodeTitle(turn)
+  if (isComicLaunchHowQuestion(request, history)) {
+    return {
+      reply: [
+        'No hay un botón llamado **Render page**.',
+        'El dibujo de las viñetas es **Generate all images** en Comic Director (barra de Comics), o dímelo aquí: **lánzalo**.',
+        'Las viñetas entran en la **misma GPU**, una detrás de otra, no en paralelo. No es un segundo motor.',
+      ].join('\n\n'),
+      actions: [{ type: 'open_tab', tab: 'comics' }],
+    }
+  }
+  if (isHowToGenerateQuestion(request)) {
+    return {
+      ...turn,
+      actions: turn.actions.filter(action => (
+        action.type === 'open_tab'
+        || action.type === 'open_story_section'
+        || action.type === 'open_series_section'
+      )),
+    }
+  }
   if (isResumePreparedStudioVideoRequest(request)) {
     const existing = turn.actions.find(
       (action): action is AgentPrepareVideoAction => action.type === 'prepare_video',
@@ -2183,29 +2206,9 @@ export async function reconcileAgentTurnWithRequest(
       actions: [rhythmic3dWorkflow],
     }
   }
-  if (isComicLaunchHowQuestion(request, history)) {
-    return {
-      reply: [
-        'No hay un botón llamado **Render page**.',
-        'El dibujo de las viñetas es **Generate all images** en Comic Director (barra de Comics), o dímelo aquí: **lánzalo**.',
-        'Las viñetas entran en la **misma GPU**, una detrás de otra, no en paralelo. No es un segundo motor.',
-      ].join('\n\n'),
-      actions: [{ type: 'open_tab', tab: 'comics' }],
-    }
-  }
   if (isLabsInventoryQuestion(request)) {
     return { ...turn, actions: [] }
   }
-  if (isHowToGenerateQuestion(request)) {
-    return {
-      ...turn,
-      actions: turn.actions.filter(action => (
-        action.type === 'open_tab'
-        || action.type === 'open_story_section'
-        || action.type === 'open_series_section'
-      )),
-    }
-  }
   if (!requestAuthorizesEditorialCommit(request)) {
     const actions = turn.actions.filter(action => !EDITORIAL_COMMIT_TYPES.has(action.type))
     if (actions.length !== turn.actions.length) turn = { ...turn, actions }

ui/tests/agentActions.test.mjs

@@ -1985,6 +1985,71 @@ test('genera el video de un mapache uses the new topic instead of the prepared f
   useStore.setState(original)
 })
 
+test('how-to launch questions do not queue a prepared Studio video', async () => {
+  const { useStore } = await import('../src/stores/useStore.ts')
+  const {
+    isHowToGenerateQuestion,
+    isResumePreparedStudioVideoRequest,
+    isExplicitVideoGenerationRequest,
+    isExplicitAudioGenerationRequest,
+    reconcileAgentTurnWithRequest,
+  } = await import('../src/features/agent/agentActions.ts')
+  const original = {
+    generationMode: useStore.getState().generationMode,
+    params: useStore.getState().params,
+    savedPromptPerMode: useStore.getState().savedPromptPerMode,
+  }
+  useStore.setState({
+    generationMode: 'video',
+    params: { ...useStore.getState().params, prompt: 'un mago en la torre' },
+    savedPromptPerMode: { ...useStore.getState().savedPromptPerMode, video: 'un mago en la torre' },
+  })
+
+  try {
+    for (const question of [
+      'How do I launch the video?',
+      'How do I start the video?',
+      'Explain how to launch the video',
+    ]) {
+      assert.equal(isHowToGenerateQuestion(question), true, question)
+      assert.equal(isResumePreparedStudioVideoRequest(question), false, question)
+      assert.equal(isExplicitVideoGenerationRequest(question), false, question)
+      const turn = await reconcileAgentTurnWithRequest(question, {
+        reply: 'I will launch it.',
+        actions: [{ type: 'prepare_video', prompt: 'un mago en la torre' }, { type: 'start_generation', confirm: true }],
+      })
+      assert.deepEqual(turn.actions.map(action => action.type), [], question)
+    }
+
+    const longVideo = `How do I generate a video? ${'Please explain the workflow in detail. '.repeat(8)}`
+    assert.ok(longVideo.length > 240)
+    assert.equal(isHowToGenerateQuestion(longVideo), true)
+    assert.equal(isExplicitVideoGenerationRequest(longVideo), false)
+    const longVideoTurn = await reconcileAgentTurnWithRequest(longVideo, { reply: 'Here is how.', actions: [] })
+    assert.deepEqual(longVideoTurn.actions.map(action => action.type), [])
+
+    const longAudio = `How can I generate music in Studio Audio? ${'Please explain the workflow in detail. '.repeat(8)}`
+    assert.ok(longAudio.length > 240)
+    assert.equal(isHowToGenerateQuestion(longAudio), true)
+    assert.equal(isExplicitAudioGenerationRequest(longAudio), false)
+    const longAudioTurn = await reconcileAgentTurnWithRequest(longAudio, {
+      reply: 'I will generate it now.',
+      actions: [
+        { type: 'open_tab', tab: 'audio' },
+        { type: 'prepare_audio', subMode: 'music', prompt: 'Piano' },
+        { type: 'start_generation', confirm: true },
+      ],
+    })
+    assert.deepEqual(longAudioTurn.actions.map(action => action.type), ['open_tab'])
+
+    const launch = await reconcileAgentTurnWithRequest('lanza el vídeo', { reply: '¿De qué?', actions: [] })
+    assert.deepEqual(launch.actions.map(action => action.type), ['prepare_video', 'start_generation'])
+    assert.equal(launch.actions[0].prompt, 'un mago en la torre')
+  } finally {
+    useStore.setState(original)
+  }
+})
+
 test('genera el video does not copy an incompatible I2V, audio or 3D model as T2V', async () => {
   const { useStore } = await import('../src/stores/useStore.ts')
   const { reconcileAgentTurnWithRequest } = await import('../src/features/agent/agentActions.ts')