Fix

.claude/commands/add-tools.md

@@ -55,21 +55,21 @@ export const {serviceName}{Action}Tool: ToolConfig<
   },
 
   params: {
-    // Hidden params (system-injected, only use hidden for oauth accessToken)
+    // Hidden params (system-injected)
     accessToken: {
       type: 'string',
       required: true,
       visibility: 'hidden',
       description: 'OAuth access token',
     },
-    // User-only params (credentials, api key, IDs user must provide)
+    // User-only params (credentials, IDs user must provide)
     someId: {
       type: 'string',
       required: true,
       visibility: 'user-only',
       description: 'The ID of the resource',
     },
-    // User-or-LLM params (everything else, can be provided by user OR computed by LLM)
+    // User-or-LLM params (can be provided by user OR computed by LLM)
     query: {
       type: 'string',
       required: false,                // Use false for optional
@@ -114,8 +114,8 @@ export const {serviceName}{Action}Tool: ToolConfig<
 
 ### Visibility Options
 - `'hidden'` - System-injected (OAuth tokens, internal params). User never sees.
-- `'user-only'` - User must provide (credentials, api keys, account-specific IDs)
-- `'user-or-llm'` - User provides OR LLM can compute (search queries, content, filters, most fall into this category)
+- `'user-only'` - User must provide (credentials, account-specific IDs)
+- `'user-or-llm'` - User provides OR LLM can compute (search queries, content, filters)
 
 ### Parameter Types
 - `'string'` - Text values

apps/sim/app/api/workflows/[id]/autolayout/route.ts

@@ -35,7 +35,8 @@ const AutoLayoutRequestSchema = z.object({
     })
     .optional()
     .default({}),
-  gridSize: z.number().min(0).max(50).optional(),
+  // Optional: if provided, use these blocks instead of loading from DB
+  // This allows using blocks with live measurements from the UI
   blocks: z.record(z.any()).optional(),
   edges: z.array(z.any()).optional(),
   loops: z.record(z.any()).optional(),
@@ -52,6 +53,7 @@ export async function POST(request: NextRequest, { params }: { params: Promise<{
   const { id: workflowId } = await params
 
   try {
+    // Get the session
     const session = await getSession()
     if (!session?.user?.id) {
       logger.warn(`[${requestId}] Unauthorized autolayout attempt for workflow ${workflowId}`)
@@ -60,6 +62,7 @@ export async function POST(request: NextRequest, { params }: { params: Promise<{
 
     const userId = session.user.id
 
+    // Parse request body
     const body = await request.json()
     const layoutOptions = AutoLayoutRequestSchema.parse(body)
 
@@ -67,6 +70,7 @@ export async function POST(request: NextRequest, { params }: { params: Promise<{
       userId,
     })
 
+    // Fetch the workflow to check ownership/access
     const accessContext = await getWorkflowAccessContext(workflowId, userId)
     const workflowData = accessContext?.workflow
 
@@ -75,6 +79,7 @@ export async function POST(request: NextRequest, { params }: { params: Promise<{
       return NextResponse.json({ error: 'Workflow not found' }, { status: 404 })
     }
 
+    // Check if user has permission to update this workflow
     const canUpdate =
       accessContext?.isOwner ||
       (workflowData.workspaceId
@@ -89,6 +94,8 @@ export async function POST(request: NextRequest, { params }: { params: Promise<{
       return NextResponse.json({ error: 'Access denied' }, { status: 403 })
     }
 
+    // Use provided blocks/edges if available (with live measurements from UI),
+    // otherwise load from database
     let currentWorkflowData: NormalizedWorkflowData | null
 
     if (layoutOptions.blocks && layoutOptions.edges) {
@@ -118,7 +125,6 @@ export async function POST(request: NextRequest, { params }: { params: Promise<{
         y: layoutOptions.padding?.y ?? DEFAULT_LAYOUT_PADDING.y,
       },
       alignment: layoutOptions.alignment,
-      gridSize: layoutOptions.gridSize,
     }
 
     const layoutResult = applyAutoLayout(

apps/sim/app/api/workflows/[id]/execute-from-block/route.ts

@@ -0,0 +1,216 @@
+import { db, workflow as workflowTable } from '@sim/db'
+import { createLogger } from '@sim/logger'
+import { eq } from 'drizzle-orm'
+import { type NextRequest, NextResponse } from 'next/server'
+import { v4 as uuidv4 } from 'uuid'
+import { z } from 'zod'
+import { checkHybridAuth } from '@/lib/auth/hybrid'
+import { generateRequestId } from '@/lib/core/utils/request'
+import { SSE_HEADERS } from '@/lib/core/utils/sse'
+import { markExecutionCancelled } from '@/lib/execution/cancellation'
+import { LoggingSession } from '@/lib/logs/execution/logging-session'
+import { executeWorkflowCore } from '@/lib/workflows/executor/execution-core'
+import { createSSECallbacks } from '@/lib/workflows/executor/execution-events'
+import { ExecutionSnapshot } from '@/executor/execution/snapshot'
+import type { ExecutionMetadata, SerializableExecutionState } from '@/executor/execution/types'
+import { hasExecutionResult } from '@/executor/utils/errors'
+
+const logger = createLogger('ExecuteFromBlockAPI')
+
+const ExecuteFromBlockSchema = z.object({
+  startBlockId: z.string().min(1, 'Start block ID is required'),
+  sourceSnapshot: z.object({
+    blockStates: z.record(z.any()),
+    executedBlocks: z.array(z.string()),
+    blockLogs: z.array(z.any()),
+    decisions: z.object({
+      router: z.record(z.string()),
+      condition: z.record(z.string()),
+    }),
+    completedLoops: z.array(z.string()),
+    loopExecutions: z.record(z.any()).optional(),
+    parallelExecutions: z.record(z.any()).optional(),
+    parallelBlockMapping: z.record(z.any()).optional(),
+    activeExecutionPath: z.array(z.string()),
+  }),
+  input: z.any().optional(),
+})
+
+export const runtime = 'nodejs'
+export const dynamic = 'force-dynamic'
+
+export async function POST(req: NextRequest, { params }: { params: Promise<{ id: string }> }) {
+  const requestId = generateRequestId()
+  const { id: workflowId } = await params
+
+  try {
+    const auth = await checkHybridAuth(req, { requireWorkflowId: false })
+    if (!auth.success || !auth.userId) {
+      return NextResponse.json({ error: auth.error || 'Unauthorized' }, { status: 401 })
+    }
+    const userId = auth.userId
+
+    let body: unknown
+    try {
+      body = await req.json()
+    } catch {
+      return NextResponse.json({ error: 'Invalid JSON body' }, { status: 400 })
+    }
+
+    const validation = ExecuteFromBlockSchema.safeParse(body)
+    if (!validation.success) {
+      logger.warn(`[${requestId}] Invalid request body:`, validation.error.errors)
+      return NextResponse.json(
+        {
+          error: 'Invalid request body',
+          details: validation.error.errors.map((e) => ({
+            path: e.path.join('.'),
+            message: e.message,
+          })),
+        },
+        { status: 400 }
+      )
+    }
+
+    const { startBlockId, sourceSnapshot, input } = validation.data
+    const executionId = uuidv4()
+
+    const [workflowRecord] = await db
+      .select({ workspaceId: workflowTable.workspaceId, userId: workflowTable.userId })
+      .from(workflowTable)
+      .where(eq(workflowTable.id, workflowId))
+      .limit(1)
+
+    if (!workflowRecord?.workspaceId) {
+      return NextResponse.json({ error: 'Workflow not found or has no workspace' }, { status: 404 })
+    }
+
+    const workspaceId = workflowRecord.workspaceId
+    const workflowUserId = workflowRecord.userId
+
+    logger.info(`[${requestId}] Starting run-from-block execution`, {
+      workflowId,
+      startBlockId,
+      executedBlocksCount: sourceSnapshot.executedBlocks.length,
+    })
+
+    const loggingSession = new LoggingSession(workflowId, executionId, 'manual', requestId)
+    const abortController = new AbortController()
+    let isStreamClosed = false
+
+    const stream = new ReadableStream<Uint8Array>({
+      async start(controller) {
+        const { sendEvent, onBlockStart, onBlockComplete, onStream } = createSSECallbacks({
+          executionId,
+          workflowId,
+          controller,
+          isStreamClosed: () => isStreamClosed,
+          setStreamClosed: () => {
+            isStreamClosed = true
+          },
+        })
+
+        const metadata: ExecutionMetadata = {
+          requestId,
+          workflowId,
+          userId,
+          executionId,
+          triggerType: 'manual',
+          workspaceId,
+          workflowUserId,
+          useDraftState: true,
+          isClientSession: true,
+          startTime: new Date().toISOString(),
+        }
+
+        const snapshot = new ExecutionSnapshot(metadata, {}, input || {}, {})
+
+        try {
+          const startTime = new Date()
+
+          sendEvent({
+            type: 'execution:started',
+            timestamp: startTime.toISOString(),
+            executionId,
+            workflowId,
+            data: { startTime: startTime.toISOString() },
+          })
+
+          const result = await executeWorkflowCore({
+            snapshot,
+            loggingSession,
+            abortSignal: abortController.signal,
+            runFromBlock: {
+              startBlockId,
+              sourceSnapshot: sourceSnapshot as SerializableExecutionState,
+            },
+            callbacks: { onBlockStart, onBlockComplete, onStream },
+          })
+
+          if (result.status === 'cancelled') {
+            sendEvent({
+              type: 'execution:cancelled',
+              timestamp: new Date().toISOString(),
+              executionId,
+              workflowId,
+              data: { duration: result.metadata?.duration || 0 },
+            })
+          } else {
+            sendEvent({
+              type: 'execution:completed',
+              timestamp: new Date().toISOString(),
+              executionId,
+              workflowId,
+              data: {
+                success: result.success,
+                output: result.output,
+                duration: result.metadata?.duration || 0,
+                startTime: result.metadata?.startTime || startTime.toISOString(),
+                endTime: result.metadata?.endTime || new Date().toISOString(),
+              },
+            })
+          }
+        } catch (error: unknown) {
+          const errorMessage = error instanceof Error ? error.message : 'Unknown error'
+          logger.error(`[${requestId}] Run-from-block execution failed: ${errorMessage}`)
+
+          const executionResult = hasExecutionResult(error) ? error.executionResult : undefined
+
+          sendEvent({
+            type: 'execution:error',
+            timestamp: new Date().toISOString(),
+            executionId,
+            workflowId,
+            data: {
+              error: executionResult?.error || errorMessage,
+              duration: executionResult?.metadata?.duration || 0,
+            },
+          })
+        } finally {
+          if (!isStreamClosed) {
+            try {
+              controller.enqueue(new TextEncoder().encode('data: [DONE]\n\n'))
+              controller.close()
+            } catch {}
+          }
+        }
+      },
+      cancel() {
+        isStreamClosed = true
+        abortController.abort()
+        markExecutionCancelled(executionId).catch(() => {})
+      },
+    })
+
+    return new NextResponse(stream, {
+      headers: { ...SSE_HEADERS, 'X-Execution-Id': executionId },
+    })
+  } catch (error: unknown) {
+    const errorMessage = error instanceof Error ? error.message : 'Unknown error'
+    logger.error(`[${requestId}] Failed to start run-from-block execution:`, error)
+    return NextResponse.json(
+      { error: errorMessage || 'Failed to start execution' },
+      { status: 500 }
+    )
+  }
+}

apps/sim/app/api/workflows/[id]/execute/route.ts

@@ -53,6 +53,7 @@ const ExecuteWorkflowSchema = z.object({
       parallels: z.record(z.any()).optional(),
     })
     .optional(),
+  stopAfterBlockId: z.string().optional(),
 })
 
 export const runtime = 'nodejs'
@@ -222,6 +223,7 @@ export async function POST(req: NextRequest, { params }: { params: Promise<{ id:
       includeFileBase64,
       base64MaxBytes,
       workflowStateOverride,
+      stopAfterBlockId,
     } = validation.data
 
     // For API key and internal JWT auth, the entire body is the input (except for our control fields)
@@ -237,6 +239,7 @@ export async function POST(req: NextRequest, { params }: { params: Promise<{ id:
               includeFileBase64,
               base64MaxBytes,
               workflowStateOverride,
+              stopAfterBlockId: _stopAfterBlockId,
               workflowId: _workflowId, // Also exclude workflowId used for internal JWT auth
               ...rest
             } = body
@@ -434,6 +437,7 @@ export async function POST(req: NextRequest, { params }: { params: Promise<{ id:
           loggingSession,
           includeFileBase64,
           base64MaxBytes,
+          stopAfterBlockId,
         })
 
         const outputWithBase64 = includeFileBase64
@@ -722,6 +726,7 @@ export async function POST(req: NextRequest, { params }: { params: Promise<{ id:
             abortSignal: abortController.signal,
             includeFileBase64,
             base64MaxBytes,
+            stopAfterBlockId,
           })
 
           if (result.status === 'paused') {

apps/sim/app/api/yaml/autolayout/route.ts

@@ -0,0 +1,108 @@
+import { createLogger } from '@sim/logger'
+import { type NextRequest, NextResponse } from 'next/server'
+import { z } from 'zod'
+import { generateRequestId } from '@/lib/core/utils/request'
+import { applyAutoLayout } from '@/lib/workflows/autolayout'
+import {
+  DEFAULT_HORIZONTAL_SPACING,
+  DEFAULT_LAYOUT_PADDING,
+  DEFAULT_VERTICAL_SPACING,
+} from '@/lib/workflows/autolayout/constants'
+
+const logger = createLogger('YamlAutoLayoutAPI')
+
+const AutoLayoutRequestSchema = z.object({
+  workflowState: z.object({
+    blocks: z.record(z.any()),
+    edges: z.array(z.any()),
+    loops: z.record(z.any()).optional().default({}),
+    parallels: z.record(z.any()).optional().default({}),
+  }),
+  options: z
+    .object({
+      spacing: z
+        .object({
+          horizontal: z.number().optional(),
+          vertical: z.number().optional(),
+        })
+        .optional(),
+      alignment: z.enum(['start', 'center', 'end']).optional(),
+      padding: z
+        .object({
+          x: z.number().optional(),
+          y: z.number().optional(),
+        })
+        .optional(),
+    })
+    .optional(),
+})
+
+export async function POST(request: NextRequest) {
+  const requestId = generateRequestId()
+
+  try {
+    const body = await request.json()
+    const { workflowState, options } = AutoLayoutRequestSchema.parse(body)
+
+    logger.info(`[${requestId}] Applying auto layout`, {
+      blockCount: Object.keys(workflowState.blocks).length,
+      edgeCount: workflowState.edges.length,
+    })
+
+    const autoLayoutOptions = {
+      horizontalSpacing: options?.spacing?.horizontal ?? DEFAULT_HORIZONTAL_SPACING,
+      verticalSpacing: options?.spacing?.vertical ?? DEFAULT_VERTICAL_SPACING,
+      padding: {
+        x: options?.padding?.x ?? DEFAULT_LAYOUT_PADDING.x,
+        y: options?.padding?.y ?? DEFAULT_LAYOUT_PADDING.y,
+      },
+      alignment: options?.alignment ?? 'center',
+    }
+
+    const layoutResult = applyAutoLayout(
+      workflowState.blocks,
+      workflowState.edges,
+      autoLayoutOptions
+    )
+
+    if (!layoutResult.success || !layoutResult.blocks) {
+      logger.error(`[${requestId}] Auto layout failed:`, {
+        error: layoutResult.error,
+      })
+      return NextResponse.json(
+        {
+          success: false,
+          errors: [layoutResult.error || 'Unknown auto layout error'],
+        },
+        { status: 500 }
+      )
+    }
+
+    logger.info(`[${requestId}] Auto layout completed successfully:`, {
+      success: true,
+      blockCount: Object.keys(layoutResult.blocks).length,
+    })
+
+    const transformedResponse = {
+      success: true,
+      workflowState: {
+        blocks: layoutResult.blocks,
+        edges: workflowState.edges,
+        loops: workflowState.loops || {},
+        parallels: workflowState.parallels || {},
+      },
+    }
+
+    return NextResponse.json(transformedResponse)
+  } catch (error) {
+    logger.error(`[${requestId}] Auto layout failed:`, error)
+
+    return NextResponse.json(
+      {
+        success: false,
+        errors: [error instanceof Error ? error.message : 'Unknown auto layout error'],
+      },
+      { status: 500 }
+    )
+  }
+}

apps/sim/app/workspace/[workspaceId]/w/[workflowId]/components/action-bar/action-bar.tsx

@@ -1,11 +1,13 @@
 import { memo, useCallback } from 'react'
-import { ArrowLeftRight, ArrowUpDown, Circle, CircleOff, LogOut } from 'lucide-react'
+import { ArrowLeftRight, ArrowUpDown, Circle, CircleOff, LogOut, Play } from 'lucide-react'
 import { Button, Copy, Tooltip, Trash2 } from '@/components/emcn'
 import { cn } from '@/lib/core/utils/cn'
 import { isInputDefinitionTrigger } from '@/lib/workflows/triggers/input-definition-triggers'
 import { useUserPermissionsContext } from '@/app/workspace/[workspaceId]/providers/workspace-permissions-provider'
+import { useWorkflowExecution } from '@/app/workspace/[workspaceId]/w/[workflowId]/hooks'
 import { validateTriggerPaste } from '@/app/workspace/[workspaceId]/w/[workflowId]/utils'
 import { useCollaborativeWorkflow } from '@/hooks/use-collaborative-workflow'
+import { useExecutionStore } from '@/stores/execution'
 import { useNotificationStore } from '@/stores/notifications'
 import { useWorkflowRegistry } from '@/stores/workflows/registry/store'
 import { useWorkflowStore } from '@/stores/workflows/workflow/store'
@@ -49,6 +51,7 @@ export const ActionBar = memo(
       collaborativeBatchToggleBlockHandles,
     } = useCollaborativeWorkflow()
     const { setPendingSelection } = useWorkflowRegistry()
+    const { handleRunFromBlock } = useWorkflowExecution()
 
     const addNotification = useNotificationStore((s) => s.addNotification)
 
@@ -97,12 +100,39 @@ export const ActionBar = memo(
       )
     )
 
+    const { activeWorkflowId } = useWorkflowRegistry()
+    const { isExecuting, getLastExecutionSnapshot } = useExecutionStore()
     const userPermissions = useUserPermissionsContext()
+    const edges = useWorkflowStore((state) => state.edges)
 
     const isStartBlock = isInputDefinitionTrigger(blockType)
     const isResponseBlock = blockType === 'response'
     const isNoteBlock = blockType === 'note'
     const isSubflowBlock = blockType === 'loop' || blockType === 'parallel'
+    const isInsideSubflow = parentId && (parentType === 'loop' || parentType === 'parallel')
+
+    const snapshot = activeWorkflowId ? getLastExecutionSnapshot(activeWorkflowId) : null
+    const incomingEdges = edges.filter((edge) => edge.target === blockId)
+    const isTriggerBlock = incomingEdges.length === 0
+
+    // Check if each source block is either executed OR is a trigger block (triggers don't need prior execution)
+    const isSourceSatisfied = (sourceId: string) => {
+      if (snapshot?.executedBlocks.includes(sourceId)) return true
+      // Check if source is a trigger (has no incoming edges itself)
+      const sourceIncomingEdges = edges.filter((edge) => edge.target === sourceId)
+      return sourceIncomingEdges.length === 0
+    }
+
+    // Non-trigger blocks need a snapshot to exist (so upstream outputs are available)
+    const dependenciesSatisfied =
+      isTriggerBlock || (snapshot && incomingEdges.every((edge) => isSourceSatisfied(edge.source)))
+    const canRunFromBlock =
+      dependenciesSatisfied && !isNoteBlock && !isInsideSubflow && !isExecuting
+
+    const handleRunFromBlockClick = useCallback(() => {
+      if (!activeWorkflowId || !canRunFromBlock) return
+      handleRunFromBlock(blockId, activeWorkflowId)
+    }, [blockId, activeWorkflowId, canRunFromBlock, handleRunFromBlock])
 
     /**
      * Get appropriate tooltip message based on disabled state
@@ -128,30 +158,35 @@ export const ActionBar = memo(
           'dark:border-transparent dark:bg-[var(--surface-4)]'
         )}
       >
-        {!isNoteBlock && (
+        {!isNoteBlock && !isInsideSubflow && (
           <Tooltip.Root>
             <Tooltip.Trigger asChild>
               <Button
                 variant='ghost'
                 onClick={(e) => {
                   e.stopPropagation()
-                  if (!disabled) {
-                    collaborativeBatchToggleBlockEnabled([blockId])
+                  if (canRunFromBlock && !disabled) {
+                    handleRunFromBlockClick()
                   }
                 }}
                 className={ACTION_BUTTON_STYLES}
-                disabled={disabled}
+                disabled={disabled || !canRunFromBlock}
               >
-                {isEnabled ? <Circle className={ICON_SIZE} /> : <CircleOff className={ICON_SIZE} />}
+                <Play className={ICON_SIZE} />
               </Button>
             </Tooltip.Trigger>
             <Tooltip.Content side='top'>
-              {getTooltipMessage(isEnabled ? 'Disable Block' : 'Enable Block')}
+              {(() => {
+                if (disabled) return getTooltipMessage('Run from block')
+                if (isExecuting) return 'Execution in progress'
+                if (!dependenciesSatisfied) return 'Run upstream blocks first'
+                return 'Run from block'
+              })()}
             </Tooltip.Content>
           </Tooltip.Root>
         )}
 
-        {isSubflowBlock && (
+        {!isNoteBlock && (
           <Tooltip.Root>
             <Tooltip.Trigger asChild>
               <Button

apps/sim/app/workspace/[workspaceId]/w/[workflowId]/components/block-menu/block-menu.tsx

@@ -40,9 +40,16 @@ export interface BlockMenuProps {
   onRemoveFromSubflow: () => void
   onOpenEditor: () => void
   onRename: () => void
+  onRunFromBlock?: () => void
+  onRunUntilBlock?: () => void
   hasClipboard?: boolean
   showRemoveFromSubflow?: boolean
+  /** Whether run from block is available (has snapshot, was executed, not inside subflow) */
+  canRunFromBlock?: boolean
   disableEdit?: boolean
+  isExecuting?: boolean
+  /** Whether the selected block is a trigger (has no incoming edges) */
+  isPositionalTrigger?: boolean
 }
 
 /**
@@ -65,9 +72,14 @@ export function BlockMenu({
   onRemoveFromSubflow,
   onOpenEditor,
   onRename,
+  onRunFromBlock,
+  onRunUntilBlock,
   hasClipboard = false,
   showRemoveFromSubflow = false,
+  canRunFromBlock = false,
   disableEdit = false,
+  isExecuting = false,
+  isPositionalTrigger = false,
 }: BlockMenuProps) {
   const isSingleBlock = selectedBlocks.length === 1
 
@@ -78,10 +90,15 @@ export function BlockMenu({
     (b) =>
       TriggerUtils.requiresSingleInstance(b.type) || TriggerUtils.isSingleInstanceBlockType(b.type)
   )
-  const hasTriggerBlock = selectedBlocks.some((b) => TriggerUtils.isTriggerBlock(b))
+  // A block is a trigger if it's explicitly a trigger type OR has no incoming edges (positional trigger)
+  const hasTriggerBlock =
+    selectedBlocks.some((b) => TriggerUtils.isTriggerBlock(b)) || isPositionalTrigger
   const allNoteBlocks = selectedBlocks.every((b) => b.type === 'note')
   const isSubflow =
     isSingleBlock && (selectedBlocks[0]?.type === 'loop' || selectedBlocks[0]?.type === 'parallel')
+  const isInsideSubflow =
+    isSingleBlock &&
+    (selectedBlocks[0]?.parentType === 'loop' || selectedBlocks[0]?.parentType === 'parallel')
 
   const canRemoveFromSubflow = showRemoveFromSubflow && !hasTriggerBlock
 
@@ -203,6 +220,38 @@ export function BlockMenu({
           </PopoverItem>
         )}
 
+        {/* Run from/until block - only for single non-note block, not inside subflows */}
+        {isSingleBlock && !allNoteBlocks && !isInsideSubflow && (
+          <>
+            <PopoverDivider />
+            <PopoverItem
+              disabled={!canRunFromBlock || isExecuting}
+              onClick={() => {
+                if (canRunFromBlock && !isExecuting) {
+                  onRunFromBlock?.()
+                  onClose()
+                }
+              }}
+            >
+              Run from block
+            </PopoverItem>
+            {/* Hide "Run until" for triggers - they're always at the start */}
+            {!hasTriggerBlock && (
+              <PopoverItem
+                disabled={isExecuting}
+                onClick={() => {
+                  if (!isExecuting) {
+                    onRunUntilBlock?.()
+                    onClose()
+                  }
+                }}
+              >
+                Run until block
+              </PopoverItem>
+            )}
+          </>
+        )}
+
         {/* Destructive action */}
         <PopoverDivider />
         <PopoverItem

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

@@ -3,7 +3,6 @@ import { createLogger } from '@sim/logger'
 import { useReactFlow } from 'reactflow'
 import type { AutoLayoutOptions } from '@/app/workspace/[workspaceId]/w/[workflowId]/utils/auto-layout-utils'
 import { applyAutoLayoutAndUpdateStore as applyAutoLayoutStandalone } from '@/app/workspace/[workspaceId]/w/[workflowId]/utils/auto-layout-utils'
-import { useSnapToGridSize } from '@/hooks/queries/general-settings'
 import { useCanvasViewport } from '@/hooks/use-canvas-viewport'
 
 export type { AutoLayoutOptions }
@@ -14,28 +13,21 @@ const logger = createLogger('useAutoLayout')
  * Hook providing auto-layout functionality for workflows.
  * Binds workflowId context and provides memoized callback for React components.
  * Includes automatic fitView animation after successful layout.
- * Automatically uses the user's snap-to-grid setting for grid-aligned layout.
  *
  * Note: This hook requires a ReactFlowProvider ancestor.
  */
 export function useAutoLayout(workflowId: string | null) {
   const reactFlowInstance = useReactFlow()
   const { fitViewToBounds } = useCanvasViewport(reactFlowInstance)
-  const snapToGridSize = useSnapToGridSize()
 
   const applyAutoLayoutAndUpdateStore = useCallback(
     async (options: AutoLayoutOptions = {}) => {
       if (!workflowId) {
         return { success: false, error: 'No workflow ID provided' }
       }
-      // Include gridSize from user's snap-to-grid setting
-      const optionsWithGrid: AutoLayoutOptions = {
-        ...options,
-        gridSize: options.gridSize ?? (snapToGridSize > 0 ? snapToGridSize : undefined),
-      }
-      return applyAutoLayoutStandalone(workflowId, optionsWithGrid)
+      return applyAutoLayoutStandalone(workflowId, options)
     },
-    [workflowId, snapToGridSize]
+    [workflowId]
   )
 
   /**

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

@@ -15,13 +15,16 @@ import {
   TriggerUtils,
 } from '@/lib/workflows/triggers/triggers'
 import { useCurrentWorkflow } from '@/app/workspace/[workspaceId]/w/[workflowId]/hooks/use-current-workflow'
-import type { BlockLog, ExecutionResult, StreamingExecution } from '@/executor/types'
+import { getBlock } from '@/blocks'
+import type { SerializableExecutionState } from '@/executor/execution/types'
+import type { BlockLog, BlockState, ExecutionResult, StreamingExecution } from '@/executor/types'
 import { hasExecutionResult } from '@/executor/utils/errors'
 import { coerceValue } from '@/executor/utils/start-block'
 import { subscriptionKeys } from '@/hooks/queries/subscription'
 import { useExecutionStream } from '@/hooks/use-execution-stream'
 import { WorkflowValidationError } from '@/serializer'
 import { useExecutionStore } from '@/stores/execution'
+import { useNotificationStore } from '@/stores/notifications'
 import { useVariablesStore } from '@/stores/panel'
 import { useEnvironmentStore } from '@/stores/settings/environment'
 import { type ConsoleEntry, useTerminalConsoleStore } from '@/stores/terminal'
@@ -98,11 +101,15 @@ export function useWorkflowExecution() {
     setActiveBlocks,
     setBlockRunStatus,
     setEdgeRunStatus,
+    setLastExecutionSnapshot,
+    getLastExecutionSnapshot,
+    clearLastExecutionSnapshot,
   } = useExecutionStore()
   const [executionResult, setExecutionResult] = useState<ExecutionResult | null>(null)
   const executionStream = useExecutionStream()
   const currentChatExecutionIdRef = useRef<string | null>(null)
   const isViewingDiff = useWorkflowDiffStore((state) => state.isShowingDiff)
+  const addNotification = useNotificationStore((state) => state.addNotification)
 
   /**
    * Validates debug state before performing debug operations
@@ -668,7 +675,8 @@ export function useWorkflowExecution() {
     onStream?: (se: StreamingExecution) => Promise<void>,
     executionId?: string,
     onBlockComplete?: (blockId: string, output: any) => Promise<void>,
-    overrideTriggerType?: 'chat' | 'manual' | 'api'
+    overrideTriggerType?: 'chat' | 'manual' | 'api',
+    stopAfterBlockId?: string
   ): Promise<ExecutionResult | StreamingExecution> => {
     // Use diff workflow for execution when available, regardless of canvas view state
     const executionWorkflowState = null as {
@@ -876,6 +884,8 @@ export function useWorkflowExecution() {
       const activeBlocksSet = new Set<string>()
       const streamedContent = new Map<string, string>()
       const accumulatedBlockLogs: BlockLog[] = []
+      const accumulatedBlockStates = new Map<string, BlockState>()
+      const executedBlockIds = new Set<string>()
 
       // Execute the workflow
       try {
@@ -887,6 +897,7 @@ export function useWorkflowExecution() {
           triggerType: overrideTriggerType || 'manual',
           useDraftState: true,
           isClientSession: true,
+          stopAfterBlockId,
           workflowStateOverride: executionWorkflowState
             ? {
                 blocks: executionWorkflowState.blocks,
@@ -916,18 +927,22 @@ export function useWorkflowExecution() {
               logger.info('onBlockCompleted received:', { data })
 
               activeBlocksSet.delete(data.blockId)
-              // Create a new Set to trigger React re-render
               setActiveBlocks(new Set(activeBlocksSet))
-
-              // Track successful block execution in run path
               setBlockRunStatus(data.blockId, 'success')
 
-              // Edges already tracked in onBlockStarted, no need to track again
+              executedBlockIds.add(data.blockId)
+              accumulatedBlockStates.set(data.blockId, {
+                output: data.output,
+                executed: true,
+                executionTime: data.durationMs,
+              })
+
+              const isContainerBlock = data.blockType === 'loop' || data.blockType === 'parallel'
+              if (isContainerBlock) return
 
               const startedAt = new Date(Date.now() - data.durationMs).toISOString()
               const endedAt = new Date().toISOString()
 
-              // Accumulate block log for the execution result
               accumulatedBlockLogs.push({
                 blockId: data.blockId,
                 blockName: data.blockName || 'Unknown Block',
@@ -1056,6 +1071,53 @@ export function useWorkflowExecution() {
                 },
                 logs: accumulatedBlockLogs,
               }
+
+              // Add trigger block to executed blocks so downstream blocks can use run-from-block
+              if (data.success && startBlockId) {
+                executedBlockIds.add(startBlockId)
+              }
+
+              if (data.success && activeWorkflowId) {
+                if (stopAfterBlockId) {
+                  const existingSnapshot = getLastExecutionSnapshot(activeWorkflowId)
+                  const mergedBlockStates = {
+                    ...(existingSnapshot?.blockStates || {}),
+                    ...Object.fromEntries(accumulatedBlockStates),
+                  }
+                  const mergedExecutedBlocks = new Set([
+                    ...(existingSnapshot?.executedBlocks || []),
+                    ...executedBlockIds,
+                  ])
+                  const snapshot: SerializableExecutionState = {
+                    blockStates: mergedBlockStates,
+                    executedBlocks: Array.from(mergedExecutedBlocks),
+                    blockLogs: [...(existingSnapshot?.blockLogs || []), ...accumulatedBlockLogs],
+                    decisions: existingSnapshot?.decisions || { router: {}, condition: {} },
+                    completedLoops: existingSnapshot?.completedLoops || [],
+                    activeExecutionPath: Array.from(mergedExecutedBlocks),
+                  }
+                  setLastExecutionSnapshot(activeWorkflowId, snapshot)
+                  logger.info('Merged execution snapshot after run-until-block', {
+                    workflowId: activeWorkflowId,
+                    newBlocksExecuted: executedBlockIds.size,
+                    totalExecutedBlocks: mergedExecutedBlocks.size,
+                  })
+                } else {
+                  const snapshot: SerializableExecutionState = {
+                    blockStates: Object.fromEntries(accumulatedBlockStates),
+                    executedBlocks: Array.from(executedBlockIds),
+                    blockLogs: accumulatedBlockLogs,
+                    decisions: { router: {}, condition: {} },
+                    completedLoops: [],
+                    activeExecutionPath: Array.from(executedBlockIds),
+                  }
+                  setLastExecutionSnapshot(activeWorkflowId, snapshot)
+                  logger.info('Stored execution snapshot for run-from-block', {
+                    workflowId: activeWorkflowId,
+                    executedBlocksCount: executedBlockIds.size,
+                  })
+                }
+              }
             },
 
             onExecutionError: (data) => {
@@ -1376,6 +1438,330 @@ export function useWorkflowExecution() {
     setActiveBlocks,
   ])
 
+  /**
+   * Handles running workflow from a specific block using cached outputs
+   */
+  const handleRunFromBlock = useCallback(
+    async (blockId: string, workflowId: string) => {
+      const snapshot = getLastExecutionSnapshot(workflowId)
+      const workflowEdges = useWorkflowStore.getState().edges
+      const incomingEdges = workflowEdges.filter((edge) => edge.target === blockId)
+      const isTriggerBlock = incomingEdges.length === 0
+
+      // Check if each source block is either executed OR is a trigger block (triggers don't need prior execution)
+      const isSourceSatisfied = (sourceId: string) => {
+        if (snapshot?.executedBlocks.includes(sourceId)) return true
+        // Check if source is a trigger (has no incoming edges itself)
+        const sourceIncomingEdges = workflowEdges.filter((edge) => edge.target === sourceId)
+        return sourceIncomingEdges.length === 0
+      }
+
+      // Non-trigger blocks need a snapshot to exist (so upstream outputs are available)
+      if (!snapshot && !isTriggerBlock) {
+        logger.error('No execution snapshot available for run-from-block', { workflowId, blockId })
+        return
+      }
+
+      const dependenciesSatisfied =
+        isTriggerBlock || incomingEdges.every((edge) => isSourceSatisfied(edge.source))
+
+      if (!dependenciesSatisfied) {
+        logger.error('Upstream dependencies not satisfied for run-from-block', {
+          workflowId,
+          blockId,
+        })
+        return
+      }
+
+      // For trigger blocks, always use empty snapshot to prevent stale data from different
+      // execution paths from being resolved. For non-trigger blocks, use the existing snapshot.
+      const emptySnapshot: SerializableExecutionState = {
+        blockStates: {},
+        executedBlocks: [],
+        blockLogs: [],
+        decisions: { router: {}, condition: {} },
+        completedLoops: [],
+        activeExecutionPath: [],
+      }
+      const effectiveSnapshot: SerializableExecutionState = isTriggerBlock
+        ? emptySnapshot
+        : snapshot || emptySnapshot
+
+      // Extract mock payload for trigger blocks
+      let workflowInput: any
+      if (isTriggerBlock) {
+        const workflowBlocks = useWorkflowStore.getState().blocks
+        const mergedStates = mergeSubblockState(workflowBlocks, workflowId)
+        const candidates = resolveStartCandidates(mergedStates, { execution: 'manual' })
+        const candidate = candidates.find((c) => c.blockId === blockId)
+
+        if (candidate) {
+          if (triggerNeedsMockPayload(candidate)) {
+            workflowInput = extractTriggerMockPayload(candidate)
+          } else if (
+            candidate.path === StartBlockPath.SPLIT_API ||
+            candidate.path === StartBlockPath.SPLIT_INPUT ||
+            candidate.path === StartBlockPath.UNIFIED
+          ) {
+            const inputFormatValue = candidate.block.subBlocks?.inputFormat?.value
+            if (Array.isArray(inputFormatValue)) {
+              const testInput: Record<string, any> = {}
+              inputFormatValue.forEach((field: any) => {
+                if (field && typeof field === 'object' && field.name && field.value !== undefined) {
+                  testInput[field.name] = coerceValue(field.type, field.value)
+                }
+              })
+              if (Object.keys(testInput).length > 0) {
+                workflowInput = testInput
+              }
+            }
+          }
+        } else {
+          // Fallback: block is trigger by position but not classified as start candidate
+          const block = mergedStates[blockId]
+          if (block) {
+            const blockConfig = getBlock(block.type)
+            const hasTriggers = blockConfig?.triggers?.available?.length
+
+            if (hasTriggers || block.triggerMode) {
+              workflowInput = extractTriggerMockPayload({
+                blockId,
+                block,
+                path: StartBlockPath.EXTERNAL_TRIGGER,
+              })
+            }
+          }
+        }
+      }
+
+      setIsExecuting(true)
+      const executionId = uuidv4()
+      const accumulatedBlockLogs: BlockLog[] = []
+      const accumulatedBlockStates = new Map<string, BlockState>()
+      const executedBlockIds = new Set<string>()
+      const activeBlocksSet = new Set<string>()
+
+      try {
+        await executionStream.executeFromBlock({
+          workflowId,
+          startBlockId: blockId,
+          sourceSnapshot: effectiveSnapshot,
+          input: workflowInput,
+          callbacks: {
+            onBlockStarted: (data) => {
+              activeBlocksSet.add(data.blockId)
+              setActiveBlocks(new Set(activeBlocksSet))
+
+              const incomingEdges = workflowEdges.filter((edge) => edge.target === data.blockId)
+              incomingEdges.forEach((edge) => {
+                setEdgeRunStatus(edge.id, 'success')
+              })
+            },
+
+            onBlockCompleted: (data) => {
+              activeBlocksSet.delete(data.blockId)
+              setActiveBlocks(new Set(activeBlocksSet))
+
+              setBlockRunStatus(data.blockId, 'success')
+
+              executedBlockIds.add(data.blockId)
+              accumulatedBlockStates.set(data.blockId, {
+                output: data.output,
+                executed: true,
+                executionTime: data.durationMs,
+              })
+
+              const isContainerBlock = data.blockType === 'loop' || data.blockType === 'parallel'
+              if (isContainerBlock) return
+
+              const startedAt = new Date(Date.now() - data.durationMs).toISOString()
+              const endedAt = new Date().toISOString()
+
+              accumulatedBlockLogs.push({
+                blockId: data.blockId,
+                blockName: data.blockName || 'Unknown Block',
+                blockType: data.blockType || 'unknown',
+                input: data.input || {},
+                output: data.output,
+                success: true,
+                durationMs: data.durationMs,
+                startedAt,
+                endedAt,
+              })
+
+              addConsole({
+                input: data.input || {},
+                output: data.output,
+                success: true,
+                durationMs: data.durationMs,
+                startedAt,
+                endedAt,
+                workflowId,
+                blockId: data.blockId,
+                executionId,
+                blockName: data.blockName || 'Unknown Block',
+                blockType: data.blockType || 'unknown',
+                iterationCurrent: data.iterationCurrent,
+                iterationTotal: data.iterationTotal,
+                iterationType: data.iterationType,
+              })
+            },
+
+            onBlockError: (data) => {
+              activeBlocksSet.delete(data.blockId)
+              setActiveBlocks(new Set(activeBlocksSet))
+
+              setBlockRunStatus(data.blockId, 'error')
+
+              const startedAt = new Date(Date.now() - data.durationMs).toISOString()
+              const endedAt = new Date().toISOString()
+
+              accumulatedBlockLogs.push({
+                blockId: data.blockId,
+                blockName: data.blockName || 'Unknown Block',
+                blockType: data.blockType || 'unknown',
+                input: data.input || {},
+                output: {},
+                success: false,
+                error: data.error,
+                durationMs: data.durationMs,
+                startedAt,
+                endedAt,
+              })
+
+              addConsole({
+                input: data.input || {},
+                output: {},
+                success: false,
+                error: data.error,
+                durationMs: data.durationMs,
+                startedAt,
+                endedAt,
+                workflowId,
+                blockId: data.blockId,
+                executionId,
+                blockName: data.blockName,
+                blockType: data.blockType,
+                iterationCurrent: data.iterationCurrent,
+                iterationTotal: data.iterationTotal,
+                iterationType: data.iterationType,
+              })
+            },
+
+            onExecutionCompleted: (data) => {
+              if (data.success) {
+                // Add the start block (trigger) to executed blocks
+                executedBlockIds.add(blockId)
+
+                const mergedBlockStates: Record<string, BlockState> = {
+                  ...effectiveSnapshot.blockStates,
+                }
+                for (const [bId, state] of accumulatedBlockStates) {
+                  mergedBlockStates[bId] = state
+                }
+
+                const mergedExecutedBlocks = new Set([
+                  ...effectiveSnapshot.executedBlocks,
+                  ...executedBlockIds,
+                ])
+
+                const updatedSnapshot: SerializableExecutionState = {
+                  ...effectiveSnapshot,
+                  blockStates: mergedBlockStates,
+                  executedBlocks: Array.from(mergedExecutedBlocks),
+                  blockLogs: [...effectiveSnapshot.blockLogs, ...accumulatedBlockLogs],
+                  activeExecutionPath: Array.from(mergedExecutedBlocks),
+                }
+                setLastExecutionSnapshot(workflowId, updatedSnapshot)
+              }
+            },
+
+            onExecutionError: (data) => {
+              const isWorkflowModified =
+                data.error?.includes('Block not found in workflow') ||
+                data.error?.includes('Upstream dependency not executed')
+
+              if (isWorkflowModified) {
+                clearLastExecutionSnapshot(workflowId)
+                addNotification({
+                  level: 'error',
+                  message:
+                    'Workflow was modified. Run the workflow again to enable running from block.',
+                  workflowId,
+                })
+              } else {
+                addNotification({
+                  level: 'error',
+                  message: data.error || 'Run from block failed',
+                  workflowId,
+                })
+              }
+            },
+          },
+        })
+      } catch (error) {
+        if ((error as Error).name !== 'AbortError') {
+          logger.error('Run-from-block failed:', error)
+        }
+      } finally {
+        setIsExecuting(false)
+        setActiveBlocks(new Set())
+      }
+    },
+    [
+      getLastExecutionSnapshot,
+      setLastExecutionSnapshot,
+      clearLastExecutionSnapshot,
+      setIsExecuting,
+      setActiveBlocks,
+      setBlockRunStatus,
+      setEdgeRunStatus,
+      addNotification,
+      addConsole,
+      executionStream,
+    ]
+  )
+
+  /**
+   * Handles running workflow until a specific block (stops after that block completes)
+   */
+  const handleRunUntilBlock = useCallback(
+    async (blockId: string, workflowId: string) => {
+      if (!workflowId || workflowId !== activeWorkflowId) {
+        logger.error('Invalid workflow ID for run-until-block', { workflowId, activeWorkflowId })
+        return
+      }
+
+      logger.info('Starting run-until-block execution', { workflowId, stopAfterBlockId: blockId })
+
+      setExecutionResult(null)
+      setIsExecuting(true)
+
+      const executionId = uuidv4()
+      try {
+        const result = await executeWorkflow(
+          undefined,
+          undefined,
+          executionId,
+          undefined,
+          'manual',
+          blockId
+        )
+        if (result && 'success' in result) {
+          setExecutionResult(result)
+        }
+      } catch (error) {
+        const errorResult = handleExecutionError(error, { executionId })
+        return errorResult
+      } finally {
+        setIsExecuting(false)
+        setIsDebugging(false)
+        setActiveBlocks(new Set())
+      }
+    },
+    [activeWorkflowId, setExecutionResult, setIsExecuting, setIsDebugging, setActiveBlocks]
+  )
+
   return {
     isExecuting,
     isDebugging,
@@ -1386,5 +1772,7 @@ export function useWorkflowExecution() {
     handleResumeDebug,
     handleCancelDebug,
     handleCancelExecution,
+    handleRunFromBlock,
+    handleRunUntilBlock,
   }
 }

apps/sim/app/workspace/[workspaceId]/w/[workflowId]/utils/auto-layout-utils.ts

@@ -21,7 +21,6 @@ export interface AutoLayoutOptions {
     x?: number
     y?: number
   }
-  gridSize?: number
 }
 
 /**
@@ -63,7 +62,6 @@ export async function applyAutoLayoutAndUpdateStore(
         x: options.padding?.x ?? DEFAULT_LAYOUT_PADDING.x,
         y: options.padding?.y ?? DEFAULT_LAYOUT_PADDING.y,
       },
-      gridSize: options.gridSize,
     }
 
     // Call the autolayout API route

apps/sim/app/workspace/[workspaceId]/w/[workflowId]/workflow.tsx

@@ -47,6 +47,7 @@ import {
   useCurrentWorkflow,
   useNodeUtilities,
   useShiftSelectionLock,
+  useWorkflowExecution,
 } from '@/app/workspace/[workspaceId]/w/[workflowId]/hooks'
 import {
   calculateContainerDimensions,
@@ -302,6 +303,8 @@ const WorkflowContent = React.memo(() => {
 
   const showTrainingModal = useCopilotTrainingStore((state) => state.showModal)
 
+  const { handleRunFromBlock, handleRunUntilBlock } = useWorkflowExecution()
+
   const snapToGridSize = useSnapToGridSize()
   const snapToGrid = snapToGridSize > 0
 
@@ -733,13 +736,16 @@ const WorkflowContent = React.memo(() => {
     [collaborativeBatchAddBlocks, setSelectedEdges, setPendingSelection]
   )
 
-  const { activeBlockIds, pendingBlocks, isDebugging } = useExecutionStore(
-    useShallow((state) => ({
-      activeBlockIds: state.activeBlockIds,
-      pendingBlocks: state.pendingBlocks,
-      isDebugging: state.isDebugging,
-    }))
-  )
+  const { activeBlockIds, pendingBlocks, isDebugging, isExecuting, getLastExecutionSnapshot } =
+    useExecutionStore(
+      useShallow((state) => ({
+        activeBlockIds: state.activeBlockIds,
+        pendingBlocks: state.pendingBlocks,
+        isDebugging: state.isDebugging,
+        isExecuting: state.isExecuting,
+        getLastExecutionSnapshot: state.getLastExecutionSnapshot,
+      }))
+    )
 
   const [dragStartParentId, setDragStartParentId] = useState<string | null>(null)
 
@@ -1102,6 +1108,50 @@ const WorkflowContent = React.memo(() => {
     }
   }, [contextMenuBlocks])
 
+  const handleContextRunFromBlock = useCallback(() => {
+    if (contextMenuBlocks.length !== 1) return
+    const blockId = contextMenuBlocks[0].id
+    handleRunFromBlock(blockId, workflowIdParam)
+  }, [contextMenuBlocks, workflowIdParam, handleRunFromBlock])
+
+  const handleContextRunUntilBlock = useCallback(() => {
+    if (contextMenuBlocks.length !== 1) return
+    const blockId = contextMenuBlocks[0].id
+    handleRunUntilBlock(blockId, workflowIdParam)
+  }, [contextMenuBlocks, workflowIdParam, handleRunUntilBlock])
+
+  const runFromBlockState = useMemo(() => {
+    if (contextMenuBlocks.length !== 1) {
+      return { canRun: false, reason: undefined }
+    }
+    const block = contextMenuBlocks[0]
+    const snapshot = getLastExecutionSnapshot(workflowIdParam)
+    const incomingEdges = edges.filter((edge) => edge.target === block.id)
+    const isTriggerBlock = incomingEdges.length === 0
+
+    // Check if each source block is either executed OR is a trigger block (triggers don't need prior execution)
+    const isSourceSatisfied = (sourceId: string) => {
+      if (snapshot?.executedBlocks.includes(sourceId)) return true
+      // Check if source is a trigger (has no incoming edges itself)
+      const sourceIncomingEdges = edges.filter((edge) => edge.target === sourceId)
+      return sourceIncomingEdges.length === 0
+    }
+
+    // Non-trigger blocks need a snapshot to exist (so upstream outputs are available)
+    const dependenciesSatisfied =
+      isTriggerBlock || (snapshot && incomingEdges.every((edge) => isSourceSatisfied(edge.source)))
+    const isNoteBlock = block.type === 'note'
+    const isInsideSubflow =
+      block.parentId && (block.parentType === 'loop' || block.parentType === 'parallel')
+
+    if (isInsideSubflow) return { canRun: false, reason: 'Cannot run from inside subflow' }
+    if (!dependenciesSatisfied) return { canRun: false, reason: 'Run upstream blocks first' }
+    if (isNoteBlock) return { canRun: false, reason: undefined }
+    if (isExecuting) return { canRun: false, reason: undefined }
+
+    return { canRun: true, reason: undefined }
+  }, [contextMenuBlocks, edges, workflowIdParam, getLastExecutionSnapshot, isExecuting])
+
   const handleContextAddBlock = useCallback(() => {
     useSearchModalStore.getState().open()
   }, [])
@@ -3418,11 +3468,19 @@ const WorkflowContent = React.memo(() => {
               onRemoveFromSubflow={handleContextRemoveFromSubflow}
               onOpenEditor={handleContextOpenEditor}
               onRename={handleContextRename}
+              onRunFromBlock={handleContextRunFromBlock}
+              onRunUntilBlock={handleContextRunUntilBlock}
               hasClipboard={hasClipboard()}
               showRemoveFromSubflow={contextMenuBlocks.some(
                 (b) => b.parentId && (b.parentType === 'loop' || b.parentType === 'parallel')
               )}
+              canRunFromBlock={runFromBlockState.canRun}
               disableEdit={!effectivePermissions.canEdit}
+              isExecuting={isExecuting}
+              isPositionalTrigger={
+                contextMenuBlocks.length === 1 &&
+                edges.filter((e) => e.target === contextMenuBlocks[0]?.id).length === 0
+              }
             />
 
             <CanvasMenu

apps/sim/executor/dag/builder.ts

@@ -33,6 +33,15 @@ export interface DAG {
   parallelConfigs: Map<string, SerializedParallel>
 }
 
+export interface DAGBuildOptions {
+  /** Trigger block ID to start path construction from */
+  triggerBlockId?: string
+  /** Saved incoming edges from snapshot for resumption */
+  savedIncomingEdges?: Record<string, string[]>
+  /** Include all enabled blocks instead of only those reachable from trigger */
+  includeAllBlocks?: boolean
+}
+
 export class DAGBuilder {
   private pathConstructor = new PathConstructor()
   private loopConstructor = new LoopConstructor()
@@ -40,11 +49,9 @@ export class DAGBuilder {
   private nodeConstructor = new NodeConstructor()
   private edgeConstructor = new EdgeConstructor()
 
-  build(
-    workflow: SerializedWorkflow,
-    triggerBlockId?: string,
-    savedIncomingEdges?: Record<string, string[]>
-  ): DAG {
+  build(workflow: SerializedWorkflow, options: DAGBuildOptions = {}): DAG {
+    const { triggerBlockId, savedIncomingEdges, includeAllBlocks } = options
+
     const dag: DAG = {
       nodes: new Map(),
       loopConfigs: new Map(),
@@ -53,7 +60,7 @@ export class DAGBuilder {
 
     this.initializeConfigs(workflow, dag)
 
-    const reachableBlocks = this.pathConstructor.execute(workflow, triggerBlockId)
+    const reachableBlocks = this.pathConstructor.execute(workflow, triggerBlockId, includeAllBlocks)
 
     this.loopConstructor.execute(dag, reachableBlocks)
     this.parallelConstructor.execute(dag, reachableBlocks)

apps/sim/executor/dag/construction/paths.ts

@@ -6,7 +6,16 @@ import type { SerializedBlock, SerializedWorkflow } from '@/serializer/types'
 const logger = createLogger('PathConstructor')
 
 export class PathConstructor {
-  execute(workflow: SerializedWorkflow, triggerBlockId?: string): Set<string> {
+  execute(
+    workflow: SerializedWorkflow,
+    triggerBlockId?: string,
+    includeAllBlocks?: boolean
+  ): Set<string> {
+    // For run-from-block mode, include all enabled blocks regardless of trigger reachability
+    if (includeAllBlocks) {
+      return this.getAllEnabledBlocks(workflow)
+    }
+
     const resolvedTriggerId = this.findTriggerBlock(workflow, triggerBlockId)
 
     if (!resolvedTriggerId) {

apps/sim/executor/execution/block-executor.ts

@@ -28,7 +28,6 @@ import type {
 } from '@/executor/types'
 import { streamingResponseFormatProcessor } from '@/executor/utils'
 import { buildBlockExecutionError, normalizeError } from '@/executor/utils/errors'
-import { isJSONString } from '@/executor/utils/json'
 import { filterOutputForLog } from '@/executor/utils/output-filter'
 import { validateBlockType } from '@/executor/utils/permission-check'
 import type { VariableResolver } from '@/executor/variables/resolver'
@@ -87,7 +86,7 @@ export class BlockExecutor {
       resolvedInputs = this.resolver.resolveInputs(ctx, node.id, block.config.params, block)
 
       if (blockLog) {
-        blockLog.input = this.parseJsonInputs(resolvedInputs)
+        blockLog.input = resolvedInputs
       }
     } catch (error) {
       cleanupSelfReference?.()
@@ -158,14 +157,7 @@ export class BlockExecutor {
         const displayOutput = filterOutputForLog(block.metadata?.id || '', normalizedOutput, {
           block,
         })
-        this.callOnBlockComplete(
-          ctx,
-          node,
-          block,
-          this.parseJsonInputs(resolvedInputs),
-          displayOutput,
-          duration
-        )
+        this.callOnBlockComplete(ctx, node, block, resolvedInputs, displayOutput, duration)
       }
 
       return normalizedOutput
@@ -241,7 +233,7 @@ export class BlockExecutor {
       blockLog.durationMs = duration
       blockLog.success = false
       blockLog.error = errorMessage
-      blockLog.input = this.parseJsonInputs(input)
+      blockLog.input = input
       blockLog.output = filterOutputForLog(block.metadata?.id || '', errorOutput, { block })
     }
 
@@ -256,14 +248,7 @@ export class BlockExecutor {
 
     if (!isSentinel) {
       const displayOutput = filterOutputForLog(block.metadata?.id || '', errorOutput, { block })
-      this.callOnBlockComplete(
-        ctx,
-        node,
-        block,
-        this.parseJsonInputs(input),
-        displayOutput,
-        duration
-      )
+      this.callOnBlockComplete(ctx, node, block, input, displayOutput, duration)
     }
 
     const hasErrorPort = this.hasErrorPortEdge(node)
@@ -351,36 +336,6 @@ export class BlockExecutor {
     return { result: output }
   }
 
-  /**
-   * Parse JSON string inputs to objects for log display only.
-   * Attempts to parse any string that looks like JSON.
-   * Returns a new object - does not mutate the original inputs.
-   */
-  private parseJsonInputs(inputs: Record<string, any>): Record<string, any> {
-    let result = inputs
-    let hasChanges = false
-
-    for (const [key, value] of Object.entries(inputs)) {
-      // isJSONString is a quick heuristic (checks for { or [), not a validator.
-      // Invalid JSON is safely caught below - this just avoids JSON.parse on every string.
-      if (typeof value !== 'string' || !isJSONString(value)) {
-        continue
-      }
-
-      try {
-        if (!hasChanges) {
-          result = { ...inputs }
-          hasChanges = true
-        }
-        result[key] = JSON.parse(value.trim())
-      } catch {
-        // Not valid JSON, keep original string
-      }
-    }
-
-    return result
-  }
-
   private callOnBlockStart(ctx: ExecutionContext, node: DAGNode, block: SerializedBlock): void {
     const blockId = node.id
     const blockName = block.metadata?.name ?? blockId

apps/sim/executor/execution/engine.ts

@@ -26,6 +26,7 @@ export class ExecutionEngine {
   private allowResumeTriggers: boolean
   private cancelledFlag = false
   private errorFlag = false
+  private stoppedEarlyFlag = false
   private executionError: Error | null = null
   private lastCancellationCheck = 0
   private readonly useRedisCancellation: boolean
@@ -105,7 +106,7 @@ export class ExecutionEngine {
       this.initializeQueue(triggerBlockId)
 
       while (this.hasWork()) {
-        if ((await this.checkCancellation()) || this.errorFlag) {
+        if ((await this.checkCancellation()) || this.errorFlag || this.stoppedEarlyFlag) {
           break
         }
         await this.processQueue()
@@ -259,6 +260,16 @@ export class ExecutionEngine {
   }
 
   private initializeQueue(triggerBlockId?: string): void {
+    if (this.context.runFromBlockContext) {
+      const { startBlockId } = this.context.runFromBlockContext
+      logger.info('Initializing queue for run-from-block mode', {
+        startBlockId,
+        dirtySetSize: this.context.runFromBlockContext.dirtySet.size,
+      })
+      this.addToQueue(startBlockId)
+      return
+    }
+
     const pendingBlocks = this.context.metadata.pendingBlocks
     const remainingEdges = (this.context.metadata as any).remainingEdges
 
@@ -385,6 +396,17 @@ export class ExecutionEngine {
       this.finalOutput = output
     }
 
+    if (this.context.stopAfterBlockId === nodeId) {
+      // For loop/parallel sentinels, only stop if the subflow has fully exited (all iterations done)
+      // shouldContinue: true means more iterations, shouldExit: true means loop is done
+      const shouldContinueLoop = output.shouldContinue === true
+      if (!shouldContinueLoop) {
+        logger.info('Stopping execution after target block', { nodeId })
+        this.stoppedEarlyFlag = true
+        return
+      }
+    }
+
     const readyNodes = this.edgeManager.processOutgoingEdges(node, output, false)
 
     logger.info('Processing outgoing edges', {

apps/sim/executor/execution/executor.ts

@@ -5,17 +5,31 @@ import { BlockExecutor } from '@/executor/execution/block-executor'
 import { EdgeManager } from '@/executor/execution/edge-manager'
 import { ExecutionEngine } from '@/executor/execution/engine'
 import { ExecutionState } from '@/executor/execution/state'
-import type { ContextExtensions, WorkflowInput } from '@/executor/execution/types'
+import type {
+  ContextExtensions,
+  SerializableExecutionState,
+  WorkflowInput,
+} from '@/executor/execution/types'
 import { createBlockHandlers } from '@/executor/handlers/registry'
 import { LoopOrchestrator } from '@/executor/orchestrators/loop'
 import { NodeExecutionOrchestrator } from '@/executor/orchestrators/node'
 import { ParallelOrchestrator } from '@/executor/orchestrators/parallel'
 import type { BlockState, ExecutionContext, ExecutionResult } from '@/executor/types'
+import {
+  computeExecutionSets,
+  type RunFromBlockContext,
+  resolveContainerToSentinelStart,
+  validateRunFromBlock,
+} from '@/executor/utils/run-from-block'
 import {
   buildResolutionFromBlock,
   buildStartBlockOutput,
   resolveExecutorStartBlock,
 } from '@/executor/utils/start-block'
+import {
+  extractLoopIdFromSentinel,
+  extractParallelIdFromSentinel,
+} from '@/executor/utils/subflow-utils'
 import { VariableResolver } from '@/executor/variables/resolver'
 import type { SerializedWorkflow } from '@/serializer/types'
 
@@ -48,7 +62,10 @@ export class DAGExecutor {
 
   async execute(workflowId: string, triggerBlockId?: string): Promise<ExecutionResult> {
     const savedIncomingEdges = this.contextExtensions.dagIncomingEdges
-    const dag = this.dagBuilder.build(this.workflow, triggerBlockId, savedIncomingEdges)
+    const dag = this.dagBuilder.build(this.workflow, {
+      triggerBlockId,
+      savedIncomingEdges,
+    })
     const { context, state } = this.createExecutionContext(workflowId, triggerBlockId)
 
     const resolver = new VariableResolver(this.workflow, this.workflowVariables, state)
@@ -89,17 +106,151 @@ export class DAGExecutor {
     }
   }
 
+  /**
+   * Execute from a specific block using cached outputs for upstream blocks.
+   */
+  async executeFromBlock(
+    workflowId: string,
+    startBlockId: string,
+    sourceSnapshot: SerializableExecutionState
+  ): Promise<ExecutionResult> {
+    // Build full DAG with all blocks to compute upstream set for snapshot filtering
+    // includeAllBlocks is needed because the startBlockId might be a trigger not reachable from the main trigger
+    const dag = this.dagBuilder.build(this.workflow, { includeAllBlocks: true })
+
+    const executedBlocks = new Set(sourceSnapshot.executedBlocks)
+    const validation = validateRunFromBlock(startBlockId, dag, executedBlocks)
+    if (!validation.valid) {
+      throw new Error(validation.error)
+    }
+
+    const { dirtySet, upstreamSet } = computeExecutionSets(dag, startBlockId)
+    const effectiveStartBlockId = resolveContainerToSentinelStart(startBlockId, dag) ?? startBlockId
+
+    // Extract container IDs from sentinel IDs in upstream set
+    const upstreamContainerIds = new Set<string>()
+    for (const nodeId of upstreamSet) {
+      const loopId = extractLoopIdFromSentinel(nodeId)
+      if (loopId) upstreamContainerIds.add(loopId)
+      const parallelId = extractParallelIdFromSentinel(nodeId)
+      if (parallelId) upstreamContainerIds.add(parallelId)
+    }
+
+    // Filter snapshot to only include upstream blocks - prevents references to non-upstream blocks
+    const filteredBlockStates: Record<string, any> = {}
+    for (const [blockId, state] of Object.entries(sourceSnapshot.blockStates)) {
+      if (upstreamSet.has(blockId) || upstreamContainerIds.has(blockId)) {
+        filteredBlockStates[blockId] = state
+      }
+    }
+    const filteredExecutedBlocks = sourceSnapshot.executedBlocks.filter(
+      (id) => upstreamSet.has(id) || upstreamContainerIds.has(id)
+    )
+
+    // Filter loop/parallel executions to only include upstream containers
+    const filteredLoopExecutions: Record<string, any> = {}
+    if (sourceSnapshot.loopExecutions) {
+      for (const [loopId, execution] of Object.entries(sourceSnapshot.loopExecutions)) {
+        if (upstreamContainerIds.has(loopId)) {
+          filteredLoopExecutions[loopId] = execution
+        }
+      }
+    }
+    const filteredParallelExecutions: Record<string, any> = {}
+    if (sourceSnapshot.parallelExecutions) {
+      for (const [parallelId, execution] of Object.entries(sourceSnapshot.parallelExecutions)) {
+        if (upstreamContainerIds.has(parallelId)) {
+          filteredParallelExecutions[parallelId] = execution
+        }
+      }
+    }
+
+    const filteredSnapshot: SerializableExecutionState = {
+      ...sourceSnapshot,
+      blockStates: filteredBlockStates,
+      executedBlocks: filteredExecutedBlocks,
+      loopExecutions: filteredLoopExecutions,
+      parallelExecutions: filteredParallelExecutions,
+    }
+
+    logger.info('Executing from block', {
+      workflowId,
+      startBlockId,
+      effectiveStartBlockId,
+      dirtySetSize: dirtySet.size,
+      upstreamSetSize: upstreamSet.size,
+    })
+
+    // Remove incoming edges from non-dirty sources so convergent blocks don't wait for cached upstream
+    for (const nodeId of dirtySet) {
+      const node = dag.nodes.get(nodeId)
+      if (!node) continue
+
+      const nonDirtyIncoming: string[] = []
+      for (const sourceId of node.incomingEdges) {
+        if (!dirtySet.has(sourceId)) {
+          nonDirtyIncoming.push(sourceId)
+        }
+      }
+
+      for (const sourceId of nonDirtyIncoming) {
+        node.incomingEdges.delete(sourceId)
+      }
+    }
+
+    const runFromBlockContext = { startBlockId: effectiveStartBlockId, dirtySet }
+    const { context, state } = this.createExecutionContext(workflowId, undefined, {
+      snapshotState: filteredSnapshot,
+      runFromBlockContext,
+    })
+
+    const resolver = new VariableResolver(this.workflow, this.workflowVariables, state)
+    const loopOrchestrator = new LoopOrchestrator(dag, state, resolver)
+    loopOrchestrator.setContextExtensions(this.contextExtensions)
+    const parallelOrchestrator = new ParallelOrchestrator(dag, state)
+    parallelOrchestrator.setResolver(resolver)
+    parallelOrchestrator.setContextExtensions(this.contextExtensions)
+    const allHandlers = createBlockHandlers()
+    const blockExecutor = new BlockExecutor(allHandlers, resolver, this.contextExtensions, state)
+    const edgeManager = new EdgeManager(dag)
+    loopOrchestrator.setEdgeManager(edgeManager)
+    const nodeOrchestrator = new NodeExecutionOrchestrator(
+      dag,
+      state,
+      blockExecutor,
+      loopOrchestrator,
+      parallelOrchestrator
+    )
+    const engine = new ExecutionEngine(context, dag, edgeManager, nodeOrchestrator)
+
+    return await engine.run()
+  }
+
   private createExecutionContext(
     workflowId: string,
-    triggerBlockId?: string
+    triggerBlockId?: string,
+    overrides?: {
+      snapshotState?: SerializableExecutionState
+      runFromBlockContext?: RunFromBlockContext
+    }
   ): { context: ExecutionContext; state: ExecutionState } {
-    const snapshotState = this.contextExtensions.snapshotState
+    const snapshotState = overrides?.snapshotState ?? this.contextExtensions.snapshotState
     const blockStates = snapshotState?.blockStates
       ? new Map(Object.entries(snapshotState.blockStates))
       : new Map<string, BlockState>()
-    const executedBlocks = snapshotState?.executedBlocks
+    let executedBlocks = snapshotState?.executedBlocks
       ? new Set(snapshotState.executedBlocks)
       : new Set<string>()
+
+    if (overrides?.runFromBlockContext) {
+      const { dirtySet } = overrides.runFromBlockContext
+      executedBlocks = new Set([...executedBlocks].filter((id) => !dirtySet.has(id)))
+      logger.info('Cleared executed status for dirty blocks', {
+        dirtySetSize: dirtySet.size,
+        remainingExecutedBlocks: executedBlocks.size,
+      })
+    }
+
     const state = new ExecutionState(blockStates, executedBlocks)
 
     const context: ExecutionContext = {
@@ -109,7 +260,7 @@ export class DAGExecutor {
       userId: this.contextExtensions.userId,
       isDeployedContext: this.contextExtensions.isDeployedContext,
       blockStates: state.getBlockStates(),
-      blockLogs: snapshotState?.blockLogs ?? [],
+      blockLogs: overrides?.runFromBlockContext ? [] : (snapshotState?.blockLogs ?? []),
       metadata: {
         ...this.contextExtensions.metadata,
         startTime: new Date().toISOString(),
@@ -169,6 +320,8 @@ export class DAGExecutor {
       abortSignal: this.contextExtensions.abortSignal,
       includeFileBase64: this.contextExtensions.includeFileBase64,
       base64MaxBytes: this.contextExtensions.base64MaxBytes,
+      runFromBlockContext: overrides?.runFromBlockContext,
+      stopAfterBlockId: this.contextExtensions.stopAfterBlockId,
     }
 
     if (this.contextExtensions.resumeFromSnapshot) {
@@ -193,6 +346,15 @@ export class DAGExecutor {
         pendingBlocks: context.metadata.pendingBlocks,
         skipStarterBlockInit: true,
       })
+    } else if (overrides?.runFromBlockContext) {
+      // In run-from-block mode, initialize the start block only if it's a regular block
+      // Skip for sentinels/containers (loop/parallel) which aren't real blocks
+      const startBlockId = overrides.runFromBlockContext.startBlockId
+      const isRegularBlock = this.workflow.blocks.some((b) => b.id === startBlockId)
+
+      if (isRegularBlock) {
+        this.initializeStarterBlock(context, state, startBlockId)
+      }
     } else {
       this.initializeStarterBlock(context, state, triggerBlockId)
     }

apps/sim/executor/execution/types.ts

@@ -1,5 +1,6 @@
 import type { Edge } from 'reactflow'
 import type { BlockLog, BlockState, NormalizedBlockOutput } from '@/executor/types'
+import type { RunFromBlockContext } from '@/executor/utils/run-from-block'
 import type { SubflowType } from '@/stores/workflows/workflow/types'
 
 export interface ExecutionMetadata {
@@ -105,6 +106,17 @@ export interface ContextExtensions {
     output: { input?: any; output: NormalizedBlockOutput; executionTime: number },
     iterationContext?: IterationContext
   ) => Promise<void>
+
+  /**
+   * Run-from-block configuration. When provided, executor runs in partial
+   * execution mode starting from the specified block.
+   */
+  runFromBlockContext?: RunFromBlockContext
+
+  /**
+   * Stop execution after this block completes. Used for "run until block" feature.
+   */
+  stopAfterBlockId?: string
 }
 
 export interface WorkflowInput {

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

@@ -936,12 +936,8 @@ export class AgentBlockHandler implements BlockHandler {
       systemPrompt: validMessages ? undefined : inputs.systemPrompt,
       context: validMessages ? undefined : stringifyJSON(messages),
       tools: formattedTools,
-      temperature:
-        inputs.temperature != null && inputs.temperature !== ''
-          ? Number(inputs.temperature)
-          : undefined,
-      maxTokens:
-        inputs.maxTokens != null && inputs.maxTokens !== '' ? Number(inputs.maxTokens) : undefined,
+      temperature: inputs.temperature,
+      maxTokens: inputs.maxTokens,
       apiKey: inputs.apiKey,
       azureEndpoint: inputs.azureEndpoint,
       azureApiVersion: inputs.azureApiVersion,

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

@@ -14,8 +14,8 @@ export interface AgentInputs {
   slidingWindowSize?: string // For message-based sliding window
   slidingWindowTokens?: string // For token-based sliding window
   // LLM parameters
-  temperature?: string
-  maxTokens?: string
+  temperature?: number
+  maxTokens?: number
   apiKey?: string
   azureEndpoint?: string
   azureApiVersion?: string

apps/sim/executor/orchestrators/loop.ts

@@ -276,7 +276,16 @@ export class LoopOrchestrator {
     scope: LoopScope
   ): LoopContinuationResult {
     const results = scope.allIterationOutputs
-    this.state.setBlockOutput(loopId, { results }, DEFAULTS.EXECUTION_TIME)
+    const output = { results }
+    this.state.setBlockOutput(loopId, output, DEFAULTS.EXECUTION_TIME)
+
+    // Emit onBlockComplete for the loop container so the UI can track it
+    if (this.contextExtensions?.onBlockComplete) {
+      this.contextExtensions.onBlockComplete(loopId, 'Loop', 'loop', {
+        output,
+        executionTime: DEFAULTS.EXECUTION_TIME,
+      })
+    }
 
     return {
       shouldContinue: false,

apps/sim/executor/orchestrators/node.ts

@@ -31,7 +31,18 @@ export class NodeExecutionOrchestrator {
       throw new Error(`Node not found in DAG: ${nodeId}`)
     }
 
-    if (this.state.hasExecuted(nodeId)) {
+    if (ctx.runFromBlockContext && !ctx.runFromBlockContext.dirtySet.has(nodeId)) {
+      const cachedOutput = this.state.getBlockOutput(nodeId) || {}
+      logger.debug('Skipping non-dirty block in run-from-block mode', { nodeId })
+      return {
+        nodeId,
+        output: cachedOutput,
+        isFinalOutput: false,
+      }
+    }
+
+    const isDirtyBlock = ctx.runFromBlockContext?.dirtySet.has(nodeId) ?? false
+    if (!isDirtyBlock && this.state.hasExecuted(nodeId)) {
       const output = this.state.getBlockOutput(nodeId) || {}
       return {
         nodeId,

apps/sim/executor/orchestrators/parallel.ts

@@ -228,9 +228,17 @@ export class ParallelOrchestrator {
       const branchOutputs = scope.branchOutputs.get(i) || []
       results.push(branchOutputs)
     }
-    this.state.setBlockOutput(parallelId, {
-      results,
-    })
+    const output = { results }
+    this.state.setBlockOutput(parallelId, output)
+
+    // Emit onBlockComplete for the parallel container so the UI can track it
+    if (this.contextExtensions?.onBlockComplete) {
+      this.contextExtensions.onBlockComplete(parallelId, 'Parallel', 'parallel', {
+        output,
+        executionTime: 0,
+      })
+    }
+
     return {
       allBranchesComplete: true,
       results,

apps/sim/executor/types.ts

@@ -1,6 +1,7 @@
 import type { TraceSpan } from '@/lib/logs/types'
 import type { PermissionGroupConfig } from '@/lib/permission-groups/types'
 import type { BlockOutput } from '@/blocks/types'
+import type { RunFromBlockContext } from '@/executor/utils/run-from-block'
 import type { SerializedBlock, SerializedWorkflow } from '@/serializer/types'
 
 export interface UserFile {
@@ -250,6 +251,17 @@ export interface ExecutionContext {
    * will not have their base64 content fetched.
    */
   base64MaxBytes?: number
+
+  /**
+   * Context for "run from block" mode. When present, only blocks in dirtySet
+   * will be executed; others return cached outputs from the source snapshot.
+   */
+  runFromBlockContext?: RunFromBlockContext
+
+  /**
+   * Stop execution after this block completes. Used for "run until block" feature.
+   */
+  stopAfterBlockId?: string
 }
 
 export interface ExecutionResult {

apps/sim/executor/utils/run-from-block.test.ts

@@ -0,0 +1,610 @@
+import { describe, expect, it } from 'vitest'
+import type { DAG, DAGNode } from '@/executor/dag/builder'
+import type { DAGEdge, NodeMetadata } from '@/executor/dag/types'
+import { computeExecutionSets, validateRunFromBlock } from '@/executor/utils/run-from-block'
+import type { SerializedLoop, SerializedParallel } from '@/serializer/types'
+
+/**
+ * Helper to extract dirty set from computeExecutionSets
+ */
+function computeDirtySet(dag: DAG, startBlockId: string): Set<string> {
+  return computeExecutionSets(dag, startBlockId).dirtySet
+}
+
+/**
+ * Helper to create a DAG node for testing
+ */
+function createNode(
+  id: string,
+  outgoingEdges: Array<{ target: string; sourceHandle?: string }> = [],
+  metadata: Partial<NodeMetadata> = {}
+): DAGNode {
+  const edges = new Map<string, DAGEdge>()
+  for (const edge of outgoingEdges) {
+    edges.set(edge.target, { target: edge.target, sourceHandle: edge.sourceHandle })
+  }
+
+  return {
+    id,
+    block: {
+      id,
+      position: { x: 0, y: 0 },
+      config: { tool: 'test', params: {} },
+      inputs: {},
+      outputs: {},
+      metadata: { id: 'test', name: `block-${id}`, category: 'tools' },
+      enabled: true,
+    },
+    incomingEdges: new Set<string>(),
+    outgoingEdges: edges,
+    metadata: {
+      isParallelBranch: false,
+      isLoopNode: false,
+      isSentinel: false,
+      ...metadata,
+    },
+  }
+}
+
+/**
+ * Helper to create a DAG for testing
+ */
+function createDAG(nodes: DAGNode[]): DAG {
+  const nodeMap = new Map<string, DAGNode>()
+  for (const node of nodes) {
+    nodeMap.set(node.id, node)
+  }
+
+  // Set up incoming edges based on outgoing edges
+  for (const node of nodes) {
+    for (const [, edge] of node.outgoingEdges) {
+      const targetNode = nodeMap.get(edge.target)
+      if (targetNode) {
+        targetNode.incomingEdges.add(node.id)
+      }
+    }
+  }
+
+  return {
+    nodes: nodeMap,
+    loopConfigs: new Map<string, SerializedLoop>(),
+    parallelConfigs: new Map<string, SerializedParallel>(),
+  }
+}
+
+describe('computeDirtySet', () => {
+  it('includes start block in dirty set', () => {
+    const dag = createDAG([createNode('A'), createNode('B'), createNode('C')])
+
+    const dirtySet = computeDirtySet(dag, 'B')
+
+    expect(dirtySet.has('B')).toBe(true)
+  })
+
+  it('includes all downstream blocks in linear workflow', () => {
+    // A → B → C → D
+    const dag = createDAG([
+      createNode('A', [{ target: 'B' }]),
+      createNode('B', [{ target: 'C' }]),
+      createNode('C', [{ target: 'D' }]),
+      createNode('D'),
+    ])
+
+    const dirtySet = computeDirtySet(dag, 'B')
+
+    expect(dirtySet.has('A')).toBe(false)
+    expect(dirtySet.has('B')).toBe(true)
+    expect(dirtySet.has('C')).toBe(true)
+    expect(dirtySet.has('D')).toBe(true)
+    expect(dirtySet.size).toBe(3)
+  })
+
+  it('handles branching paths', () => {
+    // A → B → C
+    //     ↓
+    //     D → E
+    const dag = createDAG([
+      createNode('A', [{ target: 'B' }]),
+      createNode('B', [{ target: 'C' }, { target: 'D' }]),
+      createNode('C'),
+      createNode('D', [{ target: 'E' }]),
+      createNode('E'),
+    ])
+
+    const dirtySet = computeDirtySet(dag, 'B')
+
+    expect(dirtySet.has('A')).toBe(false)
+    expect(dirtySet.has('B')).toBe(true)
+    expect(dirtySet.has('C')).toBe(true)
+    expect(dirtySet.has('D')).toBe(true)
+    expect(dirtySet.has('E')).toBe(true)
+    expect(dirtySet.size).toBe(4)
+  })
+
+  it('handles convergence points', () => {
+    // A → C
+    // B → C → D
+    const dag = createDAG([
+      createNode('A', [{ target: 'C' }]),
+      createNode('B', [{ target: 'C' }]),
+      createNode('C', [{ target: 'D' }]),
+      createNode('D'),
+    ])
+
+    // Run from A: should include A, C, D (but not B)
+    const dirtySet = computeDirtySet(dag, 'A')
+
+    expect(dirtySet.has('A')).toBe(true)
+    expect(dirtySet.has('B')).toBe(false)
+    expect(dirtySet.has('C')).toBe(true)
+    expect(dirtySet.has('D')).toBe(true)
+    expect(dirtySet.size).toBe(3)
+  })
+
+  it('handles diamond pattern', () => {
+    //     B
+    //   ↗   ↘
+    // A       D
+    //   ↘   ↗
+    //     C
+    const dag = createDAG([
+      createNode('A', [{ target: 'B' }, { target: 'C' }]),
+      createNode('B', [{ target: 'D' }]),
+      createNode('C', [{ target: 'D' }]),
+      createNode('D'),
+    ])
+
+    const dirtySet = computeDirtySet(dag, 'A')
+
+    expect(dirtySet.has('A')).toBe(true)
+    expect(dirtySet.has('B')).toBe(true)
+    expect(dirtySet.has('C')).toBe(true)
+    expect(dirtySet.has('D')).toBe(true)
+    expect(dirtySet.size).toBe(4)
+  })
+
+  it('stops at graph boundaries', () => {
+    // A → B    C → D (disconnected)
+    const dag = createDAG([
+      createNode('A', [{ target: 'B' }]),
+      createNode('B'),
+      createNode('C', [{ target: 'D' }]),
+      createNode('D'),
+    ])
+
+    const dirtySet = computeDirtySet(dag, 'A')
+
+    expect(dirtySet.has('A')).toBe(true)
+    expect(dirtySet.has('B')).toBe(true)
+    expect(dirtySet.has('C')).toBe(false)
+    expect(dirtySet.has('D')).toBe(false)
+    expect(dirtySet.size).toBe(2)
+  })
+
+  it('handles single node workflow', () => {
+    const dag = createDAG([createNode('A')])
+
+    const dirtySet = computeDirtySet(dag, 'A')
+
+    expect(dirtySet.has('A')).toBe(true)
+    expect(dirtySet.size).toBe(1)
+  })
+
+  it('handles node not in DAG gracefully', () => {
+    const dag = createDAG([createNode('A'), createNode('B')])
+
+    const dirtySet = computeDirtySet(dag, 'nonexistent')
+
+    // Should just contain the start block ID even if not found
+    expect(dirtySet.has('nonexistent')).toBe(true)
+    expect(dirtySet.size).toBe(1)
+  })
+
+  it('includes convergent block when running from one branch of parallel', () => {
+    // Parallel branches converging:
+    // A → B → D
+    // A → C → D
+    // Running from B should include B and D (but not A or C)
+    const dag = createDAG([
+      createNode('A', [{ target: 'B' }, { target: 'C' }]),
+      createNode('B', [{ target: 'D' }]),
+      createNode('C', [{ target: 'D' }]),
+      createNode('D'),
+    ])
+
+    const dirtySet = computeDirtySet(dag, 'B')
+
+    expect(dirtySet.has('A')).toBe(false)
+    expect(dirtySet.has('B')).toBe(true)
+    expect(dirtySet.has('C')).toBe(false)
+    expect(dirtySet.has('D')).toBe(true)
+    expect(dirtySet.size).toBe(2)
+  })
+
+  it('handles running from convergent block itself (all upstream non-dirty)', () => {
+    // A → C
+    // B → C
+    // Running from C should only include C
+    const dag = createDAG([
+      createNode('A', [{ target: 'C' }]),
+      createNode('B', [{ target: 'C' }]),
+      createNode('C', [{ target: 'D' }]),
+      createNode('D'),
+    ])
+
+    const dirtySet = computeDirtySet(dag, 'C')
+
+    expect(dirtySet.has('A')).toBe(false)
+    expect(dirtySet.has('B')).toBe(false)
+    expect(dirtySet.has('C')).toBe(true)
+    expect(dirtySet.has('D')).toBe(true)
+    expect(dirtySet.size).toBe(2)
+  })
+
+  it('handles deep downstream chains', () => {
+    // A → B → C → D → E → F
+    // Running from C should include C, D, E, F
+    const dag = createDAG([
+      createNode('A', [{ target: 'B' }]),
+      createNode('B', [{ target: 'C' }]),
+      createNode('C', [{ target: 'D' }]),
+      createNode('D', [{ target: 'E' }]),
+      createNode('E', [{ target: 'F' }]),
+      createNode('F'),
+    ])
+
+    const dirtySet = computeDirtySet(dag, 'C')
+
+    expect(dirtySet.has('A')).toBe(false)
+    expect(dirtySet.has('B')).toBe(false)
+    expect(dirtySet.has('C')).toBe(true)
+    expect(dirtySet.has('D')).toBe(true)
+    expect(dirtySet.has('E')).toBe(true)
+    expect(dirtySet.has('F')).toBe(true)
+    expect(dirtySet.size).toBe(4)
+  })
+})
+
+describe('validateRunFromBlock', () => {
+  it('accepts valid block', () => {
+    const dag = createDAG([createNode('A'), createNode('B')])
+    const executedBlocks = new Set(['A', 'B'])
+
+    const result = validateRunFromBlock('A', dag, executedBlocks)
+
+    expect(result.valid).toBe(true)
+    expect(result.error).toBeUndefined()
+  })
+
+  it('rejects block not found in DAG', () => {
+    const dag = createDAG([createNode('A')])
+    const executedBlocks = new Set(['A', 'B'])
+
+    const result = validateRunFromBlock('B', dag, executedBlocks)
+
+    expect(result.valid).toBe(false)
+    expect(result.error).toContain('Block not found')
+  })
+
+  it('rejects blocks inside loops', () => {
+    const dag = createDAG([createNode('A', [], { isLoopNode: true, loopId: 'loop-1' })])
+    const executedBlocks = new Set(['A'])
+
+    const result = validateRunFromBlock('A', dag, executedBlocks)
+
+    expect(result.valid).toBe(false)
+    expect(result.error).toContain('inside loop')
+    expect(result.error).toContain('loop-1')
+  })
+
+  it('rejects blocks inside parallels', () => {
+    const dag = createDAG([
+      createNode('A', [], { isParallelBranch: true, parallelId: 'parallel-1' }),
+    ])
+    const executedBlocks = new Set(['A'])
+
+    const result = validateRunFromBlock('A', dag, executedBlocks)
+
+    expect(result.valid).toBe(false)
+    expect(result.error).toContain('inside parallel')
+    expect(result.error).toContain('parallel-1')
+  })
+
+  it('rejects sentinel nodes', () => {
+    const dag = createDAG([createNode('A', [], { isSentinel: true, sentinelType: 'start' })])
+    const executedBlocks = new Set(['A'])
+
+    const result = validateRunFromBlock('A', dag, executedBlocks)
+
+    expect(result.valid).toBe(false)
+    expect(result.error).toContain('sentinel')
+  })
+
+  it('rejects blocks with unexecuted upstream dependencies', () => {
+    // A → B, only A executed but B depends on A
+    const dag = createDAG([createNode('A', [{ target: 'B' }]), createNode('B')])
+    const executedBlocks = new Set<string>() // A was not executed
+
+    const result = validateRunFromBlock('B', dag, executedBlocks)
+
+    expect(result.valid).toBe(false)
+    expect(result.error).toContain('Upstream dependency not executed')
+  })
+
+  it('allows running from block when immediate predecessor was executed (ignores transitive)', () => {
+    // A → X → B → C, where X is new (not executed)
+    // Running from C is allowed because B (immediate predecessor) was executed
+    // C will use B's cached output - doesn't matter that X is new
+    const dag = createDAG([
+      createNode('A', [{ target: 'X' }]),
+      createNode('X', [{ target: 'B' }]),
+      createNode('B', [{ target: 'C' }]),
+      createNode('C'),
+    ])
+    const executedBlocks = new Set(['A', 'B', 'C']) // X was not executed (new block)
+
+    const result = validateRunFromBlock('C', dag, executedBlocks)
+
+    // Valid because C's immediate predecessor B was executed
+    expect(result.valid).toBe(true)
+  })
+
+  it('allows blocks with no dependencies even if not previously executed', () => {
+    // A and B are independent (no edges)
+    const dag = createDAG([createNode('A'), createNode('B')])
+    const executedBlocks = new Set(['A']) // B was not executed but has no deps
+
+    const result = validateRunFromBlock('B', dag, executedBlocks)
+
+    expect(result.valid).toBe(true) // B has no incoming edges, so it's valid
+  })
+
+  it('accepts regular executed block', () => {
+    const dag = createDAG([
+      createNode('trigger', [{ target: 'A' }]),
+      createNode('A', [{ target: 'B' }]),
+      createNode('B'),
+    ])
+    const executedBlocks = new Set(['trigger', 'A', 'B'])
+
+    const result = validateRunFromBlock('A', dag, executedBlocks)
+
+    expect(result.valid).toBe(true)
+  })
+
+  it('accepts loop container when executed', () => {
+    // Loop container with sentinel nodes
+    const loopId = 'loop-container-1'
+    const sentinelStartId = `loop-${loopId}-sentinel-start`
+    const sentinelEndId = `loop-${loopId}-sentinel-end`
+    const dag = createDAG([
+      createNode('A', [{ target: sentinelStartId }]),
+      createNode(sentinelStartId, [{ target: 'B' }], {
+        isSentinel: true,
+        sentinelType: 'start',
+        loopId,
+      }),
+      createNode('B', [{ target: sentinelEndId }], { isLoopNode: true, loopId }),
+      createNode(sentinelEndId, [{ target: 'C' }], {
+        isSentinel: true,
+        sentinelType: 'end',
+        loopId,
+      }),
+      createNode('C'),
+    ])
+    dag.loopConfigs.set(loopId, { id: loopId, nodes: ['B'], iterations: 3, loopType: 'for' } as any)
+    const executedBlocks = new Set(['A', loopId, sentinelStartId, 'B', sentinelEndId, 'C'])
+
+    const result = validateRunFromBlock(loopId, dag, executedBlocks)
+
+    expect(result.valid).toBe(true)
+  })
+
+  it('accepts parallel container when executed', () => {
+    // Parallel container with sentinel nodes
+    const parallelId = 'parallel-container-1'
+    const sentinelStartId = `parallel-${parallelId}-sentinel-start`
+    const sentinelEndId = `parallel-${parallelId}-sentinel-end`
+    const dag = createDAG([
+      createNode('A', [{ target: sentinelStartId }]),
+      createNode(sentinelStartId, [{ target: 'B₍0₎' }], {
+        isSentinel: true,
+        sentinelType: 'start',
+        parallelId,
+      }),
+      createNode('B₍0₎', [{ target: sentinelEndId }], { isParallelBranch: true, parallelId }),
+      createNode(sentinelEndId, [{ target: 'C' }], {
+        isSentinel: true,
+        sentinelType: 'end',
+        parallelId,
+      }),
+      createNode('C'),
+    ])
+    dag.parallelConfigs.set(parallelId, { id: parallelId, nodes: ['B'], count: 2 } as any)
+    const executedBlocks = new Set(['A', parallelId, sentinelStartId, 'B₍0₎', sentinelEndId, 'C'])
+
+    const result = validateRunFromBlock(parallelId, dag, executedBlocks)
+
+    expect(result.valid).toBe(true)
+  })
+
+  it('allows loop container with no upstream dependencies', () => {
+    // Loop containers are validated via their sentinel nodes, not incoming edges on the container itself
+    // If the loop has no upstream dependencies, it should be valid
+    const loopId = 'loop-container-1'
+    const sentinelStartId = `loop-${loopId}-sentinel-start`
+    const dag = createDAG([
+      createNode(sentinelStartId, [], { isSentinel: true, sentinelType: 'start', loopId }),
+    ])
+    dag.loopConfigs.set(loopId, { id: loopId, nodes: [], iterations: 3, loopType: 'for' } as any)
+    const executedBlocks = new Set<string>() // Nothing executed but loop has no deps
+
+    const result = validateRunFromBlock(loopId, dag, executedBlocks)
+
+    // Loop container validation doesn't check incoming edges (containers don't have nodes in dag.nodes)
+    // So this is valid - the loop can start fresh
+    expect(result.valid).toBe(true)
+  })
+})
+
+describe('computeDirtySet with containers', () => {
+  it('includes loop container and all downstream when running from loop', () => {
+    // A → loop-sentinel-start → B (inside loop) → loop-sentinel-end → C
+    const loopId = 'loop-1'
+    const sentinelStartId = `loop-${loopId}-sentinel-start`
+    const sentinelEndId = `loop-${loopId}-sentinel-end`
+    const dag = createDAG([
+      createNode('A', [{ target: sentinelStartId }]),
+      createNode(sentinelStartId, [{ target: 'B' }], {
+        isSentinel: true,
+        sentinelType: 'start',
+        loopId,
+      }),
+      createNode('B', [{ target: sentinelEndId }], { isLoopNode: true, loopId }),
+      createNode(sentinelEndId, [{ target: 'C' }], {
+        isSentinel: true,
+        sentinelType: 'end',
+        loopId,
+      }),
+      createNode('C'),
+    ])
+    dag.loopConfigs.set(loopId, { id: loopId, nodes: ['B'], iterations: 3, loopType: 'for' } as any)
+
+    const dirtySet = computeDirtySet(dag, loopId)
+
+    // Should include loop container, sentinel-start, B, sentinel-end, C
+    expect(dirtySet.has(loopId)).toBe(true)
+    expect(dirtySet.has(sentinelStartId)).toBe(true)
+    expect(dirtySet.has('B')).toBe(true)
+    expect(dirtySet.has(sentinelEndId)).toBe(true)
+    expect(dirtySet.has('C')).toBe(true)
+    // Should NOT include A (upstream)
+    expect(dirtySet.has('A')).toBe(false)
+  })
+
+  it('includes parallel container and all downstream when running from parallel', () => {
+    // A → parallel-sentinel-start → B₍0₎ → parallel-sentinel-end → C
+    const parallelId = 'parallel-1'
+    const sentinelStartId = `parallel-${parallelId}-sentinel-start`
+    const sentinelEndId = `parallel-${parallelId}-sentinel-end`
+    const dag = createDAG([
+      createNode('A', [{ target: sentinelStartId }]),
+      createNode(sentinelStartId, [{ target: 'B₍0₎' }], {
+        isSentinel: true,
+        sentinelType: 'start',
+        parallelId,
+      }),
+      createNode('B₍0₎', [{ target: sentinelEndId }], { isParallelBranch: true, parallelId }),
+      createNode(sentinelEndId, [{ target: 'C' }], {
+        isSentinel: true,
+        sentinelType: 'end',
+        parallelId,
+      }),
+      createNode('C'),
+    ])
+    dag.parallelConfigs.set(parallelId, { id: parallelId, nodes: ['B'], count: 2 } as any)
+
+    const dirtySet = computeDirtySet(dag, parallelId)
+
+    // Should include parallel container, sentinel-start, B₍0₎, sentinel-end, C
+    expect(dirtySet.has(parallelId)).toBe(true)
+    expect(dirtySet.has(sentinelStartId)).toBe(true)
+    expect(dirtySet.has('B₍0₎')).toBe(true)
+    expect(dirtySet.has(sentinelEndId)).toBe(true)
+    expect(dirtySet.has('C')).toBe(true)
+    // Should NOT include A (upstream)
+    expect(dirtySet.has('A')).toBe(false)
+  })
+})
+
+describe('computeExecutionSets upstream set', () => {
+  it('includes all upstream blocks in linear workflow', () => {
+    // A → B → C → D
+    const dag = createDAG([
+      createNode('A', [{ target: 'B' }]),
+      createNode('B', [{ target: 'C' }]),
+      createNode('C', [{ target: 'D' }]),
+      createNode('D'),
+    ])
+
+    const { upstreamSet } = computeExecutionSets(dag, 'C')
+
+    expect(upstreamSet.has('A')).toBe(true)
+    expect(upstreamSet.has('B')).toBe(true)
+    expect(upstreamSet.has('C')).toBe(false) // start block not in upstream
+    expect(upstreamSet.has('D')).toBe(false) // downstream
+  })
+
+  it('includes all branches in convergent upstream', () => {
+    // A → C
+    // B → C → D
+    const dag = createDAG([
+      createNode('A', [{ target: 'C' }]),
+      createNode('B', [{ target: 'C' }]),
+      createNode('C', [{ target: 'D' }]),
+      createNode('D'),
+    ])
+
+    const { upstreamSet } = computeExecutionSets(dag, 'C')
+
+    expect(upstreamSet.has('A')).toBe(true)
+    expect(upstreamSet.has('B')).toBe(true)
+    expect(upstreamSet.has('C')).toBe(false)
+    expect(upstreamSet.has('D')).toBe(false)
+  })
+
+  it('excludes parallel branches not in upstream path', () => {
+    // A → B → D
+    // A → C → D
+    // Running from B: upstream is A only, not C
+    const dag = createDAG([
+      createNode('A', [{ target: 'B' }, { target: 'C' }]),
+      createNode('B', [{ target: 'D' }]),
+      createNode('C', [{ target: 'D' }]),
+      createNode('D'),
+    ])
+
+    const { upstreamSet, dirtySet } = computeExecutionSets(dag, 'B')
+
+    // Upstream should only contain A
+    expect(upstreamSet.has('A')).toBe(true)
+    expect(upstreamSet.has('C')).toBe(false) // parallel branch, not upstream of B
+    // Dirty should contain B and D
+    expect(dirtySet.has('B')).toBe(true)
+    expect(dirtySet.has('D')).toBe(true)
+    expect(dirtySet.has('C')).toBe(false)
+  })
+
+  it('handles diamond pattern upstream correctly', () => {
+    //     B
+    //   ↗   ↘
+    // A       D → E
+    //   ↘   ↗
+    //     C
+    // Running from D: upstream should be A, B, C
+    const dag = createDAG([
+      createNode('A', [{ target: 'B' }, { target: 'C' }]),
+      createNode('B', [{ target: 'D' }]),
+      createNode('C', [{ target: 'D' }]),
+      createNode('D', [{ target: 'E' }]),
+      createNode('E'),
+    ])
+
+    const { upstreamSet, dirtySet } = computeExecutionSets(dag, 'D')
+
+    expect(upstreamSet.has('A')).toBe(true)
+    expect(upstreamSet.has('B')).toBe(true)
+    expect(upstreamSet.has('C')).toBe(true)
+    expect(upstreamSet.has('D')).toBe(false)
+    expect(dirtySet.has('D')).toBe(true)
+    expect(dirtySet.has('E')).toBe(true)
+  })
+
+  it('returns empty upstream set for root block', () => {
+    const dag = createDAG([createNode('A', [{ target: 'B' }]), createNode('B')])
+
+    const { upstreamSet } = computeExecutionSets(dag, 'A')
+
+    expect(upstreamSet.size).toBe(0)
+  })
+})

apps/sim/executor/utils/run-from-block.ts

@@ -0,0 +1,193 @@
+import { LOOP, PARALLEL } from '@/executor/constants'
+import type { DAG } from '@/executor/dag/builder'
+
+/**
+ * Builds the sentinel-start node ID for a loop.
+ */
+function buildLoopSentinelStartId(loopId: string): string {
+  return `${LOOP.SENTINEL.PREFIX}${loopId}${LOOP.SENTINEL.START_SUFFIX}`
+}
+
+/**
+ * Builds the sentinel-start node ID for a parallel.
+ */
+function buildParallelSentinelStartId(parallelId: string): string {
+  return `${PARALLEL.SENTINEL.PREFIX}${parallelId}${PARALLEL.SENTINEL.START_SUFFIX}`
+}
+
+/**
+ * Checks if a block ID is a loop or parallel container and returns the sentinel-start ID if so.
+ * Returns null if the block is not a container.
+ */
+export function resolveContainerToSentinelStart(blockId: string, dag: DAG): string | null {
+  if (dag.loopConfigs.has(blockId)) {
+    return buildLoopSentinelStartId(blockId)
+  }
+  if (dag.parallelConfigs.has(blockId)) {
+    return buildParallelSentinelStartId(blockId)
+  }
+  return null
+}
+
+/**
+ * Result of validating a block for run-from-block execution.
+ */
+export interface RunFromBlockValidation {
+  valid: boolean
+  error?: string
+}
+
+/**
+ * Context for run-from-block execution mode.
+ */
+export interface RunFromBlockContext {
+  /** The block ID to start execution from */
+  startBlockId: string
+  /** Set of block IDs that need re-execution (start block + all downstream) */
+  dirtySet: Set<string>
+}
+
+/**
+ * Result of computing execution sets for run-from-block mode.
+ */
+export interface ExecutionSets {
+  /** Blocks that need re-execution (start block + all downstream) */
+  dirtySet: Set<string>
+  /** Blocks that are upstream (ancestors) of the start block */
+  upstreamSet: Set<string>
+}
+
+/**
+ * Computes both the dirty set (downstream) and upstream set in a single traversal pass.
+ * - Dirty set: start block + all blocks reachable via outgoing edges (need re-execution)
+ * - Upstream set: all blocks reachable via incoming edges (can be referenced)
+ *
+ * For loop/parallel containers, starts from the sentinel-start node and includes
+ * the container ID itself in the dirty set.
+ *
+ * @param dag - The workflow DAG
+ * @param startBlockId - The block to start execution from
+ * @returns Object containing both dirtySet and upstreamSet
+ */
+export function computeExecutionSets(dag: DAG, startBlockId: string): ExecutionSets {
+  const dirty = new Set<string>([startBlockId])
+  const upstream = new Set<string>()
+  const sentinelStartId = resolveContainerToSentinelStart(startBlockId, dag)
+  const traversalStartId = sentinelStartId ?? startBlockId
+
+  if (sentinelStartId) {
+    dirty.add(sentinelStartId)
+  }
+
+  // BFS downstream for dirty set
+  const downstreamQueue = [traversalStartId]
+  while (downstreamQueue.length > 0) {
+    const nodeId = downstreamQueue.shift()!
+    const node = dag.nodes.get(nodeId)
+    if (!node) continue
+
+    for (const [, edge] of node.outgoingEdges) {
+      if (!dirty.has(edge.target)) {
+        dirty.add(edge.target)
+        downstreamQueue.push(edge.target)
+      }
+    }
+  }
+
+  // BFS upstream for upstream set
+  const upstreamQueue = [traversalStartId]
+  while (upstreamQueue.length > 0) {
+    const nodeId = upstreamQueue.shift()!
+    const node = dag.nodes.get(nodeId)
+    if (!node) continue
+
+    for (const sourceId of node.incomingEdges) {
+      if (!upstream.has(sourceId)) {
+        upstream.add(sourceId)
+        upstreamQueue.push(sourceId)
+      }
+    }
+  }
+
+  return { dirtySet: dirty, upstreamSet: upstream }
+}
+
+/**
+ * Validates that a block can be used as a run-from-block starting point.
+ *
+ * Validation rules:
+ * - Block must exist in the DAG (or be a loop/parallel container)
+ * - Block cannot be inside a loop (but loop containers are allowed)
+ * - Block cannot be inside a parallel (but parallel containers are allowed)
+ * - Block cannot be a sentinel node
+ * - All upstream dependencies must have been executed (have cached outputs)
+ *
+ * @param blockId - The block ID to validate
+ * @param dag - The workflow DAG
+ * @param executedBlocks - Set of blocks that were executed in the source run
+ * @returns Validation result with error message if invalid
+ */
+export function validateRunFromBlock(
+  blockId: string,
+  dag: DAG,
+  executedBlocks: Set<string>
+): RunFromBlockValidation {
+  const node = dag.nodes.get(blockId)
+  const isLoopContainer = dag.loopConfigs.has(blockId)
+  const isParallelContainer = dag.parallelConfigs.has(blockId)
+  const isContainer = isLoopContainer || isParallelContainer
+
+  if (!node && !isContainer) {
+    return { valid: false, error: `Block not found in workflow: ${blockId}` }
+  }
+
+  if (isContainer) {
+    const sentinelStartId = resolveContainerToSentinelStart(blockId, dag)
+    if (!sentinelStartId || !dag.nodes.has(sentinelStartId)) {
+      return {
+        valid: false,
+        error: `Container sentinel not found for: ${blockId}`,
+      }
+    }
+  }
+
+  if (node) {
+    if (node.metadata.isLoopNode) {
+      return {
+        valid: false,
+        error: `Cannot run from block inside loop: ${node.metadata.loopId}`,
+      }
+    }
+
+    if (node.metadata.isParallelBranch) {
+      return {
+        valid: false,
+        error: `Cannot run from block inside parallel: ${node.metadata.parallelId}`,
+      }
+    }
+
+    if (node.metadata.isSentinel) {
+      return { valid: false, error: 'Cannot run from sentinel node' }
+    }
+
+    // Check immediate upstream dependencies were executed
+    for (const sourceId of node.incomingEdges) {
+      const sourceNode = dag.nodes.get(sourceId)
+      // Skip sentinel nodes - they're internal and not in executedBlocks
+      if (sourceNode?.metadata.isSentinel) continue
+
+      // Skip trigger nodes - they're entry points and don't need prior execution
+      // A trigger node has no incoming edges
+      if (sourceNode && sourceNode.incomingEdges.size === 0) continue
+
+      if (!executedBlocks.has(sourceId)) {
+        return {
+          valid: false,
+          error: `Upstream dependency not executed: ${sourceId}`,
+        }
+      }
+    }
+  }
+
+  return { valid: true }
+}

apps/sim/hooks/use-execution-stream.ts

@@ -1,10 +1,85 @@
 import { useCallback, useRef } from 'react'
 import { createLogger } from '@sim/logger'
 import type { ExecutionEvent } from '@/lib/workflows/executor/execution-events'
+import type { SerializableExecutionState } from '@/executor/execution/types'
 import type { SubflowType } from '@/stores/workflows/workflow/types'
 
 const logger = createLogger('useExecutionStream')
 
+/**
+ * Processes SSE events from a response body and invokes appropriate callbacks.
+ */
+async function processSSEStream(
+  reader: ReadableStreamDefaultReader<Uint8Array>,
+  callbacks: ExecutionStreamCallbacks,
+  logPrefix: string
+): Promise<void> {
+  const decoder = new TextDecoder()
+  let buffer = ''
+
+  try {
+    while (true) {
+      const { done, value } = await reader.read()
+
+      if (done) break
+
+      buffer += decoder.decode(value, { stream: true })
+      const lines = buffer.split('\n\n')
+      buffer = lines.pop() || ''
+
+      for (const line of lines) {
+        if (!line.trim() || !line.startsWith('data: ')) continue
+
+        const data = line.substring(6).trim()
+        if (data === '[DONE]') {
+          logger.info(`${logPrefix} stream completed`)
+          continue
+        }
+
+        try {
+          const event = JSON.parse(data) as ExecutionEvent
+
+          switch (event.type) {
+            case 'execution:started':
+              callbacks.onExecutionStarted?.(event.data)
+              break
+            case 'execution:completed':
+              callbacks.onExecutionCompleted?.(event.data)
+              break
+            case 'execution:error':
+              callbacks.onExecutionError?.(event.data)
+              break
+            case 'execution:cancelled':
+              callbacks.onExecutionCancelled?.(event.data)
+              break
+            case 'block:started':
+              callbacks.onBlockStarted?.(event.data)
+              break
+            case 'block:completed':
+              callbacks.onBlockCompleted?.(event.data)
+              break
+            case 'block:error':
+              callbacks.onBlockError?.(event.data)
+              break
+            case 'stream:chunk':
+              callbacks.onStreamChunk?.(event.data)
+              break
+            case 'stream:done':
+              callbacks.onStreamDone?.(event.data)
+              break
+            default:
+              logger.warn('Unknown event type:', (event as any).type)
+          }
+        } catch (error) {
+          logger.error('Failed to parse SSE event:', error, { data })
+        }
+      }
+    }
+  } finally {
+    reader.releaseLock()
+  }
+}
+
 export interface ExecutionStreamCallbacks {
   onExecutionStarted?: (data: { startTime: string }) => void
   onExecutionCompleted?: (data: {
@@ -68,6 +143,15 @@ export interface ExecuteStreamOptions {
     loops?: Record<string, any>
     parallels?: Record<string, any>
   }
+  stopAfterBlockId?: string
+  callbacks?: ExecutionStreamCallbacks
+}
+
+export interface ExecuteFromBlockOptions {
+  workflowId: string
+  startBlockId: string
+  sourceSnapshot: SerializableExecutionState
+  input?: any
   callbacks?: ExecutionStreamCallbacks
 }
 
@@ -119,91 +203,7 @@ export function useExecutionStream() {
       }
 
       const reader = response.body.getReader()
-      const decoder = new TextDecoder()
-      let buffer = ''
-
-      try {
-        while (true) {
-          const { done, value } = await reader.read()
-
-          if (done) {
-            break
-          }
-
-          buffer += decoder.decode(value, { stream: true })
-
-          const lines = buffer.split('\n\n')
-
-          buffer = lines.pop() || ''
-
-          for (const line of lines) {
-            if (!line.trim() || !line.startsWith('data: ')) {
-              continue
-            }
-
-            const data = line.substring(6).trim()
-
-            if (data === '[DONE]') {
-              logger.info('Stream completed')
-              continue
-            }
-
-            try {
-              const event = JSON.parse(data) as ExecutionEvent
-
-              logger.info('📡 SSE Event received:', {
-                type: event.type,
-                executionId: event.executionId,
-                data: event.data,
-              })
-
-              switch (event.type) {
-                case 'execution:started':
-                  logger.info('🚀 Execution started')
-                  callbacks.onExecutionStarted?.(event.data)
-                  break
-                case 'execution:completed':
-                  logger.info('✅ Execution completed')
-                  callbacks.onExecutionCompleted?.(event.data)
-                  break
-                case 'execution:error':
-                  logger.error('❌ Execution error')
-                  callbacks.onExecutionError?.(event.data)
-                  break
-                case 'execution:cancelled':
-                  logger.warn('🛑 Execution cancelled')
-                  callbacks.onExecutionCancelled?.(event.data)
-                  break
-                case 'block:started':
-                  logger.info('🔷 Block started:', event.data.blockId)
-                  callbacks.onBlockStarted?.(event.data)
-                  break
-                case 'block:completed':
-                  logger.info('✓ Block completed:', event.data.blockId)
-                  callbacks.onBlockCompleted?.(event.data)
-                  break
-                case 'block:error':
-                  logger.error('✗ Block error:', event.data.blockId)
-                  callbacks.onBlockError?.(event.data)
-                  break
-                case 'stream:chunk':
-                  callbacks.onStreamChunk?.(event.data)
-                  break
-                case 'stream:done':
-                  logger.info('Stream done:', event.data.blockId)
-                  callbacks.onStreamDone?.(event.data)
-                  break
-                default:
-                  logger.warn('Unknown event type:', (event as any).type)
-              }
-            } catch (error) {
-              logger.error('Failed to parse SSE event:', error, { data })
-            }
-          }
-        }
-      } finally {
-        reader.releaseLock()
-      }
+      await processSSEStream(reader, callbacks, 'Execution')
     } catch (error: any) {
       if (error.name === 'AbortError') {
         logger.info('Execution stream cancelled')
@@ -222,6 +222,70 @@ export function useExecutionStream() {
     }
   }, [])
 
+  const executeFromBlock = useCallback(async (options: ExecuteFromBlockOptions) => {
+    const { workflowId, startBlockId, sourceSnapshot, input, callbacks = {} } = options
+
+    if (abortControllerRef.current) {
+      abortControllerRef.current.abort()
+    }
+
+    const abortController = new AbortController()
+    abortControllerRef.current = abortController
+    currentExecutionRef.current = null
+
+    try {
+      const response = await fetch(`/api/workflows/${workflowId}/execute-from-block`, {
+        method: 'POST',
+        headers: {
+          'Content-Type': 'application/json',
+        },
+        body: JSON.stringify({ startBlockId, sourceSnapshot, input }),
+        signal: abortController.signal,
+      })
+
+      if (!response.ok) {
+        let errorResponse: any
+        try {
+          errorResponse = await response.json()
+        } catch {
+          throw new Error(`Server error (${response.status}): ${response.statusText}`)
+        }
+        const error = new Error(errorResponse.error || 'Failed to start execution')
+        if (errorResponse && typeof errorResponse === 'object') {
+          Object.assign(error, { executionResult: errorResponse })
+        }
+        throw error
+      }
+
+      if (!response.body) {
+        throw new Error('No response body')
+      }
+
+      const executionId = response.headers.get('X-Execution-Id')
+      if (executionId) {
+        currentExecutionRef.current = { workflowId, executionId }
+      }
+
+      const reader = response.body.getReader()
+      await processSSEStream(reader, callbacks, 'Run-from-block')
+    } catch (error: any) {
+      if (error.name === 'AbortError') {
+        logger.info('Run-from-block execution cancelled')
+        callbacks.onExecutionCancelled?.({ duration: 0 })
+      } else {
+        logger.error('Run-from-block execution error:', error)
+        callbacks.onExecutionError?.({
+          error: error.message || 'Unknown error',
+          duration: 0,
+        })
+      }
+      throw error
+    } finally {
+      abortControllerRef.current = null
+      currentExecutionRef.current = null
+    }
+  }, [])
+
   const cancel = useCallback(() => {
     const execution = currentExecutionRef.current
     if (execution) {
@@ -239,6 +303,7 @@ export function useExecutionStream() {
 
   return {
     execute,
+    executeFromBlock,
     cancel,
   }
 }

apps/sim/lib/execution/isolated-vm-worker.cjs

@@ -8,7 +8,7 @@ const ivm = require('isolated-vm')
 const USER_CODE_START_LINE = 4
 const pendingFetches = new Map()
 let fetchIdCounter = 0
-const FETCH_TIMEOUT_MS = 300000 // 5 minutes
+const FETCH_TIMEOUT_MS = 30000
 
 /**
  * Extract line and column from error stack or message

apps/sim/lib/workflows/autolayout/containers.ts

@@ -34,7 +34,6 @@ export function layoutContainers(
       : DEFAULT_CONTAINER_HORIZONTAL_SPACING,
     verticalSpacing: options.verticalSpacing ?? DEFAULT_VERTICAL_SPACING,
     padding: { x: CONTAINER_PADDING_X, y: CONTAINER_PADDING_Y },
-    gridSize: options.gridSize,
   }
 
   for (const [parentId, childIds] of children.entries()) {
@@ -57,15 +56,18 @@ export function layoutContainers(
       continue
     }
 
+    // Use the shared core layout function with container options
     const { nodes, dimensions } = layoutBlocksCore(childBlocks, childEdges, {
       isContainer: true,
       layoutOptions: containerOptions,
     })
 
+    // Apply positions back to blocks
     for (const node of nodes.values()) {
       blocks[node.id].position = node.position
     }
 
+    // Update container dimensions
     const calculatedWidth = dimensions.width
     const calculatedHeight = dimensions.height
 

apps/sim/lib/workflows/autolayout/core.ts

@@ -9,7 +9,6 @@ import {
   getBlockMetrics,
   normalizePositions,
   prepareBlockMetrics,
-  snapNodesToGrid,
 } from '@/lib/workflows/autolayout/utils'
 import { BLOCK_DIMENSIONS, HANDLE_POSITIONS } from '@/lib/workflows/blocks/block-dimensions'
 import { EDGE } from '@/executor/constants'
@@ -85,6 +84,7 @@ export function assignLayers(
 ): Map<string, GraphNode> {
   const nodes = new Map<string, GraphNode>()
 
+  // Initialize nodes
   for (const [id, block] of Object.entries(blocks)) {
     nodes.set(id, {
       id,
@@ -97,6 +97,7 @@ export function assignLayers(
     })
   }
 
+  // Build a map of target node -> edges coming into it (to check sourceHandle later)
   const incomingEdgesMap = new Map<string, Edge[]>()
   for (const edge of edges) {
     if (!incomingEdgesMap.has(edge.target)) {
@@ -105,6 +106,7 @@ export function assignLayers(
     incomingEdgesMap.get(edge.target)!.push(edge)
   }
 
+  // Build adjacency from edges
   for (const edge of edges) {
     const sourceNode = nodes.get(edge.source)
     const targetNode = nodes.get(edge.target)
@@ -115,6 +117,7 @@ export function assignLayers(
     }
   }
 
+  // Find starter nodes (no incoming edges)
   const starterNodes = Array.from(nodes.values()).filter((node) => node.incoming.size === 0)
 
   if (starterNodes.length === 0 && nodes.size > 0) {
@@ -123,6 +126,7 @@ export function assignLayers(
     logger.warn('No starter blocks found, using first block as starter', { blockId: firstNode.id })
   }
 
+  // Topological sort using Kahn's algorithm
   const inDegreeCount = new Map<string, number>()
 
   for (const node of nodes.values()) {
@@ -140,6 +144,8 @@ export function assignLayers(
     const node = nodes.get(nodeId)!
     processed.add(nodeId)
 
+    // Calculate layer based on max incoming layer + 1
+    // For edges from subflow ends, add the subflow's internal depth (minus 1 to avoid double-counting)
     if (node.incoming.size > 0) {
       let maxEffectiveLayer = -1
       const incomingEdges = incomingEdgesMap.get(nodeId) || []
@@ -147,11 +153,16 @@ export function assignLayers(
       for (const incomingId of node.incoming) {
         const incomingNode = nodes.get(incomingId)
         if (incomingNode) {
+          // Find edges from this incoming node to check if it's a subflow end edge
           const edgesFromSource = incomingEdges.filter((e) => e.source === incomingId)
           let additionalDepth = 0
 
+          // Check if any edge from this source is a subflow end edge
           const hasSubflowEndEdge = edgesFromSource.some(isSubflowEndEdge)
           if (hasSubflowEndEdge && subflowDepths) {
+            // Get the internal depth of the subflow
+            // Subtract 1 because the +1 at the end of layer calculation already accounts for one layer
+            // E.g., if subflow has 2 internal layers (depth=2), we add 1 extra so total offset is 2
             const depth = subflowDepths.get(incomingId) ?? 1
             additionalDepth = Math.max(0, depth - 1)
           }
@@ -163,6 +174,7 @@ export function assignLayers(
       node.layer = maxEffectiveLayer + 1
     }
 
+    // Add outgoing nodes when all dependencies processed
     for (const targetId of node.outgoing) {
       const currentCount = inDegreeCount.get(targetId) || 0
       inDegreeCount.set(targetId, currentCount - 1)
@@ -173,6 +185,7 @@ export function assignLayers(
     }
   }
 
+  // Handle isolated nodes
   for (const node of nodes.values()) {
     if (!processed.has(node.id)) {
       logger.debug('Isolated node detected, assigning to layer 0', { blockId: node.id })
@@ -211,6 +224,7 @@ function resolveVerticalOverlaps(nodes: GraphNode[], verticalSpacing: number): v
     hasOverlap = false
     iteration++
 
+    // Group nodes by layer for same-layer overlap resolution
     const nodesByLayer = new Map<number, GraphNode[]>()
     for (const node of nodes) {
       if (!nodesByLayer.has(node.layer)) {
@@ -219,9 +233,11 @@ function resolveVerticalOverlaps(nodes: GraphNode[], verticalSpacing: number): v
       nodesByLayer.get(node.layer)!.push(node)
     }
 
+    // Process each layer independently
     for (const [layer, layerNodes] of nodesByLayer) {
       if (layerNodes.length < 2) continue
 
+      // Sort by Y position for consistent processing
       layerNodes.sort((a, b) => a.position.y - b.position.y)
 
       for (let i = 0; i < layerNodes.length - 1; i++) {
@@ -286,6 +302,7 @@ export function calculatePositions(
 
   const layerNumbers = Array.from(layers.keys()).sort((a, b) => a - b)
 
+  // Calculate max width for each layer
   const layerWidths = new Map<number, number>()
   for (const layerNum of layerNumbers) {
     const nodesInLayer = layers.get(layerNum)!
@@ -293,6 +310,7 @@ export function calculatePositions(
     layerWidths.set(layerNum, maxWidth)
   }
 
+  // Calculate cumulative X positions for each layer based on actual widths
   const layerXPositions = new Map<number, number>()
   let cumulativeX = padding.x
 
@@ -301,6 +319,7 @@ export function calculatePositions(
     cumulativeX += layerWidths.get(layerNum)! + horizontalSpacing
   }
 
+  // Build a flat map of all nodes for quick lookups
   const allNodes = new Map<string, GraphNode>()
   for (const nodesInLayer of layers.values()) {
     for (const node of nodesInLayer) {
@@ -308,6 +327,7 @@ export function calculatePositions(
     }
   }
 
+  // Build incoming edges map for handle lookups
   const incomingEdgesMap = new Map<string, Edge[]>()
   for (const edge of edges) {
     if (!incomingEdgesMap.has(edge.target)) {
@@ -316,16 +336,20 @@ export function calculatePositions(
     incomingEdgesMap.get(edge.target)!.push(edge)
   }
 
+  // Position nodes layer by layer, aligning with connected predecessors
   for (const layerNum of layerNumbers) {
     const nodesInLayer = layers.get(layerNum)!
     const xPosition = layerXPositions.get(layerNum)!
 
+    // Separate containers and non-containers
     const containersInLayer = nodesInLayer.filter(isContainerBlock)
     const nonContainersInLayer = nodesInLayer.filter((n) => !isContainerBlock(n))
 
+    // For the first layer (layer 0), position sequentially from padding.y
     if (layerNum === 0) {
       let yOffset = padding.y
 
+      // Sort containers by height for visual balance
       containersInLayer.sort((a, b) => b.metrics.height - a.metrics.height)
 
       for (const node of containersInLayer) {
@@ -337,6 +361,7 @@ export function calculatePositions(
         yOffset += CONTAINER_VERTICAL_CLEARANCE
       }
 
+      // Sort non-containers by outgoing connections
       nonContainersInLayer.sort((a, b) => b.outgoing.size - a.outgoing.size)
 
       for (const node of nonContainersInLayer) {
@@ -346,7 +371,9 @@ export function calculatePositions(
       continue
     }
 
+    // For subsequent layers, align with connected predecessors (handle-to-handle)
     for (const node of [...containersInLayer, ...nonContainersInLayer]) {
+      // Find the bottommost predecessor handle Y (highest value) and align to it
       let bestSourceHandleY = -1
       let bestEdge: Edge | null = null
       const incomingEdges = incomingEdgesMap.get(node.id) || []
@@ -354,6 +381,7 @@ export function calculatePositions(
       for (const edge of incomingEdges) {
         const predecessor = allNodes.get(edge.source)
         if (predecessor) {
+          // Calculate actual source handle Y position based on block type and handle
           const sourceHandleOffset = getSourceHandleYOffset(predecessor.block, edge.sourceHandle)
           const sourceHandleY = predecessor.position.y + sourceHandleOffset
 
@@ -364,16 +392,20 @@ export function calculatePositions(
         }
       }
 
+      // If no predecessors found (shouldn't happen for layer > 0), use padding
       if (bestSourceHandleY < 0) {
         bestSourceHandleY = padding.y + HANDLE_POSITIONS.DEFAULT_Y_OFFSET
       }
 
+      // Calculate the target handle Y offset for this node
       const targetHandleOffset = getTargetHandleYOffset(node.block, bestEdge?.targetHandle)
 
+      // Position node so its target handle aligns with the source handle Y
       node.position = { x: xPosition, y: bestSourceHandleY - targetHandleOffset }
     }
   }
 
+  // Resolve vertical overlaps within layers (X overlaps prevented by cumulative positioning)
   resolveVerticalOverlaps(Array.from(layers.values()).flat(), verticalSpacing)
 }
 
@@ -403,7 +435,7 @@ export function layoutBlocksCore(
     return { nodes: new Map(), dimensions: { width: 0, height: 0 } }
   }
 
-  const layoutOptions: LayoutOptions =
+  const layoutOptions =
     options.layoutOptions ??
     (options.isContainer ? CONTAINER_LAYOUT_OPTIONS : DEFAULT_LAYOUT_OPTIONS)
 
@@ -420,13 +452,7 @@ export function layoutBlocksCore(
   calculatePositions(layers, edges, layoutOptions)
 
   // 5. Normalize positions
-  let dimensions = normalizePositions(nodes, { isContainer: options.isContainer })
-
-  // 6. Snap to grid if gridSize is specified (recalculates dimensions)
-  const snappedDimensions = snapNodesToGrid(nodes, layoutOptions.gridSize)
-  if (snappedDimensions) {
-    dimensions = snappedDimensions
-  }
+  const dimensions = normalizePositions(nodes, { isContainer: options.isContainer })
 
   return { nodes, dimensions }
 }

apps/sim/lib/workflows/autolayout/index.ts

@@ -36,13 +36,14 @@ export function applyAutoLayout(
     const horizontalSpacing = options.horizontalSpacing ?? DEFAULT_HORIZONTAL_SPACING
     const verticalSpacing = options.verticalSpacing ?? DEFAULT_VERTICAL_SPACING
 
+    // Pre-calculate container dimensions by laying out their children (bottom-up)
+    // This ensures accurate widths/heights before root-level layout
     prepareContainerDimensions(
       blocksCopy,
       edges,
       layoutBlocksCore,
       horizontalSpacing,
-      verticalSpacing,
-      options.gridSize
+      verticalSpacing
     )
 
     const { root: rootBlockIds } = getBlocksByParent(blocksCopy)
@@ -57,6 +58,8 @@ export function applyAutoLayout(
       (edge) => layoutRootIds.includes(edge.source) && layoutRootIds.includes(edge.target)
     )
 
+    // Calculate subflow depths before laying out root blocks
+    // This ensures blocks connected to subflow ends are positioned correctly
     const subflowDepths = calculateSubflowDepths(blocksCopy, edges, assignLayers)
 
     if (Object.keys(rootBlocks).length > 0) {
@@ -92,12 +95,13 @@ export function applyAutoLayout(
 }
 
 export type { TargetedLayoutOptions } from '@/lib/workflows/autolayout/targeted'
+// Function exports
 export { applyTargetedLayout } from '@/lib/workflows/autolayout/targeted'
+// Type exports
 export type { Edge, LayoutOptions, LayoutResult } from '@/lib/workflows/autolayout/types'
 export {
   getBlockMetrics,
   isContainerType,
   shouldSkipAutoLayout,
-  snapPositionToGrid,
   transferBlockHeights,
 } from '@/lib/workflows/autolayout/utils'

apps/sim/lib/workflows/autolayout/targeted.ts

@@ -1,3 +1,4 @@
+import { createLogger } from '@sim/logger'
 import {
   CONTAINER_PADDING,
   DEFAULT_HORIZONTAL_SPACING,
@@ -13,11 +14,12 @@ import {
   isContainerType,
   prepareContainerDimensions,
   shouldSkipAutoLayout,
-  snapPositionToGrid,
 } from '@/lib/workflows/autolayout/utils'
 import { CONTAINER_DIMENSIONS } from '@/lib/workflows/blocks/block-dimensions'
 import type { BlockState } from '@/stores/workflows/workflow/types'
 
+const logger = createLogger('AutoLayout:Targeted')
+
 export interface TargetedLayoutOptions extends LayoutOptions {
   changedBlockIds: string[]
   verticalSpacing?: number
@@ -37,7 +39,6 @@ export function applyTargetedLayout(
     changedBlockIds,
     verticalSpacing = DEFAULT_VERTICAL_SPACING,
     horizontalSpacing = DEFAULT_HORIZONTAL_SPACING,
-    gridSize,
   } = options
 
   if (!changedBlockIds || changedBlockIds.length === 0) {
@@ -47,17 +48,19 @@ export function applyTargetedLayout(
   const changedSet = new Set(changedBlockIds)
   const blocksCopy: Record<string, BlockState> = JSON.parse(JSON.stringify(blocks))
 
+  // Pre-calculate container dimensions by laying out their children (bottom-up)
+  // This ensures accurate widths/heights before root-level layout
   prepareContainerDimensions(
     blocksCopy,
     edges,
     layoutBlocksCore,
     horizontalSpacing,
-    verticalSpacing,
-    gridSize
+    verticalSpacing
   )
 
   const groups = getBlocksByParent(blocksCopy)
 
+  // Calculate subflow depths before layout to properly position blocks after subflow ends
   const subflowDepths = calculateSubflowDepths(blocksCopy, edges, assignLayers)
 
   layoutGroup(
@@ -68,8 +71,7 @@ export function applyTargetedLayout(
     changedSet,
     verticalSpacing,
     horizontalSpacing,
-    subflowDepths,
-    gridSize
+    subflowDepths
   )
 
   for (const [parentId, childIds] of groups.children.entries()) {
@@ -81,8 +83,7 @@ export function applyTargetedLayout(
       changedSet,
       verticalSpacing,
       horizontalSpacing,
-      subflowDepths,
-      gridSize
+      subflowDepths
     )
   }
 
@@ -100,8 +101,7 @@ function layoutGroup(
   changedSet: Set<string>,
   verticalSpacing: number,
   horizontalSpacing: number,
-  subflowDepths: Map<string, number>,
-  gridSize?: number
+  subflowDepths: Map<string, number>
 ): void {
   if (childIds.length === 0) return
 
@@ -116,6 +116,7 @@ function layoutGroup(
     return
   }
 
+  // Determine which blocks need repositioning
   const requestedLayout = layoutEligibleChildIds.filter((id) => {
     const block = blocks[id]
     if (!block) return false
@@ -140,6 +141,7 @@ function layoutGroup(
     return
   }
 
+  // Store old positions for anchor calculation
   const oldPositions = new Map<string, { x: number; y: number }>()
   for (const id of layoutEligibleChildIds) {
     const block = blocks[id]
@@ -147,6 +149,8 @@ function layoutGroup(
     oldPositions.set(id, { ...block.position })
   }
 
+  // Compute layout positions using core function
+  // Only pass subflowDepths for root-level layout (not inside containers)
   const layoutPositions = computeLayoutPositions(
     layoutEligibleChildIds,
     blocks,
@@ -154,8 +158,7 @@ function layoutGroup(
     parentBlock,
     horizontalSpacing,
     verticalSpacing,
-    parentId === null ? subflowDepths : undefined,
-    gridSize
+    parentId === null ? subflowDepths : undefined
   )
 
   if (layoutPositions.size === 0) {
@@ -165,6 +168,7 @@ function layoutGroup(
     return
   }
 
+  // Find anchor block (unchanged block with a layout position)
   let offsetX = 0
   let offsetY = 0
 
@@ -181,16 +185,20 @@ function layoutGroup(
     }
   }
 
+  // Apply new positions only to blocks that need layout
   for (const id of needsLayout) {
     const block = blocks[id]
     const newPos = layoutPositions.get(id)
     if (!block || !newPos) continue
-    block.position = snapPositionToGrid({ x: newPos.x + offsetX, y: newPos.y + offsetY }, gridSize)
+    block.position = {
+      x: newPos.x + offsetX,
+      y: newPos.y + offsetY,
+    }
   }
 }
 
 /**
- * Computes layout positions for a subset of blocks using the core layout function
+ * Computes layout positions for a subset of blocks using the core layout
  */
 function computeLayoutPositions(
   childIds: string[],
@@ -199,8 +207,7 @@ function computeLayoutPositions(
   parentBlock: BlockState | undefined,
   horizontalSpacing: number,
   verticalSpacing: number,
-  subflowDepths?: Map<string, number>,
-  gridSize?: number
+  subflowDepths?: Map<string, number>
 ): Map<string, { x: number; y: number }> {
   const subsetBlocks: Record<string, BlockState> = {}
   for (const id of childIds) {
@@ -221,11 +228,11 @@ function computeLayoutPositions(
     layoutOptions: {
       horizontalSpacing: isContainer ? horizontalSpacing * 0.85 : horizontalSpacing,
       verticalSpacing,
-      gridSize,
     },
     subflowDepths,
   })
 
+  // Update parent container dimensions if applicable
   if (parentBlock) {
     parentBlock.data = {
       ...parentBlock.data,
@@ -234,6 +241,7 @@ function computeLayoutPositions(
     }
   }
 
+  // Convert nodes to position map
   const positions = new Map<string, { x: number; y: number }>()
   for (const node of nodes.values()) {
     positions.set(node.id, { x: node.position.x, y: node.position.y })

apps/sim/lib/workflows/autolayout/types.ts

@@ -7,7 +7,6 @@ export interface LayoutOptions {
   horizontalSpacing?: number
   verticalSpacing?: number
   padding?: { x: number; y: number }
-  gridSize?: number
 }
 
 export interface LayoutResult {

apps/sim/lib/workflows/autolayout/utils.ts

@@ -18,61 +18,6 @@ function resolveNumeric(value: number | undefined, fallback: number): number {
   return typeof value === 'number' && Number.isFinite(value) ? value : fallback
 }
 
-/**
- * Snaps a single coordinate value to the nearest grid position
- */
-function snapToGrid(value: number, gridSize: number): number {
-  return Math.round(value / gridSize) * gridSize
-}
-
-/**
- * Snaps a position to the nearest grid point.
- * Returns the original position if gridSize is 0 or not provided.
- */
-export function snapPositionToGrid(
-  position: { x: number; y: number },
-  gridSize: number | undefined
-): { x: number; y: number } {
-  if (!gridSize || gridSize <= 0) {
-    return position
-  }
-  return {
-    x: snapToGrid(position.x, gridSize),
-    y: snapToGrid(position.y, gridSize),
-  }
-}
-
-/**
- * Snaps all node positions in a graph to grid positions and returns updated dimensions.
- * Returns null if gridSize is not set or no snapping was needed.
- */
-export function snapNodesToGrid(
-  nodes: Map<string, GraphNode>,
-  gridSize: number | undefined
-): { width: number; height: number } | null {
-  if (!gridSize || gridSize <= 0 || nodes.size === 0) {
-    return null
-  }
-
-  let minX = Number.POSITIVE_INFINITY
-  let minY = Number.POSITIVE_INFINITY
-  let maxX = Number.NEGATIVE_INFINITY
-  let maxY = Number.NEGATIVE_INFINITY
-
-  for (const node of nodes.values()) {
-    node.position = snapPositionToGrid(node.position, gridSize)
-    minX = Math.min(minX, node.position.x)
-    minY = Math.min(minY, node.position.y)
-    maxX = Math.max(maxX, node.position.x + node.metrics.width)
-    maxY = Math.max(maxY, node.position.y + node.metrics.height)
-  }
-
-  return {
-    width: maxX - minX + CONTAINER_PADDING * 2,
-    height: maxY - minY + CONTAINER_PADDING * 2,
-  }
-}
-
 /**
  * Checks if a block type is a container (loop or parallel)
  */
@@ -369,7 +314,6 @@ export type LayoutFunction = (
       horizontalSpacing?: number
       verticalSpacing?: number
       padding?: { x: number; y: number }
-      gridSize?: number
     }
     subflowDepths?: Map<string, number>
   }
@@ -385,15 +329,13 @@ export type LayoutFunction = (
  * @param layoutFn - The layout function to use for calculating dimensions
  * @param horizontalSpacing - Horizontal spacing between blocks
  * @param verticalSpacing - Vertical spacing between blocks
- * @param gridSize - Optional grid size for snap-to-grid
  */
 export function prepareContainerDimensions(
   blocks: Record<string, BlockState>,
   edges: Edge[],
   layoutFn: LayoutFunction,
   horizontalSpacing: number,
-  verticalSpacing: number,
-  gridSize?: number
+  verticalSpacing: number
 ): void {
   const { children } = getBlocksByParent(blocks)
 
@@ -460,7 +402,6 @@ export function prepareContainerDimensions(
       layoutOptions: {
         horizontalSpacing: horizontalSpacing * 0.85,
         verticalSpacing,
-        gridSize,
       },
     })
 

apps/sim/lib/workflows/executor/execution-core.ts

@@ -23,9 +23,11 @@ import type {
   ContextExtensions,
   ExecutionCallbacks,
   IterationContext,
+  SerializableExecutionState,
 } from '@/executor/execution/types'
 import type { ExecutionResult, NormalizedBlockOutput } from '@/executor/types'
 import { hasExecutionResult } from '@/executor/utils/errors'
+import { buildParallelSentinelEndId, buildSentinelEndId } from '@/executor/utils/subflow-utils'
 import { Serializer } from '@/serializer'
 
 const logger = createLogger('ExecutionCore')
@@ -40,6 +42,12 @@ export interface ExecuteWorkflowCoreOptions {
   abortSignal?: AbortSignal
   includeFileBase64?: boolean
   base64MaxBytes?: number
+  stopAfterBlockId?: string
+  /** Run-from-block mode: execute starting from a specific block using cached upstream outputs */
+  runFromBlock?: {
+    startBlockId: string
+    sourceSnapshot: SerializableExecutionState
+  }
 }
 
 function parseVariableValueByType(value: unknown, type: string): unknown {
@@ -114,6 +122,8 @@ export async function executeWorkflowCore(
     abortSignal,
     includeFileBase64,
     base64MaxBytes,
+    stopAfterBlockId,
+    runFromBlock,
   } = options
   const { metadata, workflow, input, workflowVariables, selectedOutputs } = snapshot
   const { requestId, workflowId, userId, triggerType, executionId, triggerBlockId, useDraftState } =
@@ -246,6 +256,16 @@ export async function executeWorkflowCore(
 
     processedInput = input || {}
 
+    // Resolve stopAfterBlockId for loop/parallel containers to their sentinel-end IDs
+    let resolvedStopAfterBlockId = stopAfterBlockId
+    if (stopAfterBlockId) {
+      if (serializedWorkflow.loops?.[stopAfterBlockId]) {
+        resolvedStopAfterBlockId = buildSentinelEndId(stopAfterBlockId)
+      } else if (serializedWorkflow.parallels?.[stopAfterBlockId]) {
+        resolvedStopAfterBlockId = buildParallelSentinelEndId(stopAfterBlockId)
+      }
+    }
+
     // Create and execute workflow with callbacks
     if (resumeFromSnapshot) {
       logger.info(`[${requestId}] Resume execution detected`, {
@@ -296,6 +316,7 @@ export async function executeWorkflowCore(
       abortSignal,
       includeFileBase64,
       base64MaxBytes,
+      stopAfterBlockId: resolvedStopAfterBlockId,
     }
 
     const executorInstance = new Executor({
@@ -318,10 +339,13 @@ export async function executeWorkflowCore(
       }
     }
 
-    const result = (await executorInstance.execute(
-      workflowId,
-      resolvedTriggerBlockId
-    )) as ExecutionResult
+    const result = runFromBlock
+      ? ((await executorInstance.executeFromBlock(
+          workflowId,
+          runFromBlock.startBlockId,
+          runFromBlock.sourceSnapshot
+        )) as ExecutionResult)
+      : ((await executorInstance.execute(workflowId, resolvedTriggerBlockId)) as ExecutionResult)
 
     // Build trace spans for logging from the full execution result
     const { traceSpans, totalDuration } = buildTraceSpans(result)

apps/sim/lib/workflows/executor/execution-events.ts

@@ -180,3 +180,140 @@ export function formatSSEEvent(event: ExecutionEvent): string {
 export function encodeSSEEvent(event: ExecutionEvent): Uint8Array {
   return new TextEncoder().encode(formatSSEEvent(event))
 }
+
+/**
+ * Options for creating SSE execution callbacks
+ */
+export interface SSECallbackOptions {
+  executionId: string
+  workflowId: string
+  controller: ReadableStreamDefaultController<Uint8Array>
+  isStreamClosed: () => boolean
+  setStreamClosed: () => void
+}
+
+/**
+ * Creates SSE callbacks for workflow execution streaming
+ */
+export function createSSECallbacks(options: SSECallbackOptions) {
+  const { executionId, workflowId, controller, isStreamClosed, setStreamClosed } = options
+
+  const sendEvent = (event: ExecutionEvent) => {
+    if (isStreamClosed()) return
+    try {
+      controller.enqueue(encodeSSEEvent(event))
+    } catch {
+      setStreamClosed()
+    }
+  }
+
+  const onBlockStart = async (
+    blockId: string,
+    blockName: string,
+    blockType: string,
+    iterationContext?: { iterationCurrent: number; iterationTotal: number; iterationType: string }
+  ) => {
+    sendEvent({
+      type: 'block:started',
+      timestamp: new Date().toISOString(),
+      executionId,
+      workflowId,
+      data: {
+        blockId,
+        blockName,
+        blockType,
+        ...(iterationContext && {
+          iterationCurrent: iterationContext.iterationCurrent,
+          iterationTotal: iterationContext.iterationTotal,
+          iterationType: iterationContext.iterationType as any,
+        }),
+      },
+    })
+  }
+
+  const onBlockComplete = async (
+    blockId: string,
+    blockName: string,
+    blockType: string,
+    callbackData: { input?: unknown; output: any; executionTime: number },
+    iterationContext?: { iterationCurrent: number; iterationTotal: number; iterationType: string }
+  ) => {
+    const hasError = callbackData.output?.error
+    const iterationData = iterationContext
+      ? {
+          iterationCurrent: iterationContext.iterationCurrent,
+          iterationTotal: iterationContext.iterationTotal,
+          iterationType: iterationContext.iterationType as any,
+        }
+      : {}
+
+    if (hasError) {
+      sendEvent({
+        type: 'block:error',
+        timestamp: new Date().toISOString(),
+        executionId,
+        workflowId,
+        data: {
+          blockId,
+          blockName,
+          blockType,
+          input: callbackData.input,
+          error: callbackData.output.error,
+          durationMs: callbackData.executionTime || 0,
+          ...iterationData,
+        },
+      })
+    } else {
+      sendEvent({
+        type: 'block:completed',
+        timestamp: new Date().toISOString(),
+        executionId,
+        workflowId,
+        data: {
+          blockId,
+          blockName,
+          blockType,
+          input: callbackData.input,
+          output: callbackData.output,
+          durationMs: callbackData.executionTime || 0,
+          ...iterationData,
+        },
+      })
+    }
+  }
+
+  const onStream = async (streamingExecution: unknown) => {
+    const streamingExec = streamingExecution as { stream: ReadableStream; execution: any }
+    const blockId = streamingExec.execution?.blockId
+    const reader = streamingExec.stream.getReader()
+    const decoder = new TextDecoder()
+
+    try {
+      while (true) {
+        const { done, value } = await reader.read()
+        if (done) break
+        const chunk = decoder.decode(value, { stream: true })
+        sendEvent({
+          type: 'stream:chunk',
+          timestamp: new Date().toISOString(),
+          executionId,
+          workflowId,
+          data: { blockId, chunk },
+        })
+      }
+      sendEvent({
+        type: 'stream:done',
+        timestamp: new Date().toISOString(),
+        executionId,
+        workflowId,
+        data: { blockId },
+      })
+    } finally {
+      try {
+        reader.releaseLock()
+      } catch {}
+    }
+  }
+
+  return { sendEvent, onBlockStart, onBlockComplete, onStream }
+}

apps/sim/providers/azure-openai/index.ts

@@ -102,7 +102,7 @@ export const azureOpenAIProvider: ProviderConfig = {
     }
 
     if (request.temperature !== undefined) payload.temperature = request.temperature
-    if (request.maxTokens != null) payload.max_completion_tokens = request.maxTokens
+    if (request.maxTokens !== undefined) payload.max_tokens = request.maxTokens
 
     if (request.reasoningEffort !== undefined) payload.reasoning_effort = request.reasoningEffort
     if (request.verbosity !== undefined) payload.verbosity = request.verbosity

apps/sim/providers/cerebras/index.ts

@@ -77,7 +77,7 @@ export const cerebrasProvider: ProviderConfig = {
         messages: allMessages,
       }
       if (request.temperature !== undefined) payload.temperature = request.temperature
-      if (request.maxTokens != null) payload.max_completion_tokens = request.maxTokens
+      if (request.maxTokens !== undefined) payload.max_tokens = request.maxTokens
       if (request.responseFormat) {
         payload.response_format = {
           type: 'json_schema',

apps/sim/providers/deepseek/index.ts

@@ -81,7 +81,7 @@ export const deepseekProvider: ProviderConfig = {
       }
 
       if (request.temperature !== undefined) payload.temperature = request.temperature
-      if (request.maxTokens != null) payload.max_tokens = request.maxTokens
+      if (request.maxTokens !== undefined) payload.max_tokens = request.maxTokens
 
       let preparedTools: ReturnType<typeof prepareToolsWithUsageControl> | null = null
 

apps/sim/providers/gemini/core.ts

@@ -349,7 +349,7 @@ export async function executeGeminiRequest(
     if (request.temperature !== undefined) {
       geminiConfig.temperature = request.temperature
     }
-    if (request.maxTokens != null) {
+    if (request.maxTokens !== undefined) {
       geminiConfig.maxOutputTokens = request.maxTokens
     }
     if (systemInstruction) {

apps/sim/providers/google/utils.ts

@@ -123,21 +123,17 @@ export function extractFunctionCallPart(candidate: Candidate | undefined): Part
 }
 
 /**
- * Converts usage metadata from SDK response to our format.
- * Per Gemini docs, total = promptTokenCount + candidatesTokenCount + toolUsePromptTokenCount + thoughtsTokenCount
- * We include toolUsePromptTokenCount in input and thoughtsTokenCount in output for correct billing.
+ * Converts usage metadata from SDK response to our format
  */
 export function convertUsageMetadata(
   usageMetadata: GenerateContentResponseUsageMetadata | undefined
 ): GeminiUsage {
-  const thoughtsTokenCount = usageMetadata?.thoughtsTokenCount ?? 0
-  const toolUsePromptTokenCount = usageMetadata?.toolUsePromptTokenCount ?? 0
-  const promptTokenCount = (usageMetadata?.promptTokenCount ?? 0) + toolUsePromptTokenCount
-  const candidatesTokenCount = (usageMetadata?.candidatesTokenCount ?? 0) + thoughtsTokenCount
+  const promptTokenCount = usageMetadata?.promptTokenCount ?? 0
+  const candidatesTokenCount = usageMetadata?.candidatesTokenCount ?? 0
   return {
     promptTokenCount,
     candidatesTokenCount,
-    totalTokenCount: usageMetadata?.totalTokenCount ?? 0,
+    totalTokenCount: usageMetadata?.totalTokenCount ?? promptTokenCount + candidatesTokenCount,
   }
 }
 

apps/sim/providers/groq/index.ts

@@ -74,7 +74,7 @@ export const groqProvider: ProviderConfig = {
     }
 
     if (request.temperature !== undefined) payload.temperature = request.temperature
-    if (request.maxTokens != null) payload.max_completion_tokens = request.maxTokens
+    if (request.maxTokens !== undefined) payload.max_tokens = request.maxTokens
 
     if (request.responseFormat) {
       payload.response_format = {

apps/sim/providers/mistral/index.ts

@@ -91,7 +91,7 @@ export const mistralProvider: ProviderConfig = {
     }
 
     if (request.temperature !== undefined) payload.temperature = request.temperature
-    if (request.maxTokens != null) payload.max_tokens = request.maxTokens
+    if (request.maxTokens !== undefined) payload.max_tokens = request.maxTokens
 
     if (request.responseFormat) {
       payload.response_format = {

apps/sim/providers/models.ts

@@ -1130,7 +1130,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
     id: 'cerebras',
     name: 'Cerebras',
     description: 'Cerebras Cloud LLMs',
-    defaultModel: 'cerebras/gpt-oss-120b',
+    defaultModel: 'cerebras/llama-3.3-70b',
     modelPatterns: [/^cerebras/],
     icon: CerebrasIcon,
     capabilities: {
@@ -1138,64 +1138,44 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
     },
     models: [
       {
-        id: 'cerebras/gpt-oss-120b',
-        pricing: {
-          input: 0.35,
-          output: 0.75,
-          updatedAt: '2026-01-27',
-        },
-        capabilities: {},
-        contextWindow: 131000,
-      },
-      {
-        id: 'cerebras/llama3.1-8b',
+        id: 'cerebras/llama-3.1-8b',
         pricing: {
           input: 0.1,
           output: 0.1,
-          updatedAt: '2026-01-27',
+          updatedAt: '2025-10-11',
         },
         capabilities: {},
         contextWindow: 32000,
       },
       {
-        id: 'cerebras/llama-3.3-70b',
+        id: 'cerebras/llama-3.1-70b',
         pricing: {
-          input: 0.85,
-          output: 1.2,
-          updatedAt: '2026-01-27',
+          input: 0.6,
+          output: 0.6,
+          updatedAt: '2025-10-11',
         },
         capabilities: {},
         contextWindow: 128000,
       },
       {
-        id: 'cerebras/qwen-3-32b',
-        pricing: {
-          input: 0.4,
-          output: 0.8,
-          updatedAt: '2026-01-27',
-        },
-        capabilities: {},
-        contextWindow: 131000,
-      },
-      {
-        id: 'cerebras/qwen-3-235b-a22b-instruct-2507',
+        id: 'cerebras/llama-3.3-70b',
         pricing: {
           input: 0.6,
-          output: 1.2,
-          updatedAt: '2026-01-27',
+          output: 0.6,
+          updatedAt: '2025-10-11',
         },
         capabilities: {},
-        contextWindow: 131000,
+        contextWindow: 128000,
       },
       {
-        id: 'cerebras/zai-glm-4.7',
+        id: 'cerebras/llama-4-scout-17b-16e-instruct',
         pricing: {
-          input: 2.25,
-          output: 2.75,
-          updatedAt: '2026-01-27',
+          input: 0.11,
+          output: 0.34,
+          updatedAt: '2025-10-11',
         },
         capabilities: {},
-        contextWindow: 131000,
+        contextWindow: 10000000,
       },
     ],
   },
@@ -1214,8 +1194,8 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         id: 'groq/openai/gpt-oss-120b',
         pricing: {
           input: 0.15,
-          output: 0.6,
-          updatedAt: '2026-01-27',
+          output: 0.75,
+          updatedAt: '2025-10-11',
         },
         capabilities: {},
         contextWindow: 131072,
@@ -1223,29 +1203,9 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
       {
         id: 'groq/openai/gpt-oss-20b',
         pricing: {
-          input: 0.075,
-          output: 0.3,
-          updatedAt: '2026-01-27',
-        },
-        capabilities: {},
-        contextWindow: 131072,
-      },
-      {
-        id: 'groq/openai/gpt-oss-safeguard-20b',
-        pricing: {
-          input: 0.075,
-          output: 0.3,
-          updatedAt: '2026-01-27',
-        },
-        capabilities: {},
-        contextWindow: 131072,
-      },
-      {
-        id: 'groq/qwen/qwen3-32b',
-        pricing: {
-          input: 0.29,
-          output: 0.59,
-          updatedAt: '2026-01-27',
+          input: 0.01,
+          output: 0.25,
+          updatedAt: '2025-10-11',
         },
         capabilities: {},
         contextWindow: 131072,
@@ -1255,7 +1215,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         pricing: {
           input: 0.05,
           output: 0.08,
-          updatedAt: '2026-01-27',
+          updatedAt: '2025-10-11',
         },
         capabilities: {},
         contextWindow: 131072,
@@ -1265,17 +1225,27 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         pricing: {
           input: 0.59,
           output: 0.79,
-          updatedAt: '2026-01-27',
+          updatedAt: '2025-10-11',
         },
         capabilities: {},
         contextWindow: 131072,
       },
       {
-        id: 'groq/meta-llama/llama-4-scout-17b-16e-instruct',
+        id: 'groq/llama-4-scout-17b-instruct',
         pricing: {
           input: 0.11,
           output: 0.34,
-          updatedAt: '2026-01-27',
+          updatedAt: '2025-10-11',
+        },
+        capabilities: {},
+        contextWindow: 131072,
+      },
+      {
+        id: 'groq/llama-4-maverick-17b-instruct',
+        pricing: {
+          input: 0.5,
+          output: 0.77,
+          updatedAt: '2025-10-11',
         },
         capabilities: {},
         contextWindow: 131072,
@@ -1283,9 +1253,9 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
       {
         id: 'groq/meta-llama/llama-4-maverick-17b-128e-instruct',
         pricing: {
-          input: 0.2,
-          output: 0.6,
-          updatedAt: '2026-01-27',
+          input: 0.5,
+          output: 0.77,
+          updatedAt: '2025-10-11',
         },
         capabilities: {},
         contextWindow: 131072,
@@ -1295,7 +1265,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         pricing: {
           input: 0.04,
           output: 0.04,
-          updatedAt: '2026-01-27',
+          updatedAt: '2025-10-11',
         },
         capabilities: {},
         contextWindow: 8192,
@@ -1305,37 +1275,27 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         pricing: {
           input: 0.59,
           output: 0.79,
-          updatedAt: '2026-01-27',
+          updatedAt: '2025-10-11',
         },
         capabilities: {},
         contextWindow: 128000,
       },
       {
-        id: 'groq/deepseek-r1-distill-qwen-32b',
-        pricing: {
-          input: 0.69,
-          output: 0.69,
-          updatedAt: '2026-01-27',
-        },
-        capabilities: {},
-        contextWindow: 128000,
-      },
-      {
-        id: 'groq/moonshotai/kimi-k2-instruct-0905',
+        id: 'groq/moonshotai/kimi-k2-instruct',
         pricing: {
           input: 1.0,
           output: 3.0,
-          updatedAt: '2026-01-27',
+          updatedAt: '2025-10-11',
         },
         capabilities: {},
-        contextWindow: 262144,
+        contextWindow: 131072,
       },
       {
         id: 'groq/meta-llama/llama-guard-4-12b',
         pricing: {
           input: 0.2,
           output: 0.2,
-          updatedAt: '2026-01-27',
+          updatedAt: '2025-10-11',
         },
         capabilities: {},
         contextWindow: 131072,

apps/sim/providers/ollama/index.ts

@@ -105,7 +105,7 @@ export const ollamaProvider: ProviderConfig = {
     }
 
     if (request.temperature !== undefined) payload.temperature = request.temperature
-    if (request.maxTokens != null) payload.max_tokens = request.maxTokens
+    if (request.maxTokens !== undefined) payload.max_tokens = request.maxTokens
 
     if (request.responseFormat) {
       payload.response_format = {

apps/sim/providers/openai/index.ts

@@ -81,7 +81,7 @@ export const openaiProvider: ProviderConfig = {
     }
 
     if (request.temperature !== undefined) payload.temperature = request.temperature
-    if (request.maxTokens != null) payload.max_completion_tokens = request.maxTokens
+    if (request.maxTokens !== undefined) payload.max_tokens = request.maxTokens
 
     if (request.reasoningEffort !== undefined) payload.reasoning_effort = request.reasoningEffort
     if (request.verbosity !== undefined) payload.verbosity = request.verbosity

apps/sim/providers/openrouter/index.ts

@@ -121,7 +121,7 @@ export const openRouterProvider: ProviderConfig = {
     }
 
     if (request.temperature !== undefined) payload.temperature = request.temperature
-    if (request.maxTokens != null) payload.max_tokens = request.maxTokens
+    if (request.maxTokens !== undefined) payload.max_tokens = request.maxTokens
 
     let preparedTools: ReturnType<typeof prepareToolsWithUsageControl> | null = null
     let hasActiveTools = false
@@ -516,7 +516,7 @@ export const openRouterProvider: ProviderConfig = {
         return streamingResult as StreamingExecution
       }
 
-      if (request.responseFormat && hasActiveTools) {
+      if (request.responseFormat && hasActiveTools && toolCalls.length > 0) {
         const finalPayload: any = {
           model: payload.model,
           messages: [...currentMessages],

apps/sim/providers/vllm/index.ts

@@ -135,7 +135,7 @@ export const vllmProvider: ProviderConfig = {
     }
 
     if (request.temperature !== undefined) payload.temperature = request.temperature
-    if (request.maxTokens != null) payload.max_completion_tokens = request.maxTokens
+    if (request.maxTokens !== undefined) payload.max_tokens = request.maxTokens
 
     if (request.responseFormat) {
       payload.response_format = {

apps/sim/providers/xai/index.ts

@@ -92,7 +92,7 @@ export const xAIProvider: ProviderConfig = {
     }
 
     if (request.temperature !== undefined) basePayload.temperature = request.temperature
-    if (request.maxTokens != null) basePayload.max_completion_tokens = request.maxTokens
+    if (request.maxTokens !== undefined) basePayload.max_tokens = request.maxTokens
     let preparedTools: ReturnType<typeof prepareToolsWithUsageControl> | null = null
 
     if (tools?.length) {

apps/sim/stores/execution/store.ts

@@ -35,4 +35,23 @@ export const useExecutionStore = create<ExecutionState & ExecutionActions>()((se
   },
   clearRunPath: () => set({ lastRunPath: new Map(), lastRunEdges: new Map() }),
   reset: () => set(initialState),
+
+  setLastExecutionSnapshot: (workflowId, snapshot) => {
+    const { lastExecutionSnapshots } = get()
+    const newSnapshots = new Map(lastExecutionSnapshots)
+    newSnapshots.set(workflowId, snapshot)
+    set({ lastExecutionSnapshots: newSnapshots })
+  },
+
+  getLastExecutionSnapshot: (workflowId) => {
+    const { lastExecutionSnapshots } = get()
+    return lastExecutionSnapshots.get(workflowId)
+  },
+
+  clearLastExecutionSnapshot: (workflowId) => {
+    const { lastExecutionSnapshots } = get()
+    const newSnapshots = new Map(lastExecutionSnapshots)
+    newSnapshots.delete(workflowId)
+    set({ lastExecutionSnapshots: newSnapshots })
+  },
 }))

apps/sim/stores/execution/types.ts

@@ -1,4 +1,5 @@
 import type { Executor } from '@/executor'
+import type { SerializableExecutionState } from '@/executor/execution/types'
 import type { ExecutionContext } from '@/executor/types'
 
 /**
@@ -18,16 +19,9 @@ export interface ExecutionState {
   pendingBlocks: string[]
   executor: Executor | null
   debugContext: ExecutionContext | null
-  /**
-   * Tracks blocks from the last execution run and their success/error status.
-   * Cleared when a new run starts. Used to show run path indicators (rings on blocks).
-   */
   lastRunPath: Map<string, BlockRunStatus>
-  /**
-   * Tracks edges from the last execution run and their success/error status.
-   * Cleared when a new run starts. Used to show run path indicators on edges.
-   */
   lastRunEdges: Map<string, EdgeRunStatus>
+  lastExecutionSnapshots: Map<string, SerializableExecutionState>
 }
 
 export interface ExecutionActions {
@@ -41,6 +35,9 @@ export interface ExecutionActions {
   setEdgeRunStatus: (edgeId: string, status: EdgeRunStatus) => void
   clearRunPath: () => void
   reset: () => void
+  setLastExecutionSnapshot: (workflowId: string, snapshot: SerializableExecutionState) => void
+  getLastExecutionSnapshot: (workflowId: string) => SerializableExecutionState | undefined
+  clearLastExecutionSnapshot: (workflowId: string) => void
 }
 
 export const initialState: ExecutionState = {
@@ -52,4 +49,5 @@ export const initialState: ExecutionState = {
   debugContext: null,
   lastRunPath: new Map(),
   lastRunEdges: new Map(),
+  lastExecutionSnapshots: new Map(),
 }

apps/sim/tools/index.test.ts

@@ -253,6 +253,23 @@ describe('executeTool Function', () => {
     vi.restoreAllMocks()
   })
 
+  it('should handle errors from tools', async () => {
+    setupFetchMock({ status: 400, ok: false, json: { error: 'Bad request' } })
+
+    const result = await executeTool(
+      'http_request',
+      {
+        url: 'https://api.example.com/data',
+        method: 'GET',
+      },
+      true
+    )
+
+    expect(result.success).toBe(false)
+    expect(result.error).toBeDefined()
+    expect(result.timing).toBeDefined()
+  })
+
   it('should add timing information to results', async () => {
     const result = await executeTool(
       'http_request',

helm/sim/templates/cert-manager-issuers.yaml

@@ -1,84 +0,0 @@
-{{- if .Values.certManager.enabled }}
-{{- /*
-  cert-manager Issuer Bootstrap Pattern
-
-  PREREQUISITE: cert-manager must be installed in your cluster before enabling this.
-  The root CA Certificate is created in the namespace specified by certManager.rootCA.namespace
-  (defaults to "cert-manager"). Ensure this namespace exists and cert-manager is running there.
-
-  Install cert-manager: https://cert-manager.io/docs/installation/
-
-  This implements the recommended pattern from cert-manager documentation:
-  1. A self-signed ClusterIssuer (for bootstrapping the root CA only)
-  2. A root CA Certificate (self-signed, used to sign other certificates)
-  3. A CA ClusterIssuer (uses the root CA to sign certificates)
-
-  Reference: https://cert-manager.io/docs/configuration/selfsigned/
-*/ -}}
-
----
-# 1. Self-Signed ClusterIssuer (Bootstrap Only)
-# This issuer is used ONLY to create the root CA certificate.
-# It should NOT be used directly for application certificates.
-apiVersion: cert-manager.io/v1
-kind: ClusterIssuer
-metadata:
-  name: {{ .Values.certManager.selfSignedIssuer.name }}
-  labels:
-    {{- include "sim.labels" . | nindent 4 }}
-    app.kubernetes.io/component: cert-manager
-spec:
-  selfSigned: {}
-
----
-# 2. Root CA Certificate
-# This certificate is signed by the self-signed issuer and becomes the root of trust.
-# The secret created here will be used by the CA issuer to sign certificates.
-# NOTE: This must be created in the cert-manager namespace (or the namespace specified
-# in certManager.rootCA.namespace). Ensure cert-manager is installed there first.
-apiVersion: cert-manager.io/v1
-kind: Certificate
-metadata:
-  name: {{ .Values.certManager.rootCA.certificateName }}
-  namespace: {{ .Values.certManager.rootCA.namespace | default "cert-manager" }}  # Must match cert-manager's cluster-resource-namespace
-  labels:
-    {{- include "sim.labels" . | nindent 4 }}
-    app.kubernetes.io/component: cert-manager
-spec:
-  isCA: true
-  commonName: {{ .Values.certManager.rootCA.commonName }}
-  secretName: {{ .Values.certManager.rootCA.secretName }}
-  duration: {{ .Values.certManager.rootCA.duration | default "87600h" }}
-  renewBefore: {{ .Values.certManager.rootCA.renewBefore | default "2160h" }}
-  privateKey:
-    algorithm: {{ .Values.certManager.rootCA.privateKey.algorithm | default "RSA" }}
-    size: {{ .Values.certManager.rootCA.privateKey.size | default 4096 }}
-  subject:
-    organizations:
-      {{- if .Values.certManager.rootCA.subject.organizations }}
-      {{- toYaml .Values.certManager.rootCA.subject.organizations | nindent 6 }}
-      {{- else }}
-      - {{ .Release.Name }}
-      {{- end }}
-  issuerRef:
-    name: {{ .Values.certManager.selfSignedIssuer.name }}
-    kind: ClusterIssuer
-    group: cert-manager.io
-
----
-# 3. CA ClusterIssuer
-# This is the issuer that should be used by applications to obtain certificates.
-# It signs certificates using the root CA created above.
-# NOTE: This issuer may briefly show "not ready" on first install while cert-manager
-# processes the Certificate above and creates the secret. It will auto-reconcile.
-apiVersion: cert-manager.io/v1
-kind: ClusterIssuer
-metadata:
-  name: {{ .Values.certManager.caIssuer.name }}
-  labels:
-    {{- include "sim.labels" . | nindent 4 }}
-    app.kubernetes.io/component: cert-manager
-spec:
-  ca:
-    secretName: {{ .Values.certManager.rootCA.secretName }}
-{{- end }}

helm/sim/templates/gpu-device-plugin.yaml

@@ -1,36 +1,6 @@
 {{- if and .Values.ollama.enabled .Values.ollama.gpu.enabled }}
 ---
-# 1. ConfigMap for NVIDIA Device Plugin Configuration
-apiVersion: v1
-kind: ConfigMap
-metadata:
-  name: {{ include "sim.fullname" . }}-nvidia-device-plugin-config
-  namespace: {{ .Release.Namespace }}
-  labels:
-    {{- include "sim.labels" . | nindent 4 }}
-    app.kubernetes.io/component: nvidia-device-plugin
-data:
-  config.yaml: |
-    version: v1
-    flags:
-      {{- if eq .Values.ollama.gpu.strategy "mig" }}
-      migStrategy: "single"
-      {{- else }}
-      migStrategy: "none"
-      {{- end }}
-      failOnInitError: false
-    plugin:
-      passDeviceSpecs: true
-      deviceListStrategy: envvar
-    {{- if eq .Values.ollama.gpu.strategy "time-slicing" }}
-    sharing:
-      timeSlicing:
-        resources:
-          - name: nvidia.com/gpu
-            replicas: {{ .Values.ollama.gpu.timeSlicingReplicas | default 5 }}
-    {{- end }}
----
-# 2. NVIDIA Device Plugin DaemonSet for GPU support
+# NVIDIA Device Plugin DaemonSet for GPU support
 apiVersion: apps/v1
 kind: DaemonSet
 metadata:
@@ -65,6 +35,9 @@ spec:
         # Only schedule on nodes with NVIDIA GPUs
         accelerator: nvidia
       priorityClassName: system-node-critical
+      runtimeClassName: nvidia
+      hostNetwork: true
+      hostPID: true
       volumes:
         - name: device-plugin
           hostPath:
@@ -75,21 +48,22 @@ spec:
         - name: sys
           hostPath:
             path: /sys
-        # Volume to mount the ConfigMap
-        - name: nvidia-device-plugin-config
-          configMap:
-            name: {{ include "sim.fullname" . }}-nvidia-device-plugin-config
+        - name: proc-driver-nvidia
+          hostPath:
+            path: /proc/driver/nvidia
       containers:
         - name: nvidia-device-plugin
-          image: nvcr.io/nvidia/k8s-device-plugin:v0.18.2
+          image: nvcr.io/nvidia/k8s-device-plugin:v0.14.5
           imagePullPolicy: Always
           args:
-            - "--config-file=/etc/device-plugin/config.yaml"
-          {{- if eq .Values.ollama.gpu.strategy "mig" }}
+            - --mig-strategy=single
+            - --pass-device-specs=true
+            - --fail-on-init-error=false
+            - --device-list-strategy=envvar
+            - --nvidia-driver-root=/host-sys/fs/cgroup
           env:
             - name: NVIDIA_MIG_MONITOR_DEVICES
               value: all
-          {{- end }}
           securityContext:
             allowPrivilegeEscalation: false
             capabilities:
@@ -100,16 +74,29 @@ spec:
             - name: dev
               mountPath: /dev
             - name: sys
-              mountPath: /sys
+              mountPath: /host-sys
               readOnly: true
-            - name: nvidia-device-plugin-config
-              mountPath: /etc/device-plugin/
+            - name: proc-driver-nvidia
+              mountPath: /proc/driver/nvidia
               readOnly: true
           resources:
             requests:
               cpu: 50m
-              memory: 20Mi
+              memory: 10Mi
             limits:
               cpu: 50m
-              memory: 50Mi
-{{- end }}
+              memory: 20Mi
+  {{- if .Values.nodeSelector }}
+      nodeSelector:
+        {{- toYaml .Values.nodeSelector | nindent 8 }}
+  {{- end }}
+---
+# RuntimeClass for NVIDIA Container Runtime
+apiVersion: node.k8s.io/v1
+kind: RuntimeClass
+metadata:
+  name: {{ include "sim.fullname" . }}-nvidia
+  labels:
+    {{- include "sim.labels" . | nindent 4 }}
+handler: nvidia
+{{- end }}

helm/sim/values.yaml

@@ -400,10 +400,8 @@ postgresql:
       algorithm: RSA  # RSA or ECDSA
       size: 4096  # Key size in bits
     # Issuer reference (REQUIRED if tls.enabled is true)
-    # By default, references the CA issuer created by certManager.caIssuer
-    # Make sure certManager.enabled is true, or provide your own issuer
     issuerRef:
-      name: sim-ca-issuer  # Name of your cert-manager Issuer/ClusterIssuer
+      name: selfsigned-cluster-issuer  # Name of your cert-manager Issuer/ClusterIssuer
       kind: ClusterIssuer  # ClusterIssuer or Issuer
       group: ""  # Optional: cert-manager.io (leave empty for default)
     # Additional DNS names (optional)
@@ -465,26 +463,20 @@ externalDatabase:
 ollama:
   # Enable/disable Ollama deployment
   enabled: false
-
+  
   # Image configuration
   image:
     repository: ollama/ollama
     tag: latest
     pullPolicy: Always
-
+  
   # Number of replicas
   replicaCount: 1
-
+  
   # GPU configuration
   gpu:
     enabled: false
     count: 1
-    # GPU sharing strategy: "mig" (Multi-Instance GPU) or "time-slicing"
-    # - mig: Hardware-level GPU partitioning (requires supported GPUs like A100)
-    # - time-slicing: Software-level GPU sharing (works with most NVIDIA GPUs)
-    strategy: "time-slicing"
-    # Number of time-slicing replicas (only used when strategy is "time-slicing")
-    timeSlicingReplicas: 5
   
   # Node selector for GPU workloads (adjust labels based on your cluster configuration)
   nodeSelector:
@@ -1193,53 +1185,4 @@ externalSecrets:
     # External database password (when using managed database services)
     externalDatabase:
       # Path to external database password in external store
-      password: ""
-
-# cert-manager configuration
-# Prerequisites: Install cert-manager in your cluster first
-# See: https://cert-manager.io/docs/installation/
-#
-# This implements the recommended CA bootstrap pattern from cert-manager:
-# 1. Self-signed ClusterIssuer (bootstrap only - creates root CA)
-# 2. Root CA Certificate (self-signed, becomes the trust anchor)
-# 3. CA ClusterIssuer (signs application certificates using root CA)
-#
-# Reference: https://cert-manager.io/docs/configuration/selfsigned/
-certManager:
-  # Enable/disable cert-manager issuer resources
-  enabled: false
-
-  # Self-signed ClusterIssuer (used ONLY to bootstrap the root CA)
-  # Do not reference this issuer directly for application certificates
-  selfSignedIssuer:
-    name: "sim-selfsigned-bootstrap-issuer"
-
-  # Root CA Certificate configuration
-  # This certificate is signed by the self-signed issuer and used as the trust anchor
-  rootCA:
-    # Name of the Certificate resource
-    certificateName: "sim-root-ca"
-    # Namespace where the root CA certificate and secret will be created
-    # Must match cert-manager's cluster-resource-namespace (default: cert-manager)
-    namespace: "cert-manager"
-    # Common name for the root CA certificate
-    commonName: "sim-root-ca"
-    # Secret name where the root CA certificate and key will be stored
-    secretName: "sim-root-ca-secret"
-    # Certificate validity duration (default: 10 years)
-    duration: "87600h"
-    # Renew before expiry (default: 90 days)
-    renewBefore: "2160h"
-    # Private key configuration
-    privateKey:
-      algorithm: RSA
-      size: 4096
-    # Subject configuration
-    subject:
-      organizations: []
-      # If empty, defaults to the release name
-
-  # CA ClusterIssuer configuration
-  # This is the issuer that applications should reference for obtaining certificates
-  caIssuer:
-    name: "sim-ca-issuer"
+      password: ""