fix(media): preflight all writable outputs

Author: j15z

apps/sim/lib/copilot/generated/tool-catalog-v1.ts

@@ -1630,7 +1630,7 @@ export const GenerateAudio: ToolCatalogEntry = {
       },
       voice: { type: 'string', description: 'Optional voice name or id for speech.' },
     },
-    required: ['prompt'],
+    required: ['prompt', 'outputs'],
   },
   requiredPermission: 'write',
   capabilities: ['file_input', 'file_output', 'generated_media'],
@@ -1713,7 +1713,7 @@ export const GenerateImage: ToolCatalogEntry = {
           'Detailed text description of the image to generate, or editing instructions when editing the image(s) passed in `inputs.files`.',
       },
     },
-    required: ['prompt'],
+    required: ['prompt', 'outputs'],
   },
   requiredPermission: 'write',
   capabilities: ['file_input', 'file_output', 'generated_media'],
@@ -1836,7 +1836,7 @@ export const GenerateVideo: ToolCatalogEntry = {
         enum: ['720p', '1080p', '4k'],
       },
     },
-    required: ['prompt'],
+    required: ['prompt', 'outputs'],
   },
   requiredPermission: 'write',
   capabilities: ['file_input', 'file_output', 'generated_media'],

apps/sim/lib/copilot/generated/tool-schemas-v1.ts

@@ -1441,7 +1441,7 @@ export const TOOL_RUNTIME_SCHEMAS: Record<string, ToolRuntimeSchemaEntry> = {
           description: 'Optional voice name or id for speech.',
         },
       },
-      required: ['prompt'],
+      required: ['prompt', 'outputs'],
     },
     resultSchema: undefined,
   },
@@ -1518,7 +1518,7 @@ export const TOOL_RUNTIME_SCHEMAS: Record<string, ToolRuntimeSchemaEntry> = {
             'Detailed text description of the image to generate, or editing instructions when editing the image(s) passed in `inputs.files`.',
         },
       },
-      required: ['prompt'],
+      required: ['prompt', 'outputs'],
     },
     resultSchema: undefined,
   },
@@ -1635,7 +1635,7 @@ export const TOOL_RUNTIME_SCHEMAS: Record<string, ToolRuntimeSchemaEntry> = {
           enum: ['720p', '1080p', '4k'],
         },
       },
-      required: ['prompt'],
+      required: ['prompt', 'outputs'],
     },
     resultSchema: undefined,
   },

apps/sim/lib/copilot/tools/server/image/generate-image.test.ts

@@ -172,7 +172,11 @@ describe('generateImageServerTool', () => {
 
   it('rejects explicitly empty input and output containers before generation', async () => {
     const emptyInputs = await generateImageServerTool.execute(
-      { prompt: 'Edit the supplied image', inputs: {} },
+      {
+        prompt: 'Edit the supplied image',
+        inputs: {},
+        outputs: { files: [{ path: 'files/result.png' }] },
+      },
       CONTEXT
     )
     const emptyOutputs = await generateImageServerTool.execute(

apps/sim/lib/copilot/tools/server/image/generate-image.ts

@@ -163,10 +163,9 @@ export const generateImageServerTool: BaseServerTool<GenerateImageArgs, Generate
         }
       }
 
-      const ext = mimeType.includes('jpeg') || mimeType.includes('jpg') ? '.jpg' : '.png'
-      const resolvedOutputPath = outputFile?.path || `files/generated-image${ext}`
+      const resolvedOutputPath = outputFile.path
       const imageBuffer = Buffer.from(imageBase64, 'base64')
-      const mode = outputFile?.mode ?? 'create'
+      const mode = outputFile.mode
 
       assertServerToolNotAborted(context)
       const written = await writeWorkspaceFileByPath({
@@ -175,7 +174,7 @@ export const generateImageServerTool: BaseServerTool<GenerateImageArgs, Generate
         target: {
           path: resolvedOutputPath,
           mode,
-          mimeType: outputFile?.mimeType,
+          mimeType: outputFile.mimeType,
         },
         buffer: imageBuffer,
         inferredMimeType: mimeType,

apps/sim/lib/copilot/tools/server/media/ffmpeg.ts

@@ -82,11 +82,14 @@ export const ffmpegServerTool: BaseServerTool<FfmpegArgs, FfmpegResult> = {
     }
 
     try {
-      const outputFile = await prepareMediaOutput({
-        output: params.outputs,
-        workspaceId,
-        userId: context.userId,
-      })
+      const outputFile =
+        params.operation === 'probe'
+          ? undefined
+          : await prepareMediaOutput({
+              output: params.outputs,
+              workspaceId,
+              userId: context.userId,
+            })
 
       const mediaFiles: MediaFile[] = []
       for (const filePath of inputPaths) {
@@ -127,18 +130,21 @@ export const ffmpegServerTool: BaseServerTool<FfmpegArgs, FfmpegResult> = {
         }
       }
 
-      if (!result.buffer || !result.ext) {
+      if (!result.buffer) {
         return { success: false, message: `ffmpeg ${params.operation} produced no output` }
       }
+      if (!outputFile) {
+        return { success: false, message: `ffmpeg ${params.operation} requires an output file` }
+      }
 
-      const outputPath = outputFile?.path || `files/ffmpeg-${params.operation}.${result.ext}`
-      const mode = outputFile?.mode ?? 'create'
+      const outputPath = outputFile.path
+      const mode = outputFile.mode
 
       assertServerToolNotAborted(context)
       const written = await writeWorkspaceFileByPath({
         workspaceId,
         userId: context.userId,
-        target: { path: outputPath, mode, mimeType: outputFile?.mimeType },
+        target: { path: outputPath, mode, mimeType: outputFile.mimeType },
         buffer: result.buffer,
         inferredMimeType: result.contentType || 'application/octet-stream',
       })

apps/sim/lib/copilot/tools/server/media/file-paths.test.ts

@@ -23,6 +23,7 @@ vi.mock('@/lib/copilot/vfs/resource-writer', () => ({
 
 import {
   getSingleMediaFileDeclaration,
+  prepareMediaOutput,
   resolveMediaInputFile,
   validateMediaOutputFile,
 } from '@/lib/copilot/tools/server/media/file-paths'
@@ -50,6 +51,20 @@ describe('media file paths', () => {
     expect(mocks.resolveWorkspaceFileReference).not.toHaveBeenCalled()
   })
 
+  it('resolves uploads/ inputs with the read-compatible /content suffix', async () => {
+    mocks.resolveChatUpload.mockResolvedValue(FILE_RECORD)
+
+    await expect(
+      resolveMediaInputFile({
+        workspaceId: 'workspace-1',
+        chatId: 'chat-1',
+        path: 'uploads/My%20Portrait.png/content',
+      })
+    ).resolves.toBe(FILE_RECORD)
+
+    expect(mocks.resolveChatUpload).toHaveBeenCalledWith('My%20Portrait.png', 'chat-1')
+  })
+
   it('rejects unresolved inputs instead of dropping them', async () => {
     mocks.resolveWorkspaceFileReference.mockResolvedValue(null)
 
@@ -155,4 +170,15 @@ describe('media file paths', () => {
       'Output requires exactly one file; received 0'
     )
   })
+
+  it('requires an explicit output before provider work can begin', async () => {
+    await expect(
+      prepareMediaOutput({
+        workspaceId: 'workspace-1',
+        userId: 'user-1',
+      })
+    ).rejects.toThrow('Output requires exactly one file; received 0')
+
+    expect(mocks.validateWorkspaceFileWriteTarget).not.toHaveBeenCalled()
+  })
 })

apps/sim/lib/copilot/tools/server/media/file-paths.ts

@@ -49,8 +49,8 @@ export async function resolveMediaInputFile(args: {
   let file: WorkspaceFileRecord | null
 
   if (path.startsWith(UPLOADS_PREFIX)) {
-    const filename = path.slice(UPLOADS_PREFIX.length)
-    if (!filename || filename.includes('/')) {
+    const filename = path.slice(UPLOADS_PREFIX.length).split('/')[0]
+    if (!filename) {
       throw new Error(`Upload input path must identify a file: ${args.path}`)
     }
     if (!args.chatId) {
@@ -68,7 +68,7 @@ export async function resolveMediaInputFile(args: {
 }
 
 /**
- * Preflight a user-supplied media output through the canonical workspace writer policy.
+ * Preflight a media output through the canonical workspace writer policy.
  */
 export async function validateMediaOutputFile(args: {
   workspaceId: string
@@ -93,10 +93,8 @@ export async function prepareMediaOutput<T extends MediaOutputDeclaration>(args:
   output?: { files?: T[] }
   workspaceId: string
   userId: string
-}): Promise<(T & { mode: WorkspaceFileWriteMode }) | undefined> {
-  if (!args.output) return undefined
-
-  const file = getSingleMediaFileDeclaration(args.output.files, 'Output')
+}): Promise<T & { mode: WorkspaceFileWriteMode }> {
+  const file = getSingleMediaFileDeclaration(args.output?.files, 'Output')
   const output = { ...file, mode: file.mode ?? 'create' }
   await validateMediaOutputFile({
     workspaceId: args.workspaceId,

apps/sim/lib/copilot/tools/server/media/generate-audio.ts

@@ -46,16 +46,6 @@ interface GenerateAudioResult {
   _serviceCost?: { service: string; cost: number }
 }
 
-function audioExtFromContentType(contentType: string): string {
-  if (contentType.includes('wav')) return 'wav'
-  if (contentType.includes('mp4') || contentType.includes('m4a')) return 'm4a'
-  if (contentType.includes('ogg')) return 'ogg'
-  if (contentType.includes('flac')) return 'flac'
-  if (contentType.includes('aac')) return 'aac'
-  if (contentType.includes('opus')) return 'opus'
-  return 'mp3'
-}
-
 export const generateAudioServerTool: BaseServerTool<GenerateAudioArgs, GenerateAudioResult> = {
   name: GenerateAudio.id,
 
@@ -123,15 +113,14 @@ export const generateAudioServerTool: BaseServerTool<GenerateAudioArgs, Generate
         voiceSampleDataUri,
       })
 
-      const ext = audioExtFromContentType(result.contentType)
-      const outputPath = outputFile?.path || `files/generated-audio.${ext}`
-      const mode = outputFile?.mode ?? 'create'
+      const outputPath = outputFile.path
+      const mode = outputFile.mode
 
       assertServerToolNotAborted(context)
       const written = await writeWorkspaceFileByPath({
         workspaceId,
         userId: context.userId,
-        target: { path: outputPath, mode, mimeType: outputFile?.mimeType },
+        target: { path: outputPath, mode, mimeType: outputFile.mimeType },
         buffer: result.buffer,
         inferredMimeType: result.contentType,
       })

apps/sim/lib/copilot/tools/server/media/generate-video.ts

@@ -66,7 +66,7 @@ export const generateVideoServerTool: BaseServerTool<GenerateVideoArgs, Generate
         workspaceId,
         userId: context.userId,
       })
-      const outputPath = outputFile?.path ?? 'files/generated-video.mp4'
+      const outputPath = outputFile.path
 
       let imageDataUri: string | undefined
       const inputFile = params.inputs
@@ -101,13 +101,13 @@ export const generateVideoServerTool: BaseServerTool<GenerateVideoArgs, Generate
         imageDataUri,
       })
 
-      const mode = outputFile?.mode ?? 'create'
+      const mode = outputFile.mode
 
       assertServerToolNotAborted(context)
       const written = await writeWorkspaceFileByPath({
         workspaceId,
         userId: context.userId,
-        target: { path: outputPath, mode, mimeType: outputFile?.mimeType },
+        target: { path: outputPath, mode, mimeType: outputFile.mimeType },
         buffer: result.buffer,
         inferredMimeType: result.contentType,
       })

apps/sim/lib/copilot/tools/server/media/media-tools-validation.test.ts

@@ -94,7 +94,11 @@ describe('media tool path validation', () => {
     mocks.resolveWorkspaceFileReference.mockResolvedValue(null)
 
     const result = await generateAudioServerTool.execute(
-      { prompt: 'Clone this voice', inputs: { files: [{ path: '' }] } },
+      {
+        prompt: 'Clone this voice',
+        inputs: { files: [{ path: '' }] },
+        outputs: { files: [{ path: 'files/voice.mp3' }] },
+      },
       CONTEXT
     )
 
@@ -140,6 +144,34 @@ describe('media tool path validation', () => {
     expect(mocks.runFfmpegOperation).not.toHaveBeenCalled()
   })
 
+  it('ignores unused output declarations for ffmpeg probes', async () => {
+    mocks.resolveWorkspaceFileReference.mockResolvedValue(UPLOAD_RECORD)
+    mocks.runFfmpegOperation.mockResolvedValue({ probe: { duration: 3 } })
+
+    const result = await ffmpegServerTool.execute(
+      {
+        operation: 'probe',
+        inputs: { files: [{ path: 'files/input.mov' }] },
+        outputs: { files: [{ path: 'uploads/unused.json' }] },
+      },
+      CONTEXT
+    )
+
+    expect(result.success).toBe(true)
+    expect(mocks.validateWorkspaceFileWriteTarget).not.toHaveBeenCalled()
+    expect(mocks.writeWorkspaceFileByPath).not.toHaveBeenCalled()
+  })
+
+  it('rejects missing generated-media outputs before calling the provider', async () => {
+    const result = await generateVideoServerTool.execute(
+      { prompt: 'Create a launch video' },
+      CONTEXT
+    )
+
+    expect(result.success).toBe(false)
+    expect(mocks.generateFalVideo).not.toHaveBeenCalled()
+  })
+
   it('passes uploads/ media files to ffmpeg', async () => {
     mocks.resolveChatUpload.mockResolvedValue(UPLOAD_RECORD)
 
@@ -166,6 +198,7 @@ describe('media tool path validation', () => {
       {
         prompt: 'Animate these portraits',
         inputs: { files: [{ path: 'files/one.png' }, { path: 'files/two.png' }] },
+        outputs: { files: [{ path: 'files/output.mp4' }] },
       },
       CONTEXT
     )