v1

apps/sim/app/workspace/[workspaceId]/w/[workflowId]/components/panel/components/editor/components/sub-block/components/messages-input/messages-input.tsx

@@ -9,6 +9,8 @@ import {
 } from 'react'
 import { isEqual } from 'lodash'
 import { ArrowLeftRight, ChevronDown, ChevronsUpDown, ChevronUp, Plus } from 'lucide-react'
+import { useParams } from 'next/navigation'
+import { createLogger } from '@sim/logger'
 import {
   Button,
   Popover,
@@ -22,15 +24,28 @@ import { cn } from '@/lib/core/utils/cn'
 import { EnvVarDropdown } from '@/app/workspace/[workspaceId]/w/[workflowId]/components/panel/components/editor/components/sub-block/components/env-var-dropdown'
 import { FileUpload } from '@/app/workspace/[workspaceId]/w/[workflowId]/components/panel/components/editor/components/sub-block/components/file-upload/file-upload'
 import { formatDisplayText } from '@/app/workspace/[workspaceId]/w/[workflowId]/components/panel/components/editor/components/sub-block/components/formatted-text'
+import { ShortInput } from '@/app/workspace/[workspaceId]/w/[workflowId]/components/panel/components/editor/components/sub-block/components/short-input/short-input'
 import { TagDropdown } from '@/app/workspace/[workspaceId]/w/[workflowId]/components/panel/components/editor/components/sub-block/components/tag-dropdown/tag-dropdown'
 import { useSubBlockInput } from '@/app/workspace/[workspaceId]/w/[workflowId]/components/panel/components/editor/components/sub-block/hooks/use-sub-block-input'
 import { useSubBlockValue } from '@/app/workspace/[workspaceId]/w/[workflowId]/components/panel/components/editor/components/sub-block/hooks/use-sub-block-value'
+import { useWorkflowRegistry } from '@/stores/workflows/registry/store'
+import { useSubBlockStore } from '@/stores/workflows/subblock/store'
 import type { WandControlHandlers } from '@/app/workspace/[workspaceId]/w/[workflowId]/components/panel/components/editor/components/sub-block/sub-block'
 import { useAccessibleReferencePrefixes } from '@/app/workspace/[workspaceId]/w/[workflowId]/hooks/use-accessible-reference-prefixes'
 import { useWand } from '@/app/workspace/[workspaceId]/w/[workflowId]/hooks/use-wand'
 import type { SubBlockConfig } from '@/blocks/types'
 
+const logger = createLogger('MessagesInput')
+
 const MIN_TEXTAREA_HEIGHT_PX = 80
+
+/** Workspace file record from API */
+interface WorkspaceFile {
+  id: string
+  name: string
+  path: string
+  type: string
+}
 const MAX_TEXTAREA_HEIGHT_PX = 320
 
 /** Pattern to match complete message objects in JSON */
@@ -49,14 +64,20 @@ const ROLE_BEFORE_CONTENT_PATTERN = /"role"\s*:\s*"(system|user|assistant|media)
 const unescapeContent = (str: string): string =>
   str.replace(/\\n/g, '\n').replace(/\\"/g, '"').replace(/\\\\/g, '\\')
 
-
 /**
  * Media content for multimodal messages
  */
 interface MediaContent {
+  /** Source type: how the data was provided */
+  sourceType: 'url' | 'base64' | 'file'
+  /** The URL or base64 data */
   data: string
+  /** MIME type (e.g., 'image/png', 'application/pdf', 'audio/mp3') */
   mimeType?: string
+  /** Optional filename for file uploads */
   fileName?: string
+  /** Optional workspace file ID (used by wand to select existing files) */
+  fileId?: string
 }
 
 /**
@@ -93,14 +114,109 @@ export function MessagesInput({
   disabled = false,
   wandControlRef,
 }: MessagesInputProps) {
+  const params = useParams()
+  const workspaceId = params?.workspaceId as string
   const [messages, setMessages] = useSubBlockValue<Message[]>(blockId, subBlockId, false)
   const [localMessages, setLocalMessages] = useState<Message[]>([{ role: 'user', content: '' }])
   const accessiblePrefixes = useAccessibleReferencePrefixes(blockId)
   const [openPopoverIndex, setOpenPopoverIndex] = useState<number | null>(null)
+  const { activeWorkflowId } = useWorkflowRegistry()
 
   // Local media mode state - basic = FileUpload, advanced = URL/base64 textarea
   const [mediaMode, setMediaMode] = useState<'basic' | 'advanced'>('basic')
 
+  // Workspace files for wand context
+  const [workspaceFiles, setWorkspaceFiles] = useState<WorkspaceFile[]>([])
+
+  // Fetch workspace files for wand context
+  const loadWorkspaceFiles = useCallback(async () => {
+    if (!workspaceId || isPreview) return
+
+    try {
+      const response = await fetch(`/api/workspaces/${workspaceId}/files`)
+      const data = await response.json()
+      if (data.success) {
+        setWorkspaceFiles(data.files || [])
+      }
+    } catch (error) {
+      logger.error('Error loading workspace files:', error)
+    }
+  }, [workspaceId, isPreview])
+
+  // Load workspace files on mount
+  useEffect(() => {
+    void loadWorkspaceFiles()
+  }, [loadWorkspaceFiles])
+
+  // Build sources string for wand - available workspace files
+  const sourcesInfo = useMemo(() => {
+    if (workspaceFiles.length === 0) {
+      return 'No workspace files available. The user can upload files manually after generation.'
+    }
+
+    const filesList = workspaceFiles
+      .filter((f) => f.type.startsWith('image/') || f.type.startsWith('audio/') || f.type.startsWith('video/') || f.type === 'application/pdf')
+      .map((f) => `  - id: "${f.id}", name: "${f.name}", type: "${f.type}"`)
+      .join('\n')
+
+    if (!filesList) {
+      return 'No media files in workspace. The user can upload files manually after generation.'
+    }
+
+    return `AVAILABLE WORKSPACE FILES (optional - you don't have to select one):\n${filesList}\n\nTo use a file, include "fileId": "<id>" in the media object. If not selecting a file, omit the fileId field.`
+  }, [workspaceFiles])
+
+
+  // Effect to sync FileUpload values to message media objects
+  useEffect(() => {
+    if (!activeWorkflowId || isPreview) return
+
+    // Get all subblock values for this workflow
+    const workflowValues = useSubBlockStore.getState().workflowValues[activeWorkflowId]
+    if (!workflowValues?.[blockId]) return
+
+    let hasChanges = false
+    const updatedMessages = localMessages.map((msg, index) => {
+      if (msg.role !== 'media') return msg
+
+      // Check if there's a FileUpload value for this media message
+      const fileUploadKey = `${subBlockId}-media-${index}`
+      const fileValue = workflowValues[blockId][fileUploadKey]
+
+      if (fileValue && typeof fileValue === 'object' && 'path' in fileValue) {
+        const uploadedFile = fileValue as { name: string; path: string; type: string; size: number }
+        const newMedia: MediaContent = {
+          sourceType: 'file',
+          data: uploadedFile.path,
+          mimeType: uploadedFile.type,
+          fileName: uploadedFile.name,
+        }
+
+        // Only update if different
+        if (
+          msg.media?.data !== newMedia.data ||
+          msg.media?.sourceType !== newMedia.sourceType ||
+          msg.media?.mimeType !== newMedia.mimeType ||
+          msg.media?.fileName !== newMedia.fileName
+        ) {
+          hasChanges = true
+          return {
+            ...msg,
+            content: uploadedFile.name || msg.content,
+            media: newMedia,
+          }
+        }
+      }
+
+      return msg
+    })
+
+    if (hasChanges) {
+      setLocalMessages(updatedMessages)
+      setMessages(updatedMessages)
+    }
+  }, [activeWorkflowId, blockId, subBlockId, localMessages, isPreview, setMessages])
+
   const subBlockInput = useSubBlockInput({
     blockId,
     subBlockId,
@@ -186,6 +302,7 @@ export function MessagesInput({
   const wandHook = useWand({
     wandConfig: config.wandConfig,
     currentValue: getMessagesJson(),
+    sources: sourcesInfo,
     onStreamStart: () => {
       streamBufferRef.current = ''
       setLocalMessages([{ role: 'system', content: '' }])
@@ -200,6 +317,46 @@ export function MessagesInput({
     onGeneratedContent: (content) => {
       const validMessages = parseMessages(content)
       if (validMessages) {
+        // Process media messages - only allow fileId to set files, sanitize other attempts
+        validMessages.forEach((msg, index) => {
+          if (msg.role === 'media') {
+            // Check if this is an existing file with valid data (preserve it)
+            const hasExistingFile = msg.media?.sourceType === 'file' && 
+              msg.media?.data?.startsWith('/api/') && 
+              msg.media?.fileName
+
+            if (hasExistingFile) {
+              // Preserve existing file data as-is
+              return
+            }
+
+            // Check if wand provided a fileId to select a workspace file
+            if (msg.media?.fileId) {
+              const file = workspaceFiles.find((f) => f.id === msg.media?.fileId)
+              if (file) {
+                // Set the file value in SubBlockStore so FileUpload picks it up
+                const fileUploadKey = `${subBlockId}-media-${index}`
+                const uploadedFile = {
+                  name: file.name,
+                  path: file.path,
+                  type: file.type,
+                  size: 0, // Size not available from workspace files list
+                }
+                useSubBlockStore.getState().setValue(blockId, fileUploadKey, uploadedFile)
+
+                // Clear the media object - the FileUpload will sync the file data via useEffect
+                // DON'T set media.data here as it would appear in the ShortInput (advanced mode)
+                msg.media = undefined
+                return
+              }
+            }
+
+            // Sanitize: clear any media object that isn't a valid existing file or fileId match
+            // This prevents the LLM from setting arbitrary data/variable references
+            msg.media = undefined
+          }
+        })
+
         setLocalMessages(validMessages)
         setMessages(validMessages)
       } else {
@@ -283,6 +440,7 @@ export function MessagesInput({
           role,
           content: updatedMessages[index].content || '',
           media: updatedMessages[index].media || {
+            sourceType: 'file',
             data: '',
           },
         }
@@ -700,29 +858,41 @@ export function MessagesInput({
                         disabled={disabled}
                       />
                     ) : (
-                      <textarea
-                        ref={(el) => {
-                          textareaRefs.current[fieldId] = el
+                      <ShortInput
+                        blockId={blockId}
+                        subBlockId={`${subBlockId}-media-ref-${index}`}
+                        placeholder='Reference file from previous block...'
+                        config={{
+                          id: `${subBlockId}-media-ref-${index}`,
+                          type: 'short-input',
                         }}
-                        className='relative z-[2] m-0 box-border h-auto min-h-[60px] w-full resize-none overflow-y-auto overflow-x-hidden whitespace-pre-wrap break-words rounded-[4px] border border-[var(--border-1)] bg-transparent px-[8px] py-[6px] font-medium font-sans text-[var(--text-primary)] text-sm leading-[1.5] outline-none transition-colors [-ms-overflow-style:none] [scrollbar-width:none] placeholder:text-[var(--text-muted)] hover:border-[var(--border-2)] focus:border-[var(--border-2)] focus:outline-none focus-visible:outline-none disabled:cursor-not-allowed [&::-webkit-scrollbar]:hidden'
-                        placeholder='Enter URL or paste base64 data...'
-                        value={message.media?.data || ''}
-                        onChange={(e) => {
+                        value={
+                          // Only show value for variable references, not file uploads
+                          message.media?.sourceType === 'file' ? '' : (message.media?.data || '')
+                        }
+                        onChange={(newValue: string) => {
                           const updatedMessages = [...localMessages]
                           if (updatedMessages[index].role === 'media') {
+                            // Determine sourceType based on content
+                            let sourceType: 'url' | 'base64' = 'url'
+                            if (newValue.startsWith('data:') || newValue.includes(';base64,')) {
+                              sourceType = 'base64'
+                            }
                             updatedMessages[index] = {
                               ...updatedMessages[index],
-                              content: e.target.value.substring(0, 50),
+                              content: newValue.substring(0, 50),
                               media: {
                                 ...updatedMessages[index].media,
-                                data: e.target.value,
+                                sourceType,
+                                data: newValue,
                               },
                             }
+                            setLocalMessages(updatedMessages)
+                            setMessages(updatedMessages)
                           }
-                          setLocalMessages(updatedMessages)
-                          setMessages(updatedMessages)
                         }}
-                        disabled={isPreview || disabled}
+                        isPreview={isPreview}
+                        disabled={disabled}
                       />
                     )}
                   </div>

apps/sim/app/workspace/[workspaceId]/w/[workflowId]/hooks/use-wand.ts

@@ -63,6 +63,8 @@ export interface WandConfig {
 interface UseWandProps {
   wandConfig?: WandConfig
   currentValue?: string
+  /** Additional context about available sources/references for the prompt */
+  sources?: string
   onGeneratedContent: (content: string) => void
   onStreamChunk?: (chunk: string) => void
   onStreamStart?: () => void
@@ -72,6 +74,7 @@ interface UseWandProps {
 export function useWand({
   wandConfig,
   currentValue,
+  sources,
   onGeneratedContent,
   onStreamChunk,
   onStreamStart,
@@ -154,6 +157,9 @@ export function useWand({
         if (systemPrompt.includes('{context}')) {
           systemPrompt = systemPrompt.replace('{context}', contextInfo)
         }
+        if (systemPrompt.includes('{sources}')) {
+          systemPrompt = systemPrompt.replace('{sources}', sources || 'No upstream sources available')
+        }
 
         const userMessage = prompt
 

apps/sim/blocks/blocks/agent.ts

@@ -95,10 +95,12 @@ export const AgentBlock: BlockConfig<AgentResponse> = {
 
 Current messages: {context}
 
+{sources}
+
 RULES:
 1. Generate ONLY a valid JSON array - no markdown, no explanations
-2. Each message object must have "role" (system/user/assistant) and "content" (string)
-3. You can generate any number of messages as needed
+2. Each message object must have "role" and "content" properties
+3. Valid roles are: "system", "user", "assistant", "media"
 4. Content can be as long as necessary - don't truncate
 5. If editing existing messages, preserve structure unless asked to change it
 6. For new agents, create DETAILED, PROFESSIONAL system prompts that include:
@@ -108,6 +110,16 @@ RULES:
    - Critical thinking or quality guidelines
    - How to handle edge cases and uncertainty
 
+MEDIA MESSAGES:
+- Use role "media" to include images, audio, video, or documents in a multimodal conversation
+- IMPORTANT: If a media message in the current context has a "media" object with file data, ALWAYS preserve that entire "media" object exactly as-is
+- When creating NEW media messages, you can either:
+  1. Just set role to "media" with descriptive content - user will upload the file manually
+  2. Select a file from the available workspace files by including "fileId" in the media object (optional)
+- You do NOT have to select a file - it's completely optional
+- Example without file: {"role": "media", "content": "Analyze this image for text and objects"}
+- Example with file selection: {"role": "media", "content": "Analyze this image", "media": {"fileId": "abc123"}}
+
 EXAMPLES:
 
 Research agent:
@@ -116,8 +128,8 @@ Research agent:
 Code reviewer:
 [{"role": "system", "content": "You are a Senior Code Reviewer with expertise in software architecture, security, and best practices. Your role is to provide thorough, constructive code reviews that improve code quality and help developers grow.\\n\\n## Review Methodology\\n\\n1. **Security First**: Check for vulnerabilities including injection attacks, authentication flaws, data exposure, and insecure dependencies.\\n\\n2. **Code Quality**: Evaluate readability, maintainability, adherence to DRY/SOLID principles, and appropriate abstraction levels.\\n\\n3. **Performance**: Identify potential bottlenecks, unnecessary computations, memory leaks, and optimization opportunities.\\n\\n4. **Testing**: Assess test coverage, edge case handling, and testability of the code structure.\\n\\n## Output Format\\n\\n### Summary\\nBrief overview of the code's purpose and overall assessment.\\n\\n### Critical Issues\\nSecurity vulnerabilities or bugs that must be fixed before merging.\\n\\n### Improvements\\nSuggested enhancements with clear explanations of why and how.\\n\\n### Positive Aspects\\nHighlight well-written code to reinforce good practices.\\n\\nBe specific with line references. Provide code examples for suggested changes. Balance critique with encouragement."}, {"role": "user", "content": "<start.input>"}]
 
-Writing assistant:
-[{"role": "system", "content": "You are a skilled Writing Editor and Coach. Your role is to help users improve their writing through constructive feedback, editing suggestions, and guidance on style, clarity, and structure.\\n\\n## Editing Approach\\n\\n1. **Clarity**: Ensure ideas are expressed clearly and concisely. Eliminate jargon unless appropriate for the audience.\\n\\n2. **Structure**: Evaluate logical flow, paragraph organization, and transitions between ideas.\\n\\n3. **Voice & Tone**: Maintain consistency and appropriateness for the intended audience and purpose.\\n\\n4. **Grammar & Style**: Correct errors while respecting the author's voice.\\n\\n## Output Format\\n\\n### Overall Impression\\nBrief assessment of the piece's strengths and areas for improvement.\\n\\n### Structural Feedback\\nComments on organization, flow, and logical progression.\\n\\n### Line-Level Edits\\nSpecific suggestions with explanations, not just corrections.\\n\\n### Revised Version\\nWhen appropriate, provide an edited version demonstrating improvements.\\n\\nBe encouraging while honest. Explain the reasoning behind suggestions to help the writer improve."}, {"role": "user", "content": "<start.input>"}]
+Image analysis agent:
+[{"role": "system", "content": "You are an expert image analyst. Describe images in detail, identify objects, text, and patterns. Provide structured analysis."}, {"role": "media", "content": "Analyze this image"}]
 
 Return ONLY the JSON array.`,
         placeholder: 'Describe what you want to create or change...',

apps/sim/executor/handlers/agent/agent-handler.ts

@@ -811,17 +811,41 @@ export class AgentBlockHandler implements BlockHandler {
     return messages.length > 0 ? messages : undefined
   }
 
-  private extractValidMessages(messages?: Message[]): Message[] {
-    if (!messages || !Array.isArray(messages)) return []
+  private extractValidMessages(messages?: Message[] | string): Message[] {
+    if (!messages) return []
 
-    return messages.filter(
-      (msg): msg is Message =>
-        msg &&
-        typeof msg === 'object' &&
-        'role' in msg &&
-        'content' in msg &&
-        ['system', 'user', 'assistant', 'media'].includes(msg.role)
-    )
+    // Handle raw JSON string input (from advanced mode)
+    let messageArray: unknown[]
+    if (typeof messages === 'string') {
+      const trimmed = messages.trim()
+      if (!trimmed) return []
+      try {
+        const parsed = JSON.parse(trimmed)
+        if (!Array.isArray(parsed)) {
+          logger.warn('Parsed messages JSON is not an array', { parsed })
+          return []
+        }
+        messageArray = parsed
+      } catch (error) {
+        logger.warn('Failed to parse messages JSON string', { error, messages: trimmed.substring(0, 100) })
+        return []
+      }
+    } else if (Array.isArray(messages)) {
+      messageArray = messages
+    } else {
+      return []
+    }
+
+    return messageArray.filter((msg): msg is Message => {
+      if (!msg || typeof msg !== 'object') return false
+      const m = msg as Record<string, unknown>
+      return (
+        'role' in m &&
+        'content' in m &&
+        typeof m.role === 'string' &&
+        ['system', 'user', 'assistant', 'media'].includes(m.role)
+      )
+    })
   }
 
   /**
@@ -914,6 +938,35 @@ export class AgentBlockHandler implements BlockHandler {
 
     const { sourceType, data, mimeType } = media
 
+    // Validate data is not empty
+    if (!data || !data.trim()) {
+      logger.warn('Empty media data, skipping media content')
+      return null
+    }
+
+    // Validate URL format if sourceType is URL
+    if (sourceType === 'url' || sourceType === 'file') {
+      const trimmedData = data.trim()
+      // Must start with http://, https://, or / (relative path for workspace files)
+      if (!trimmedData.startsWith('http://') && 
+          !trimmedData.startsWith('https://') && 
+          !trimmedData.startsWith('/')) {
+        logger.warn('Invalid URL format for media content', { data: trimmedData.substring(0, 50) })
+        // Try to salvage by treating as text
+        return { type: 'text', text: `[Invalid media URL: ${trimmedData.substring(0, 30)}...]` }
+      }
+    }
+
+    // Validate base64 format
+    if (sourceType === 'base64') {
+      const trimmedData = data.trim()
+      // Should be a data URL or raw base64
+      if (!trimmedData.startsWith('data:') && !/^[A-Za-z0-9+/]+=*$/.test(trimmedData.replace(/\s/g, ''))) {
+        logger.warn('Invalid base64 format for media content', { data: trimmedData.substring(0, 50) })
+        return { type: 'text', text: `[Invalid base64 data]` }
+      }
+    }
+
     switch (providerId) {
       case 'anthropic':
         return this.createAnthropicMediaContent(sourceType, data, mimeType)
@@ -922,8 +975,14 @@ export class AgentBlockHandler implements BlockHandler {
       case 'vertex':
         return this.createGeminiMediaContent(sourceType, data, mimeType)
 
+      case 'mistral':
+        return this.createMistralMediaContent(sourceType, data, mimeType)
+
+      case 'bedrock':
+        return this.createBedrockMediaContent(sourceType, data, mimeType)
+
       default:
-        // OpenAI format (used by OpenAI, Azure, xAI, Mistral, etc.)
+        // OpenAI format (used by OpenAI, Azure, xAI, Groq, etc.)
         return this.createOpenAIMediaContent(sourceType, data, mimeType)
     }
   }
@@ -938,15 +997,10 @@ export class AgentBlockHandler implements BlockHandler {
   ): any {
     const isImage = mimeType?.startsWith('image/')
     const isAudio = mimeType?.startsWith('audio/')
+    // Treat 'file' as 'url' since workspace files are served via URL
+    const isUrl = sourceType === 'url' || sourceType === 'file'
 
     if (isImage) {
-      if (sourceType === 'url') {
-        return {
-          type: 'image_url',
-          image_url: { url: data, detail: 'auto' },
-        }
-      }
-      // base64 or file (already converted to base64)
       return {
         type: 'image_url',
         image_url: { url: data, detail: 'auto' },
@@ -990,9 +1044,11 @@ export class AgentBlockHandler implements BlockHandler {
   ): any {
     const isImage = mimeType?.startsWith('image/')
     const isPdf = mimeType === 'application/pdf'
+    // Treat 'file' as 'url' since workspace files are served via URL
+    const isUrl = sourceType === 'url' || sourceType === 'file'
 
     if (isImage) {
-      if (sourceType === 'url') {
+      if (isUrl) {
         return {
           type: 'image',
           source: { type: 'url', url: data },
@@ -1011,7 +1067,7 @@ export class AgentBlockHandler implements BlockHandler {
     }
 
     if (isPdf) {
-      if (sourceType === 'url') {
+      if (isUrl) {
         return {
           type: 'document',
           source: { type: 'url', url: data },
@@ -1043,7 +1099,10 @@ export class AgentBlockHandler implements BlockHandler {
     data: string,
     mimeType?: string
   ): any {
-    if (sourceType === 'url') {
+    // Treat 'file' as 'url' since workspace files are served via URL
+    const isUrl = sourceType === 'url' || sourceType === 'file'
+
+    if (isUrl) {
       return {
         fileData: {
           mimeType: mimeType || 'application/octet-stream',
@@ -1062,6 +1121,114 @@ export class AgentBlockHandler implements BlockHandler {
     }
   }
 
+  /**
+   * Creates Mistral-compatible media content
+   * Note: Mistral uses a simplified format where image_url is a direct string,
+   * NOT a nested object like OpenAI
+   */
+  private createMistralMediaContent(
+    sourceType: string,
+    data: string,
+    mimeType?: string
+  ): any {
+    const isImage = mimeType?.startsWith('image/')
+    // Treat 'file' as 'url' since workspace files are served via URL
+    const isUrl = sourceType === 'url' || sourceType === 'file'
+
+    if (isImage) {
+      if (isUrl) {
+        // Mistral uses direct string for image_url, not nested object
+        return {
+          type: 'image_url',
+          image_url: data,
+        }
+      }
+      // Base64 - Mistral accepts data URLs directly
+      const base64Data = data.includes(',') ? data : `data:${mimeType || 'image/png'};base64,${data}`
+      return {
+        type: 'image_url',
+        image_url: base64Data,
+      }
+    }
+
+    // Fallback for non-image types
+    return {
+      type: 'text',
+      text: `[File: ${mimeType || 'unknown type'}]`,
+    }
+  }
+
+  /**
+   * Creates AWS Bedrock Converse API-compatible media content
+   * Bedrock uses a different structure: { image: { format, source: { bytes } } }
+   * Note: The actual bytes conversion happens in the provider layer
+   */
+  private createBedrockMediaContent(
+    sourceType: string,
+    data: string,
+    mimeType?: string
+  ): any {
+    const isImage = mimeType?.startsWith('image/')
+    // Treat 'file' as 'url' since workspace files are served via URL
+    const isUrl = sourceType === 'url' || sourceType === 'file'
+
+    // Determine format from mimeType
+    const getFormat = (mime?: string): string => {
+      if (!mime) return 'png'
+      if (mime.includes('jpeg') || mime.includes('jpg')) return 'jpeg'
+      if (mime.includes('png')) return 'png'
+      if (mime.includes('gif')) return 'gif'
+      if (mime.includes('webp')) return 'webp'
+      return 'png'
+    }
+
+    if (isImage) {
+      if (isUrl) {
+        // For URLs, Bedrock needs S3 URIs or we need to fetch and convert
+        // Mark this for the provider layer to handle
+        return {
+          type: 'bedrock_image',
+          format: getFormat(mimeType),
+          sourceType: 'url',
+          url: data,
+        }
+      }
+      // Base64 - extract raw base64 data
+      const base64Data = data.includes(',') ? data.split(',')[1] : data
+      return {
+        type: 'bedrock_image',
+        format: getFormat(mimeType),
+        sourceType: 'base64',
+        data: base64Data,
+      }
+    }
+
+    // Documents (PDFs) - Bedrock supports document content type
+    if (mimeType === 'application/pdf') {
+      if (isUrl) {
+        return {
+          type: 'bedrock_document',
+          format: 'pdf',
+          sourceType: 'url',
+          url: data,
+        }
+      }
+      const base64Data = data.includes(',') ? data.split(',')[1] : data
+      return {
+        type: 'bedrock_document',
+        format: 'pdf',
+        sourceType: 'base64',
+        data: base64Data,
+      }
+    }
+
+    // Fallback for unsupported types
+    return {
+      type: 'text',
+      text: `[File: ${mimeType || 'unknown type'}]`,
+    }
+  }
+
   private processMemories(memories: any): Message[] {
     if (!memories) return []
 

apps/sim/executor/handlers/agent/types.ts

@@ -6,8 +6,8 @@ export interface AgentInputs {
   systemPrompt?: string
   userPrompt?: string | object
   memories?: any // Legacy memory block output
-  // New message array input (from messages-input subblock)
-  messages?: Message[]
+  // New message array input (from messages-input subblock or raw JSON from advanced mode)
+  messages?: Message[] | string
   // Memory configuration
   memoryType?: 'none' | 'conversation' | 'sliding_window' | 'sliding_window_tokens'
   conversationId?: string // Required for all non-none memory types
@@ -46,8 +46,8 @@ export interface ToolInput {
  * Media content for multimodal messages
  */
 export interface MediaContent {
-  /** Mode: basic (file upload) or advanced (URL/base64 text input) */
-  mode: 'basic' | 'advanced'
+  /** Source type: how the data was provided */
+  sourceType: 'url' | 'base64' | 'file'
   /** The URL or base64 data */
   data: string
   /** MIME type (e.g., 'image/png', 'application/pdf', 'audio/mp3') */

apps/sim/providers/bedrock/index.ts

@@ -16,6 +16,7 @@ import type { StreamingExecution } from '@/executor/types'
 import { MAX_TOOL_ITERATIONS } from '@/providers'
 import {
   checkForForcedToolUsage,
+  convertToBedrockContentBlocks,
   createReadableStreamFromBedrockStream,
   generateToolUseId,
   getBedrockInferenceProfileId,
@@ -116,9 +117,11 @@ export const bedrockProvider: ProviderConfig = {
           }
         } else {
           const role: ConversationRole = msg.role === 'assistant' ? 'assistant' : 'user'
+          // Handle multimodal content arrays
+          const contentBlocks = convertToBedrockContentBlocks(msg.content || '')
           messages.push({
             role,
-            content: [{ text: msg.content || '' }],
+            content: contentBlocks,
           })
         }
       }

apps/sim/providers/bedrock/utils.ts

@@ -1,9 +1,195 @@
-import type { ConverseStreamOutput } from '@aws-sdk/client-bedrock-runtime'
+import type { ContentBlock, ConverseStreamOutput, ImageFormat } from '@aws-sdk/client-bedrock-runtime'
 import { createLogger } from '@sim/logger'
 import { trackForcedToolUsage } from '@/providers/utils'
 
 const logger = createLogger('BedrockUtils')
 
+/**
+ * Converts message content (string or array) to Bedrock ContentBlock array.
+ * Handles multimodal content including images and documents.
+ */
+export function convertToBedrockContentBlocks(content: string | any[]): ContentBlock[] {
+  // Simple string content
+  if (typeof content === 'string') {
+    return [{ text: content || '' }]
+  }
+
+  // Array content - could be multimodal
+  if (!Array.isArray(content)) {
+    return [{ text: String(content) || '' }]
+  }
+
+  const blocks: ContentBlock[] = []
+
+  for (const item of content) {
+    if (!item) continue
+
+    // Text content
+    if (item.type === 'text' && item.text) {
+      blocks.push({ text: item.text })
+      continue
+    }
+
+    // Gemini-style text (just { text: "..." })
+    if (typeof item.text === 'string' && !item.type) {
+      blocks.push({ text: item.text })
+      continue
+    }
+
+    // Bedrock image content (from agent handler)
+    if (item.type === 'bedrock_image') {
+      const imageBlock = createBedrockImageBlock(item)
+      if (imageBlock) {
+        blocks.push(imageBlock)
+      }
+      continue
+    }
+
+    // Bedrock document content (from agent handler)
+    if (item.type === 'bedrock_document') {
+      const docBlock = createBedrockDocumentBlock(item)
+      if (docBlock) {
+        blocks.push(docBlock)
+      }
+      continue
+    }
+
+    // OpenAI-style image_url (fallback for direct OpenAI format)
+    if (item.type === 'image_url' && item.image_url) {
+      const url = typeof item.image_url === 'string' ? item.image_url : item.image_url?.url
+      if (url) {
+        const imageBlock = createBedrockImageBlockFromUrl(url)
+        if (imageBlock) {
+          blocks.push(imageBlock)
+        }
+      }
+      continue
+    }
+
+    // Unknown type - log warning and skip
+    logger.warn('Unknown content block type in Bedrock conversion:', { type: item.type })
+  }
+
+  // Ensure at least one text block
+  if (blocks.length === 0) {
+    blocks.push({ text: '' })
+  }
+
+  return blocks
+}
+
+/**
+ * Creates a Bedrock image ContentBlock from a bedrock_image item
+ */
+function createBedrockImageBlock(item: {
+  format: string
+  sourceType: string
+  data?: string
+  url?: string
+}): ContentBlock | null {
+  const format = (item.format || 'png') as ImageFormat
+
+  if (item.sourceType === 'base64' && item.data) {
+    // Convert base64 to Uint8Array
+    const bytes = base64ToUint8Array(item.data)
+    return {
+      image: {
+        format,
+        source: { bytes },
+      },
+    }
+  }
+
+  if (item.sourceType === 'url' && item.url) {
+    // For URLs, we need to fetch the image and convert to bytes
+    // This is a limitation - Bedrock doesn't support URL sources directly
+    // The provider layer should handle this, or we log a warning
+    logger.warn('Bedrock does not support image URLs directly. Image will be skipped.', {
+      url: item.url,
+    })
+    // Return a text placeholder
+    return { text: `[Image from URL: ${item.url}]` }
+  }
+
+  return null
+}
+
+/**
+ * Creates a Bedrock document ContentBlock from a bedrock_document item
+ */
+function createBedrockDocumentBlock(item: {
+  format: string
+  sourceType: string
+  data?: string
+  url?: string
+}): ContentBlock | null {
+  if (item.sourceType === 'base64' && item.data) {
+    const bytes = base64ToUint8Array(item.data)
+    return {
+      document: {
+        format: 'pdf',
+        name: 'document',
+        source: { bytes },
+      },
+    }
+  }
+
+  if (item.sourceType === 'url' && item.url) {
+    logger.warn('Bedrock does not support document URLs directly. Document will be skipped.', {
+      url: item.url,
+    })
+    return { text: `[Document from URL: ${item.url}]` }
+  }
+
+  return null
+}
+
+/**
+ * Creates a Bedrock image ContentBlock from a data URL or regular URL
+ */
+function createBedrockImageBlockFromUrl(url: string): ContentBlock | null {
+  // Check if it's a data URL (base64)
+  if (url.startsWith('data:')) {
+    const match = url.match(/^data:image\/(\w+);base64,(.+)$/)
+    if (match) {
+      let format: ImageFormat = match[1] as ImageFormat
+      // Normalize jpg to jpeg
+      if (format === ('jpg' as ImageFormat)) {
+        format = 'jpeg'
+      }
+      const base64Data = match[2]
+      const bytes = base64ToUint8Array(base64Data)
+      return {
+        image: {
+          format,
+          source: { bytes },
+        },
+      }
+    }
+  }
+
+  // Regular URL - Bedrock doesn't support this directly
+  logger.warn('Bedrock does not support image URLs directly. Image will be skipped.', { url })
+  return { text: `[Image from URL: ${url}]` }
+}
+
+/**
+ * Converts a base64 string to Uint8Array
+ */
+function base64ToUint8Array(base64: string): Uint8Array {
+  // Handle browser and Node.js environments
+  if (typeof Buffer !== 'undefined') {
+    return Buffer.from(base64, 'base64')
+  }
+  // Browser fallback
+  const binaryString = atob(base64)
+  const bytes = new Uint8Array(binaryString.length)
+  for (let i = 0; i < binaryString.length; i++) {
+    bytes[i] = binaryString.charCodeAt(i)
+  }
+  return bytes
+}
+
 export interface BedrockStreamUsage {
   inputTokens: number
   outputTokens: number

apps/sim/providers/models.ts

@@ -34,6 +34,8 @@ export interface ModelCapabilities {
   toolUsageControl?: boolean
   computerUse?: boolean
   nativeStructuredOutputs?: boolean
+  /** Whether the model supports vision/multimodal inputs (images, audio, video, PDFs) */
+  vision?: boolean
   maxOutputTokens?: {
     /** Maximum tokens for streaming requests */
     max: number
@@ -120,6 +122,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 2 },
+          vision: true,
         },
         contextWindow: 128000,
       },
@@ -132,6 +135,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
           updatedAt: '2025-12-11',
         },
         capabilities: {
+          vision: true,
           reasoningEffort: {
             values: ['none', 'minimal', 'low', 'medium', 'high', 'xhigh'],
           },
@@ -150,6 +154,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
           updatedAt: '2025-11-14',
         },
         capabilities: {
+          vision: true,
           reasoningEffort: {
             values: ['none', 'low', 'medium', 'high'],
           },
@@ -222,6 +227,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
           updatedAt: '2025-08-07',
         },
         capabilities: {
+          vision: true,
           reasoningEffort: {
             values: ['minimal', 'low', 'medium', 'high'],
           },
@@ -240,6 +246,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
           updatedAt: '2025-08-07',
         },
         capabilities: {
+          vision: true,
           reasoningEffort: {
             values: ['minimal', 'low', 'medium', 'high'],
           },
@@ -258,6 +265,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
           updatedAt: '2025-08-07',
         },
         capabilities: {
+          vision: true,
           reasoningEffort: {
             values: ['minimal', 'low', 'medium', 'high'],
           },
@@ -287,6 +295,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
           updatedAt: '2025-06-17',
         },
         capabilities: {
+          vision: true,
           reasoningEffort: {
             values: ['low', 'medium', 'high'],
           },
@@ -302,6 +311,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
           updatedAt: '2025-06-17',
         },
         capabilities: {
+          vision: true,
           reasoningEffort: {
             values: ['low', 'medium', 'high'],
           },
@@ -317,6 +327,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
           updatedAt: '2025-06-17',
         },
         capabilities: {
+          vision: true,
           reasoningEffort: {
             values: ['low', 'medium', 'high'],
           },
@@ -333,6 +344,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 2 },
+          vision: true,
         },
         contextWindow: 1000000,
       },
@@ -346,6 +358,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 2 },
+          vision: true,
         },
         contextWindow: 1000000,
       },
@@ -359,6 +372,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 2 },
+          vision: true,
         },
         contextWindow: 1000000,
       },
@@ -385,6 +399,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 2 },
+          vision: true,
         },
         contextWindow: 128000,
       },
@@ -397,6 +412,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
           updatedAt: '2025-12-11',
         },
         capabilities: {
+          vision: true,
           reasoningEffort: {
             values: ['none', 'minimal', 'low', 'medium', 'high', 'xhigh'],
           },
@@ -415,6 +431,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
           updatedAt: '2025-11-14',
         },
         capabilities: {
+          vision: true,
           reasoningEffort: {
             values: ['none', 'low', 'medium', 'high'],
           },
@@ -433,6 +450,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
           updatedAt: '2025-11-14',
         },
         capabilities: {
+          vision: true,
           reasoningEffort: {
             values: ['none', 'low', 'medium', 'high'],
           },
@@ -451,6 +469,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
           updatedAt: '2025-11-14',
         },
         capabilities: {
+          vision: true,
           reasoningEffort: {
             values: ['none', 'low', 'medium', 'high'],
           },
@@ -469,6 +488,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
           updatedAt: '2025-11-14',
         },
         capabilities: {
+          vision: true,
           reasoningEffort: {
             values: ['none', 'medium', 'high'],
           },
@@ -487,6 +507,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
           updatedAt: '2025-08-07',
         },
         capabilities: {
+          vision: true,
           reasoningEffort: {
             values: ['minimal', 'low', 'medium', 'high'],
           },
@@ -505,6 +526,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
           updatedAt: '2025-08-07',
         },
         capabilities: {
+          vision: true,
           reasoningEffort: {
             values: ['minimal', 'low', 'medium', 'high'],
           },
@@ -523,6 +545,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
           updatedAt: '2025-08-07',
         },
         capabilities: {
+          vision: true,
           reasoningEffort: {
             values: ['minimal', 'low', 'medium', 'high'],
           },
@@ -552,6 +575,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
           updatedAt: '2025-06-15',
         },
         capabilities: {
+          vision: true,
           reasoningEffort: {
             values: ['low', 'medium', 'high'],
           },
@@ -567,6 +591,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
           updatedAt: '2025-06-15',
         },
         capabilities: {
+          vision: true,
           reasoningEffort: {
             values: ['low', 'medium', 'high'],
           },
@@ -581,7 +606,9 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
           output: 8.0,
           updatedAt: '2025-06-15',
         },
-        capabilities: {},
+        capabilities: {
+          vision: true,
+        },
         contextWindow: 1000000,
       },
       {
@@ -620,6 +647,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
           temperature: { min: 0, max: 1 },
           nativeStructuredOutputs: true,
           maxOutputTokens: { max: 64000, default: 8192 },
+          vision: true,
         },
         contextWindow: 200000,
       },
@@ -635,6 +663,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
           temperature: { min: 0, max: 1 },
           nativeStructuredOutputs: true,
           maxOutputTokens: { max: 64000, default: 8192 },
+          vision: true,
         },
         contextWindow: 200000,
       },
@@ -649,6 +678,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         capabilities: {
           temperature: { min: 0, max: 1 },
           maxOutputTokens: { max: 64000, default: 8192 },
+          vision: true,
         },
         contextWindow: 200000,
       },
@@ -664,6 +694,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
           temperature: { min: 0, max: 1 },
           nativeStructuredOutputs: true,
           maxOutputTokens: { max: 64000, default: 8192 },
+          vision: true,
         },
         contextWindow: 200000,
       },
@@ -679,6 +710,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
           temperature: { min: 0, max: 1 },
           nativeStructuredOutputs: true,
           maxOutputTokens: { max: 64000, default: 8192 },
+          vision: true,
         },
         contextWindow: 200000,
       },
@@ -693,6 +725,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         capabilities: {
           temperature: { min: 0, max: 1 },
           maxOutputTokens: { max: 64000, default: 8192 },
+          vision: true,
         },
         contextWindow: 200000,
       },
@@ -708,6 +741,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
           temperature: { min: 0, max: 1 },
           computerUse: true,
           maxOutputTokens: { max: 8192, default: 8192 },
+          vision: true,
         },
         contextWindow: 200000,
       },
@@ -723,6 +757,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
           temperature: { min: 0, max: 1 },
           computerUse: true,
           maxOutputTokens: { max: 8192, default: 8192 },
+          vision: true,
         },
         contextWindow: 200000,
       },
@@ -736,6 +771,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
     modelPatterns: [/^gemini/],
     capabilities: {
       toolUsageControl: true,
+      vision: true,
     },
     icon: GeminiIcon,
     models: [
@@ -847,6 +883,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
     icon: VertexIcon,
     capabilities: {
       toolUsageControl: true,
+      vision: true,
     },
     models: [
       {
@@ -1005,6 +1042,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
     icon: xAIIcon,
     capabilities: {
       toolUsageControl: true,
+      vision: true,
     },
     models: [
       {
@@ -1277,7 +1315,9 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
           output: 0.34,
           updatedAt: '2026-01-27',
         },
-        capabilities: {},
+        capabilities: {
+          vision: true,
+        },
         contextWindow: 131072,
       },
       {
@@ -1287,7 +1327,9 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
           output: 0.6,
           updatedAt: '2026-01-27',
         },
-        capabilities: {},
+        capabilities: {
+          vision: true,
+        },
         contextWindow: 131072,
       },
       {
@@ -1369,6 +1411,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 1 },
+          vision: true,
         },
         contextWindow: 256000,
       },
@@ -1381,6 +1424,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 1 },
+          vision: true,
         },
         contextWindow: 256000,
       },
@@ -1453,6 +1497,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 1 },
+          vision: true,
         },
         contextWindow: 128000,
       },
@@ -1465,6 +1510,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 1 },
+          vision: true,
         },
         contextWindow: 128000,
       },
@@ -1489,6 +1535,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 1 },
+          vision: true,
         },
         contextWindow: 128000,
       },
@@ -1501,6 +1548,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 1 },
+          vision: true,
         },
         contextWindow: 128000,
       },
@@ -1549,6 +1597,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 1 },
+          vision: true,
         },
         contextWindow: 128000,
       },
@@ -1561,6 +1610,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 1 },
+          vision: true,
         },
         contextWindow: 128000,
       },
@@ -1585,6 +1635,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 1 },
+          vision: true,
         },
         contextWindow: 256000,
       },
@@ -1597,6 +1648,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 1 },
+          vision: true,
         },
         contextWindow: 256000,
       },
@@ -1609,6 +1661,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 1 },
+          vision: true,
         },
         contextWindow: 256000,
       },
@@ -1621,6 +1674,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 1 },
+          vision: true,
         },
         contextWindow: 256000,
       },
@@ -1645,6 +1699,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 1 },
+          vision: true,
         },
         contextWindow: 256000,
       },
@@ -1657,6 +1712,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 1 },
+          vision: true,
         },
         contextWindow: 256000,
       },
@@ -1710,6 +1766,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
           temperature: { min: 0, max: 1 },
           nativeStructuredOutputs: true,
           maxOutputTokens: { max: 64000, default: 8192 },
+          vision: true,
         },
         contextWindow: 200000,
       },
@@ -1724,6 +1781,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
           temperature: { min: 0, max: 1 },
           nativeStructuredOutputs: true,
           maxOutputTokens: { max: 64000, default: 8192 },
+          vision: true,
         },
         contextWindow: 200000,
       },
@@ -1738,6 +1796,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
           temperature: { min: 0, max: 1 },
           nativeStructuredOutputs: true,
           maxOutputTokens: { max: 64000, default: 8192 },
+          vision: true,
         },
         contextWindow: 200000,
       },
@@ -1752,6 +1811,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
           temperature: { min: 0, max: 1 },
           nativeStructuredOutputs: true,
           maxOutputTokens: { max: 64000, default: 8192 },
+          vision: true,
         },
         contextWindow: 200000,
       },
@@ -1764,6 +1824,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 1 },
+          vision: true,
         },
         contextWindow: 1000000,
       },
@@ -1776,6 +1837,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 1 },
+          vision: true,
         },
         contextWindow: 1000000,
       },
@@ -1788,6 +1850,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 1 },
+          vision: true,
         },
         contextWindow: 1000000,
       },
@@ -1800,6 +1863,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 1 },
+          vision: true,
         },
         contextWindow: 300000,
       },
@@ -1812,6 +1876,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 1 },
+          vision: true,
         },
         contextWindow: 300000,
       },
@@ -1836,6 +1901,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 1 },
+          vision: true,
         },
         contextWindow: 1000000,
       },
@@ -1848,6 +1914,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 1 },
+          vision: true,
         },
         contextWindow: 3500000,
       },
@@ -1872,6 +1939,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 1 },
+          vision: true,
         },
         contextWindow: 128000,
       },
@@ -1884,6 +1952,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 1 },
+          vision: true,
         },
         contextWindow: 128000,
       },
@@ -1956,6 +2025,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 1 },
+          vision: true,
         },
         contextWindow: 128000,
       },
@@ -1992,6 +2062,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 1 },
+          vision: true,
         },
         contextWindow: 128000,
       },
@@ -2016,6 +2087,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 1 },
+          vision: true,
         },
         contextWindow: 128000,
       },
@@ -2028,6 +2100,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 1 },
+          vision: true,
         },
         contextWindow: 128000,
       },
@@ -2040,6 +2113,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 1 },
+          vision: true,
         },
         contextWindow: 128000,
       },
@@ -2211,6 +2285,32 @@ export function getMaxTemperature(modelId: string): number | undefined {
   return capabilities?.temperature?.max
 }
 
+/**
+ * Checks if a model supports vision/multimodal inputs (images, audio, video, PDFs)
+ */
+export function supportsVision(modelId: string): boolean {
+  const capabilities = getModelCapabilities(modelId)
+  return !!capabilities?.vision
+}
+
+/**
+ * Returns a list of all vision-capable models
+ */
+export function getVisionModels(): string[] {
+  const models: string[] = []
+  for (const provider of Object.values(PROVIDER_DEFINITIONS)) {
+    // Check if the provider has vision capability at the provider level
+    const providerHasVision = provider.capabilities?.vision
+    for (const model of provider.models) {
+      // Model has vision if either the model or provider has vision capability
+      if (model.capabilities.vision || providerHasVision) {
+        models.push(model.id)
+      }
+    }
+  }
+  return models
+}
+
 export function supportsToolUsageControl(providerId: string): boolean {
   return getProvidersWithToolUsageControl().includes(providerId)
 }

apps/sim/providers/types.ts

@@ -115,8 +115,8 @@ export interface ProviderToolConfig {
  * Media content for multimodal messages
  */
 export interface MediaContent {
-  /** Mode: basic (file upload) or advanced (URL/base64 text input) */
-  mode: 'basic' | 'advanced'
+  /** Source type: how the data was provided */
+  sourceType: 'url' | 'base64' | 'file'
   /** The URL or base64 data */
   data: string
   /** MIME type (e.g., 'image/png', 'application/pdf', 'audio/mp3') */

apps/sim/providers/utils.ts

@@ -23,9 +23,11 @@ import {
   getReasoningEffortValuesForModel as getReasoningEffortValuesForModelFromDefinitions,
   getThinkingLevelsForModel as getThinkingLevelsForModelFromDefinitions,
   getVerbosityValuesForModel as getVerbosityValuesForModelFromDefinitions,
+  getVisionModels,
   PROVIDER_DEFINITIONS,
   supportsTemperature as supportsTemperatureFromDefinitions,
   supportsToolUsageControl as supportsToolUsageControlFromDefinitions,
+  supportsVision,
   updateOllamaModels as updateOllamaModelsInDefinitions,
 } from '@/providers/models'
 import type { ProviderId, ProviderToolConfig } from '@/providers/types'
@@ -1152,3 +1154,6 @@ export function checkForForcedToolUsageOpenAI(
 
   return { hasUsedForcedTool, usedForcedTools: updatedUsedForcedTools }
 }
+
+// Re-export vision capability functions
+export { supportsVision, getVisionModels }