fix(video): keep fal settings enabled without a start frame

- Drop the optional start frame fields from `dependsOn`: an array there
  disables the field until every entry is filled, which greyed out the
  settings for text-only runs. Options and conditions already re-read
  all block values.
- Pass the run's abort signal through to the start frame download.
- The fal tool now offers `visualReference` to agents for models with an
  image-to-video endpoint; update the provider selection test.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Author: myxamediyar

apps/sim/blocks/blocks/video_generator.ts

@@ -47,6 +47,8 @@ const FALAI_AUDIO_DEFAULT_OFF_MODELS = ['kling-o3-pro', 'kling-o3-4k']
 const FAL_MODEL_DEFAULT = { id: 'default', label: 'Model default' } as const
 const endpointModels = falVideoModels.filter((model) => !model.legacyId)
 
+// Never list these optional fields in `dependsOn`: an array there disables the field until every
+// entry has a value. Options and conditions already re-read all block values on any change.
 const START_FRAME_FIELDS = ['visualReferenceInput', 'visualReferenceUpload']
 const HIDDEN = { field: 'provider', value: [] }
 
@@ -78,7 +80,7 @@ function tuningField(id: string, title: string, key: FalSettingKey): SubBlockCon
         ? { field: 'provider', value: 'falai' }
         : HIDDEN
     },
-    dependsOn: ['model', ...START_FRAME_FIELDS],
+    dependsOn: ['model'],
     options: (params) => {
       if (!params) {
         const values = endpointModels.flatMap((model) =>
@@ -179,10 +181,6 @@ function withFalVideoCatalog(subBlocks: SubBlockConfig[]): SubBlockConfig[] {
             const image = getFalVideoModel(values?.model)?.image
             return image && !image[key] && hasStartFrame(values) ? HIDDEN : condition
           },
-          dependsOn: [
-            ...(Array.isArray(subBlock.dependsOn) ? subBlock.dependsOn : []),
-            ...START_FRAME_FIELDS,
-          ],
         },
       ]
     return [subBlock]
@@ -207,7 +205,7 @@ const directVideoFields = (
     title,
     type: key === 'promptOptimizer' ? 'switch' : 'dropdown',
     value: key === 'promptOptimizer' ? undefined : () => FAL_MODEL_DEFAULT.id,
-    dependsOn: ['provider', 'model', ...START_FRAME_FIELDS],
+    dependsOn: ['provider', 'model'],
     condition: (values) => ({
       field: 'provider',
       value: selectedDirectModel(values)?.settings?.[key] ? [String(values?.provider)] : [],

apps/sim/lib/internal/video/client.ts

@@ -117,12 +117,13 @@ const MAX_REFERENCE_IMAGE_BYTES = Math.floor(((5 * 1024 * 1024 - 512) * 3) / 4)
 
 async function readReferenceImageDataUri(
   file: UserFile,
-  requestId: string,
+  { requestId, signal }: VideoGenerationContext,
   label: string
 ): Promise<string> {
   assertKnownSizeWithinLimit(file.size, MAX_REFERENCE_IMAGE_BYTES, label)
   const buffer = await downloadFileFromStorage(file, requestId, logger, {
     maxBytes: MAX_REFERENCE_IMAGE_BYTES,
+    signal,
   })
   assertKnownSizeWithinLimit(buffer.length, MAX_REFERENCE_IMAGE_BYTES, label)
   return `data:${file.type};base64,${buffer.toString('base64')}`
@@ -134,7 +135,7 @@ async function generateWithDirectProvider(
 ): Promise<VideoGenerationResult> {
   const prepared = buildDirectVideoRequest(input)
   const { protocol } = prepared.model
-  const { signal, requestId } = context
+  const { signal } = context
   const headers: Record<string, string> = { 'Content-Type': 'application/json' }
   if (protocol === 'gemini') {
     headers['x-goog-api-key'] = input.apiKey
@@ -145,7 +146,7 @@ async function generateWithDirectProvider(
   if (input.visualReference) {
     prepared.body.promptImage = await readReferenceImageDataUri(
       input.visualReference,
-      requestId,
+      context,
       'Runway reference image'
     )
   }
@@ -378,7 +379,7 @@ async function generateWithFalAI(
   if (input.visualReference && imageField) {
     body[imageField] = await readReferenceImageDataUri(
       input.visualReference,
-      requestId,
+      context,
       'Fal.ai start frame image'
     )
     signal?.throwIfAborted()

apps/sim/providers/utils.test.ts

@@ -2245,7 +2245,7 @@ describe('transformBlockTool configured selectors', () => {
     )
 
     expect(result?.id).toBe('video_falai')
-    expect(result?.parameters?.properties).not.toHaveProperty('visualReference')
+    expect(result?.parameters?.properties).toHaveProperty('visualReference')
     const prepared = prepareToolExecution(
       result!,
       { prompt: 'A paper boat', provider: 'runway', model: 'gen-4-turbo' },