chore(release): merge staging into the v0.9.43 release

Author: waleedlatif1

apps/sim/blocks/blocks/image_generator.ts

@@ -459,7 +459,10 @@ export const ImageGeneratorV2Block: BlockConfig = {
   },
   inputs: {
     provider: { type: 'string', description: 'Image generation provider' },
-    prompt: { type: 'string', description: 'Image description prompt' },
+    prompt: {
+      type: 'string',
+      description: 'Image description, up to 4,000 characters. Some models have a lower limit.',
+    },
     model: { type: 'string', description: 'Image generation model' },
     size: { type: 'string', description: 'Image size' },
     aspectRatio: { type: 'string', description: 'Image aspect ratio' },

apps/sim/blocks/blocks/video_generator.ts

@@ -1834,7 +1834,7 @@ export const VideoGeneratorV2Block: BlockConfig = {
         const model = params.provider === 'falai' ? getFalVideoModel(params.model) : undefined
         const useFalFields = model && !model.legacyId
         const setting = (legacy: string, catalog: string) => {
-          const value = params[useFalFields ? catalog : legacy]
+          const value = useFalFields ? (params[catalog] ?? params[legacy]) : params[legacy]
           return useFalFields && value === FAL_MODEL_DEFAULT.id ? undefined : value
         }
         const duration = setting('duration', 'falDuration')
@@ -1956,7 +1956,7 @@ export const VideoGeneratorV3Block: BlockConfig = {
           promptOptimizer: parseOptionalBooleanInput(
             setting('promptOptimizer', model?.settings?.promptOptimizer)
           ),
-          endpoint: model?.protocol === 'minimax-v1' ? params.endpoint : undefined,
+          endpoint: params.provider === 'minimax' ? params.endpoint : undefined,
           cameraControl:
             model?.protocol === 'luma-legacy' && params.cameraControl
               ? typeof params.cameraControl === 'string'

apps/sim/lib/internal/image/request.ts

@@ -1,6 +1,9 @@
 import type { ImageGenerationInput, ImageProvider } from '@/lib/internal/image/schema'
 import { getImageModel, imageModelDefaults, imageOptionFields } from '@/lib/media/image-catalog'
 
+/** Sim limits image prompts independently of larger provider model allowances. */
+const MAX_IMAGE_PROMPT_CHARACTERS = 4000
+
 /** Builds only supported settings for the selected image model before a paid request is sent. */
 export function buildImageGenerationRequest(input: ImageGenerationInput) {
   const model = getImageModel(
@@ -12,7 +15,10 @@ export function buildImageGenerationRequest(input: ImageGenerationInput) {
     throw new Error('Image generation returns one image per tool execution; numImages must be 1')
   }
   const minLength = Math.max(3, model.minPromptLength ?? 3)
-  const maxLength = Math.min(4000, model.maxPromptLength ?? 4000)
+  const maxLength = Math.min(
+    MAX_IMAGE_PROMPT_CHARACTERS,
+    model.maxPromptLength ?? MAX_IMAGE_PROMPT_CHARACTERS
+  )
   if (input.prompt.length < minLength || input.prompt.length > maxLength) {
     throw new Error(
       `Prompt must be between ${minLength} and ${maxLength} characters for ${model.label}`

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

@@ -28,9 +28,9 @@ describe('Video provider client', () => {
       isValid: true,
       resolvedIP: '203.0.113.1',
     })
-    inputValidationMockFns.mockSecureFetchWithPinnedIP.mockResolvedValue(
-      new Response(Buffer.from('video'))
-    )
+    inputValidationMockFns.mockSecureFetchWithPinnedIP
+      .mockReset()
+      .mockResolvedValue(new Response(Buffer.from('video')))
   })
 
   afterEach(() => {
@@ -204,23 +204,6 @@ describe('Video provider client', () => {
         },
       },
     },
-    {
-      provider: 'veo' as const,
-      model: 'gemini-omni-1.1-flash',
-      submit: '/v1beta/interactions',
-      poll: '/v1beta/interactions/job',
-      payload: { model: 'gemini-omni-1.1-flash', input: 'A cinematic sunrise' },
-      created: { id: 'job', status: 'in_progress' },
-      completed: {
-        status: 'completed',
-        steps: [
-          {
-            type: 'model_output',
-            content: [{ type: 'video', mime_type: 'video/mp4', data: 'dmlkZW8=' }],
-          },
-        ],
-      },
-    },
   ])('executes $model through its native submit, poll and result protocol', async (testCase) => {
     let submissions = 0
     vi.stubGlobal('fetch', async (url: string | URL, init?: RequestInit) => {
@@ -251,6 +234,156 @@ describe('Video provider client', () => {
     expect(submissions).toBe(1)
   })
 
+  describe('Gemini URI delivery', () => {
+    const input = { provider: 'veo' as const, apiKey: 'key', prompt: 'A cinematic sunrise' }
+    const videoUri =
+      'https://generativelanguage.googleapis.com/v1beta/files/video-file:download?alt=media'
+    const completed = (uri = videoUri) => ({
+      id: 'interaction-1',
+      status: 'completed',
+      steps: [{ type: 'model_output', content: [{ type: 'video', uri }] }],
+    })
+
+    it('downloads a normal-sized video after Files becomes ACTIVE without reading inline polling data', async () => {
+      const video = Buffer.alloc(2 * 1024 * 1024)
+      let filePolls = 0
+      let submissions = 0
+      vi.stubGlobal('fetch', async (url: string | URL, init?: RequestInit) => {
+        const path = new URL(url).pathname
+        if (path === '/v1beta/interactions' && init?.method === 'POST') {
+          submissions++
+          const body = JSON.parse(String(init.body))
+          return jsonResponse(
+            body.background ? { id: 'interaction-1', status: 'in_progress' } : completed()
+          )
+        }
+        if (path === '/v1beta/interactions/interaction-1')
+          return jsonResponse({
+            ...completed(),
+            steps: [
+              {
+                type: 'model_output',
+                content: [{ type: 'video', data: video.toString('base64') }],
+              },
+            ],
+          })
+        if (path === '/v1beta/files/video-file') {
+          filePolls++
+          expect(new Headers(init?.headers).get('x-goog-api-key')).toBe('key')
+          return jsonResponse({ state: filePolls === 1 ? 'PROCESSING' : 'ACTIVE' })
+        }
+        return new Response(null, { status: 404 })
+      })
+      inputValidationMockFns.mockSecureFetchWithPinnedIP.mockImplementationOnce(
+        async (_url, _ip, options) => {
+          expect(new Headers(options.headers).get('x-goog-api-key')).toBe('key')
+          return filePolls >= 2
+            ? new Response(video)
+            : new Response('File is not ACTIVE', { status: 409 })
+        }
+      )
+      const result = generateVideo(input, { requestId: 'google-uri' })
+      const completion = result.then((output) => {
+        expect(output.buffer.equals(video)).toBe(true)
+        expect(output.jobId).toBe('interaction-1')
+      })
+      await Promise.all([
+        completion,
+        vi.advanceTimersByTimeAsync(0).then(() => vi.advanceTimersByTimeAsync(5000)),
+      ])
+      expect(submissions).toBe(1)
+    })
+
+    it.each([
+      { state: 'FAILED', error: /Google video processing failed/ },
+      { state: 'PROCESSING', error: /Google video processing timed out/ },
+    ])('does not download a file that remains $state', async ({ state, error }) => {
+      vi.stubGlobal(
+        'fetch',
+        vi
+          .fn()
+          .mockResolvedValueOnce(jsonResponse(completed()))
+          .mockImplementation(() => Promise.resolve(jsonResponse({ state })))
+      )
+      const rejection = expect(
+        generateVideo(input, { requestId: 'google-processing' })
+      ).rejects.toThrow(error)
+      await Promise.all([rejection, vi.advanceTimersByTimeAsync(5000)])
+      expect(inputValidationMockFns.mockSecureFetchWithPinnedIP).not.toHaveBeenCalled()
+    })
+
+    it('stops during Files processing without downloading or polling again', async () => {
+      const controller = new AbortController()
+      const fetchMock = vi
+        .fn()
+        .mockResolvedValueOnce(jsonResponse(completed()))
+        .mockResolvedValueOnce(jsonResponse({ state: 'PROCESSING' }))
+      vi.stubGlobal('fetch', fetchMock)
+      const result = generateVideo(input, {
+        requestId: 'google-stopped',
+        signal: controller.signal,
+      })
+      const rejection = expect(result).rejects.toMatchObject({ name: 'AbortError' })
+      await Promise.all([
+        rejection,
+        vi.advanceTimersByTimeAsync(0).then(() => {
+          controller.abort(new DOMException('Stopped', 'AbortError'))
+        }),
+      ])
+      expect(fetchMock).toHaveBeenCalledTimes(2)
+      expect(inputValidationMockFns.mockSecureFetchWithPinnedIP).not.toHaveBeenCalled()
+    })
+
+    it.each([
+      'https://foreign.example/v1beta/files/video-file:download?alt=media',
+      'https://generativelanguage.googleapis.com/v1beta/models/model:download?alt=media',
+      'https://generativelanguage.googleapis.com/v1beta/files/nested%2Ffile:download?alt=media',
+    ])('rejects a non-Files download URI before sending credentials: %s', async (uri) => {
+      const fetchMock = vi.fn().mockResolvedValueOnce(jsonResponse(completed(uri)))
+      vi.stubGlobal('fetch', fetchMock)
+      await expect(generateVideo(input, { requestId: 'google-invalid-uri' })).rejects.toThrow(
+        /Invalid Google video file URI/
+      )
+      expect(fetchMock).toHaveBeenCalledTimes(1)
+      expect(inputValidationMockFns.mockSecureFetchWithPinnedIP).not.toHaveBeenCalled()
+    })
+
+    it('keeps Files metadata within the ordinary JSON response limit', async () => {
+      vi.stubGlobal(
+        'fetch',
+        vi
+          .fn()
+          .mockResolvedValueOnce(jsonResponse(completed()))
+          .mockResolvedValueOnce(
+            jsonResponse(
+              { state: 'ACTIVE' },
+              { headers: { 'Content-Length': String(2 * 1024 * 1024 + 1) } }
+            )
+          )
+      )
+      await expect(
+        generateVideo(input, { requestId: 'google-metadata-limit' })
+      ).rejects.toMatchObject({
+        name: 'PayloadSizeLimitError',
+      })
+      expect(inputValidationMockFns.mockSecureFetchWithPinnedIP).not.toHaveBeenCalled()
+    })
+
+    it('does not poll a full interaction when synchronous generation does not complete', async () => {
+      const fetchMock = vi
+        .fn()
+        .mockImplementation(() =>
+          Promise.resolve(jsonResponse({ id: 'interaction-1', status: 'in_progress' }))
+        )
+      vi.stubGlobal('fetch', fetchMock)
+      const rejection = expect(
+        generateVideo(input, { requestId: 'google-incomplete' })
+      ).rejects.toThrow(/Google synchronous video generation did not complete/)
+      await Promise.all([rejection, vi.advanceTimersByTimeAsync(5000)])
+      expect(fetchMock).toHaveBeenCalledTimes(1)
+    })
+  })
+
   it('reports a failed MiniMax job immediately instead of polling until timeout', async () => {
     const fetchMock = vi
       .fn()

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

@@ -227,10 +227,29 @@ async function generateWithDirectProvider(
         }
       }
       if (!buffer && videoUrl) {
-        if (
-          protocol === 'gemini' &&
-          new URL(videoUrl).origin === 'https://generativelanguage.googleapis.com'
-        ) {
+        if (protocol === 'gemini') {
+          const fileUrl = new URL(videoUrl)
+          const fileId = fileUrl.pathname.match(/^\/v1beta\/files\/([a-zA-Z0-9_-]+):download$/)?.[1]
+          if (
+            fileUrl.origin !== 'https://generativelanguage.googleapis.com' ||
+            fileUrl.username ||
+            fileUrl.password ||
+            fileUrl.hash ||
+            !fileId
+          )
+            throw new Error('Invalid Google video file URI')
+          const metadataUrl = `https://generativelanguage.googleapis.com/v1beta/files/${fileId}`
+          for (let fileAttempt = 0; ; fileAttempt++) {
+            const file = await readProviderResponse(
+              await fetch(metadataUrl, { headers, redirect: 'error', signal })
+            )
+            const state = getStringProperty(file, 'state')
+            if (state === 'ACTIVE') break
+            if (state === 'FAILED') throw new Error('Google video processing failed')
+            if (fileAttempt === maxAttempts) throw new Error('Google video processing timed out')
+            await waitForProvider(context)
+          }
+          signal?.throwIfAborted()
           const validation = await validateUrlWithDNS(videoUrl, 'videoUrl', 'contentFetch', {
             signal,
           })
@@ -297,6 +316,8 @@ async function generateWithDirectProvider(
         `${input.provider} generation failed: ${getStringProperty(task, 'failure_reason') || getStringProperty(task, 'failure') || getStringProperty(task, 'error') || 'Request failed'}`
       )
     }
+    if (protocol === 'gemini')
+      throw new Error('Google synchronous video generation did not complete')
     if (attempt === maxAttempts) break
     await waitForProvider(context)
     data = await readProviderResponse(await fetch(pollUrl, { headers, redirect: 'error', signal }))

apps/sim/lib/internal/video/direct-request.ts

@@ -15,6 +15,10 @@ export function buildDirectVideoRequest(input: VideoGenerationInput) {
     throw new Error(
       `${model.label} ${input.visualReference ? 'does not accept a reference image' : 'requires a reference image'}`
     )
+  if (model.protocol === 'minimax-v2' && input.endpoint)
+    throw new Error(
+      `${model.label} does not support quality endpoints; select a resolution instead`
+    )
   const maxLength = spec.maxPromptLength ?? 2000
   if (input.prompt.length < 3 || input.prompt.length > maxLength)
     throw new Error(`Prompt must be between 3 and ${maxLength} characters for ${model.label}`)
@@ -95,7 +99,9 @@ export function buildDirectVideoRequest(input: VideoGenerationInput) {
       body = {
         model: model.id,
         input: input.prompt,
-        background: true,
+        background: false,
+        store: false,
+        stream: false,
         response_format: {
           type: 'video',
           delivery: 'uri',

apps/sim/tools/generated/tool-metadata.ts

@@ -35,7 +35,8 @@ export const imageGenerateTool: ImageGenerationToolConfig = {
       type: 'string',
       required: true,
       visibility: 'user-or-llm',
-      description: 'Text prompt describing the image to generate',
+      description:
+        'Text prompt describing the image to generate, up to 4,000 characters. Some models have a lower limit.',
     },
     size: {
       type: 'string',