aigen/server/api/generate.post.ts

import { frameLength, runGeneration } from '~/server/utils/videoChain'
import { createShotQueue, finishQueueBurst, setQueueJob } from '~/server/utils/shotQueue'
import { defaultVideoSteps, isLtxWorkflow, isTextToVideo, parseVideoWorkflow } from '~/utils/videoModels'

function isComfyBusyTimeout(error: unknown) {
  const err = error as { statusCode?: number; status?: number; data?: { code?: string }; message?: string; statusMessage?: string }
  if (err?.statusCode === 409 || err?.status === 409 || err?.data?.code === 'COMFY_BUSY') return true
  const text = `${err?.statusMessage || ''} ${err?.message || ''}`
  return text.includes('still busy after 3 minutes')
}

function parseDuration(raw: unknown) {
  const seconds = Number(raw)
  if (!Number.isFinite(seconds)) return 5
  return Math.min(120, Math.max(0.5, Math.round(seconds * 10) / 10))
}

function parseExtendDuration(raw: unknown) {
  return parseDuration(raw)
}

function parseExtensions(raw: string | undefined) {
  if (!raw) return [] as { prompt: string; duration: number }[]
  try {
    const parsed = JSON.parse(raw)
    if (!Array.isArray(parsed)) return []
    return parsed
      .map((item: { prompt?: unknown; duration?: unknown }) => ({
        prompt: String(item?.prompt || '').trim(),
        duration: parseExtendDuration(item?.duration)
      }))
      .filter(item => item.prompt)
  } catch {
    return []
  }
}

const SAMPLERS = new Set(['res_multistep', 'euler', 'dpmpp_2m'])
const SCHEDULERS = new Set(['simple', 'ddim_uniform', 'sgm_uniform'])

function parseFps(raw: string | undefined) {
  const fps = Number(raw)
  return fps === 12 || fps === 30 || fps === 24 ? fps : 24
}

function parseCfg(raw: string | undefined, turbo: boolean, ltx = false) {
  const fallback = ltx ? 3.5 : (turbo ? 1.5 : 4)
  const value = Number(raw)
  if (!Number.isFinite(value)) return fallback
  const snapped = Math.round(value * 2) / 2
  return Math.min(10, Math.max(1, snapped))
}

function parseSampler(raw: string | undefined, ltx = false) {
  return SAMPLERS.has(raw || '') ? raw! : (ltx ? 'euler' : 'res_multistep')
}

function parseScheduler(raw: string | undefined) {
  return SCHEDULERS.has(raw || '') ? raw! : 'simple'
}

export default defineEventHandler(async (event) => {
  const form = await readMultipartFormData(event)
  if (!form?.length) {
    throw createError({ statusCode: 400, statusMessage: 'Expected multipart form data' })
  }

  const fields: Record<string, string> = {}
  let image: { filename: string; data: Buffer; type?: string } | null = null
  const referenceImages: Array<{ filename: string; data: Buffer; type?: string } | null> = [null, null, null, null]
  let compactRefSlot = 0
  for (const part of form) {
    if (part.name === 'image' && part.filename && part.data?.length) {
      image = { filename: part.filename, data: part.data, type: part.type }
    } else if (part.name && /^refImage(\d)$/.test(part.name) && part.filename && part.data?.length) {
      const slot = Number(part.name.replace('refImage', ''))
      if (slot >= 0 && slot < 4) {
        referenceImages[slot] = { filename: part.filename, data: part.data, type: part.type }
      }
    } else if (part.name === 'refImage' && part.filename && part.data?.length) {
      if (compactRefSlot < 4) {
        referenceImages[compactRefSlot] = { filename: part.filename, data: part.data, type: part.type }
        compactRefSlot += 1
      }
    } else if (part.name && part.data) {
      fields[part.name] = part.data.toString('utf8')
    }
  }

  const workflow = parseVideoWorkflow(fields.workflow)
  const textToVideo = isTextToVideo(workflow)
  if (!image && !textToVideo) {
    throw createError({ statusCode: 400, statusMessage: 'An initial image is required' })
  }
  const prompt = (fields.prompt || '').trim()
  if (!prompt) {
    throw createError({ statusCode: 400, statusMessage: 'A motion prompt is required' })
  }
  const extensions = parseExtensions(fields.extensions)
  const useIdentityRefs = workflow === 'v2' && !textToVideo && !isLtxWorkflow(workflow) && fields.useIdentityRefs === 'true'
  const { width, height } = resolveOutputSize(
    fields.aspect,
    image?.data || Buffer.alloc(0),
    Number(fields.width),
    Number(fields.height)
  )
  const turbo = fields.turbo !== 'false'
  const steps = Number(fields.steps || defaultVideoSteps(turbo, workflow))
  const seed = fields.seed && fields.seed !== 'random'
    ? Number(fields.seed)
    : Math.floor(Math.random() * 2_147_483_647)
  const durationSeconds = parseDuration(fields.duration || '5')
  const fps = parseFps(fields.fps)
  const cfg = parseCfg(fields.cfg, turbo, isLtxWorkflow(workflow))
  const samplerName = parseSampler(fields.sampler_name, isLtxWorkflow(workflow))
  const scheduler = parseScheduler(fields.scheduler)
  const length = frameLength(durationSeconds, fps)
  const hideThumbnail = fields.hideThumbnail === 'true'
  const hideInput = fields.hideInput === 'true'
  const sound = !isLtxWorkflow(workflow) && fields.sound !== 'false'
  const clipName = (fields.name || '').trim().slice(0, 80)
  const ownerKey = libraryOwnerKey(event)
  const library = publicLibrary(event)
  const folderId = library.folders.some(folder => folder.id === fields.folderId)
    ? fields.folderId
    : library.folders[0]?.id
  if (!folderId) {
    throw createError({ statusCode: 400, statusMessage: 'Create a library folder before generating' })
  }
  const destFolder = library.folders.find(folder => folder.id === folderId)
  const folderLocked = Boolean(destFolder?.protected && !destFolder.unlocked)
  assertFolderExists(event, folderId)
  const still = image
    ? await rememberInputStill({
      ownerKey,
      folderId,
      filename: image.filename,
      data: image.data,
      width,
      height,
      hideInput
    })
    : null
  const referenceStillIds: Array<string | null> = [null, null, null, null]
  if (useIdentityRefs) {
    for (const [index, ref] of referenceImages.entries()) {
      if (!ref?.data?.length || isPipelineFrameFilename(ref.filename)) continue
      const saved = await rememberInputStill({
        ownerKey,
        folderId,
        filename: ref.filename || `identity-ref-${index + 2}.png`,
        data: ref.data,
        width,
        height,
        hideInput
      })
      if (saved?.id) referenceStillIds[index] = saved.id
    }
  }

  const chainTotal = 1 + extensions.length
  const job = createJob()
  job.kind = 'video'
  job.maxStep = steps
  job.hideThumbnail = hideThumbnail
  const familyId = crypto.randomUUID()
  const queueAutoRun = fields.queueAutoRun === 'true'
  job.library = {
    ownerKey,
    folderId,
    hideThumbnail,
    hideInput,
    folderLocked,
    name: clipName,
    prompt,
    aspect: fields.aspect || 'auto',
    width,
    height,
    steps,
    turbo,
    seed,
    cfg,
    fps,
    samplerName,
    scheduler,
    thumb: image && !isPipelineFrameFilename(image.filename) ? image.data : undefined,
    stillId: still?.id,
    stillFilename: still?.filename,
    referenceStillIds,
    duration: durationSeconds,
    sound,
    draftId: (fields.draftId || '').trim() || undefined,
    extensions,
    chainIndex: 0,
    chainStep: 1,
    chainTotal,
    chainLabel: extensions.length ? 'Initial' : undefined,
    familyId,
    workflow,
    useIdentityRefs,
    queueAutoRun: extensions.length ? queueAutoRun : false,
    queueBudget: extensions.length && queueAutoRun ? extensions.length : 0
  }
  if (extensions.length) {
    const queue = await createShotQueue({
      ownerKey,
      name: clipName,
      familyId,
      folderId,
      autoRun: queueAutoRun,
      stillId: still?.id,
      stillFilename: still?.filename,
      hideThumbnail,
      hideInput,
      aspect: fields.aspect || 'auto',
      width,
      height,
      steps,
      turbo,
      cfg,
      fps,
      samplerName,
      scheduler,
      workflow,
      sound,
      useIdentityRefs,
      referenceStillIds,
      initial: { prompt, duration: durationSeconds },
      extensions,
      jobId: job.id
    })
    job.library.queueId = queue.id
    setQueueJob(queue.id, job.id)
  }
  emitChainJob(job, { type: 'status', message: 'Checking ComfyUI...', progress: 1 })

  void runGeneration(job, {
    prompt,
    image,
    width,
    height,
    steps,
    seed,
    turbo,
    length,
    sound,
    cfg,
    fps,
    samplerName,
    scheduler,
    extensions,
    workflow,
    duration: durationSeconds,
    useIdentityRefs,
    referenceImages
  }).catch(async (error) => {
    removeExtendTemp(job.library?.extendTmpDir)
    if (isComfyBusyTimeout(error) && job.library) {
      try {
        const draft = await saveRetryDraft({
          ownerKey: job.library.ownerKey,
          id: job.library.draftId,
          folderId: job.library.folderId,
          stillId: job.library.stillId,
          stillFilename: job.library.stillFilename,
          name: job.library.name,
          prompt: job.library.prompt,
          aspect: job.library.aspect,
          width: job.library.width,
          height: job.library.height,
          steps: job.library.steps,
          turbo: job.library.turbo,
          seed: job.library.seed,
          duration: job.library.duration || durationSeconds,
          hideThumbnail: job.library.hideThumbnail,
          hideInput: job.library.hideInput === true,
          sound: job.library.sound !== false,
          cfg: job.library.cfg,
          fps: job.library.fps,
          samplerName: job.library.samplerName,
          scheduler: job.library.scheduler,
          extensions: job.library.extensions,
          workflow: job.library.workflow
        })
        job.library.draftId = draft.id
        job.status = 'deferred'
        const message = 'ComfyUI was still busy after 3 minutes. Saved this job — click it under Held jobs to restore the form.'
        emitJob(job, { type: 'deferred', draftId: draft.id, message, progress: job.progress })
        if (job.library.queueId) {
          await finishQueueBurst(job.library.ownerKey, job.library.queueId, true, 'Held because ComfyUI was busy').catch(() => null)
          setQueueJob(job.library.queueId, null)
        }
        return
      } catch {
        // fall through to a normal error if the held job could not be saved
      }
    }
    if (job.status === 'error' || job.status === 'cancelled' || job.status === 'deferred') return
    const message = error instanceof Error ? error.message : String(error)
    job.status = 'error'
    job.error = message
    emitJob(job, { type: 'error', error: message, message })
    if (job.library?.queueId) {
      await finishQueueBurst(job.library.ownerKey, job.library.queueId, false, message).catch(() => null)
      setQueueJob(job.library.queueId, null)
    }
  })

  return {
    jobId: job.id,
    clientId: job.clientId,
    seed,
    width,
    height,
    aspect: fields.aspect || 'auto',
    cfg,
    fps,
    sampler_name: samplerName,
    scheduler,
    steps,
    turbo,
    folderId,
    hideThumbnail,
    folderLocked,
    chainTotal,
    queueId: job.library?.queueId,
    queuedRemaining: extensions.length
  }
})