fix(max-tokens): anthropic models streaming vs non-streaming

* fix(docs): update requirements to be more accurate for deploying the app

* updated kb to support 1536 dimension vectors for models other than text embedding 3 small

* fix(storage): support Azure connection string for presigned URLs

* fix(kb): update test for embedding dimensions parameter

* fix(storage): align credential source ordering for consistency

.devcontainer/docker-compose.yml

@@ -44,7 +44,7 @@ services:
     deploy:
       resources:
         limits:
-          memory: 4G
+          memory: 1G
     environment:
       - NODE_ENV=development
       - DATABASE_URL=postgresql://postgres:postgres@db:5432/simstudio

apps/docs/content/docs/de/self-hosting/index.mdx

@@ -10,12 +10,20 @@ Stellen Sie Sim auf Ihrer eigenen Infrastruktur mit Docker oder Kubernetes berei
 
 ## Anforderungen
 
-| Ressource | Minimum | Empfohlen |
+| Ressource | Klein | Standard | Produktion |
-|----------|---------|-------------|
+|----------|-------|----------|------------|
-| CPU | 2 Kerne | 4+ Kerne |
+| CPU | 2 Kerne | 4 Kerne | 8+ Kerne |
-| RAM | 12 GB | 16+ GB |
+| RAM | 12 GB | 16 GB | 32+ GB |
-| Speicher | 20 GB SSD | 50+ GB SSD |
+| Speicher | 20 GB SSD | 50 GB SSD | 100+ GB SSD |
-| Docker | 20.10+ | Neueste Version |
+| Docker | 20.10+ | 20.10+ | Neueste Version |
+
+**Klein**: Entwicklung, Tests, Einzelnutzer (1-5 Nutzer)
+**Standard**: Teams (5-50 Nutzer), moderate Arbeitslasten
+**Produktion**: Große Teams (50+ Nutzer), Hochverfügbarkeit, intensive Workflow-Ausführung
+
+<Callout type="info">
+Die Ressourcenanforderungen werden durch Workflow-Ausführung (isolated-vm Sandboxing), Dateiverarbeitung (In-Memory-Dokumentenparsing) und Vektoroperationen (pgvector) bestimmt. Arbeitsspeicher ist typischerweise der limitierende Faktor, nicht CPU. Produktionsdaten zeigen, dass die Hauptanwendung durchschnittlich 4-8 GB und bei hoher Last bis zu 12 GB benötigt.
+</Callout>
 
 ## Schnellstart
 

apps/docs/content/docs/en/self-hosting/index.mdx

@@ -16,12 +16,20 @@ Deploy Sim on your own infrastructure with Docker or Kubernetes.
 
 ## Requirements
 
-| Resource | Minimum | Recommended |
+| Resource | Small | Standard | Production |
-|----------|---------|-------------|
+|----------|-------|----------|------------|
-| CPU | 2 cores | 4+ cores |
+| CPU | 2 cores | 4 cores | 8+ cores |
-| RAM | 12 GB | 16+ GB |
+| RAM | 12 GB | 16 GB | 32+ GB |
-| Storage | 20 GB SSD | 50+ GB SSD |
+| Storage | 20 GB SSD | 50 GB SSD | 100+ GB SSD |
-| Docker | 20.10+ | Latest |
+| Docker | 20.10+ | 20.10+ | Latest |
+
+**Small**: Development, testing, single user (1-5 users)
+**Standard**: Teams (5-50 users), moderate workloads
+**Production**: Large teams (50+ users), high availability, heavy workflow execution
+
+<Callout type="info">
+Resource requirements are driven by workflow execution (isolated-vm sandboxing), file processing (in-memory document parsing), and vector operations (pgvector). Memory is typically the constraining factor rather than CPU. Production telemetry shows the main app uses 4-8 GB average with peaks up to 12 GB under heavy load.
+</Callout>
 
 ## Quick Start
 

apps/docs/content/docs/es/self-hosting/index.mdx

@@ -10,12 +10,20 @@ Despliega Sim en tu propia infraestructura con Docker o Kubernetes.
 
 ## Requisitos
 
-| Recurso | Mínimo | Recomendado |
+| Recurso | Pequeño | Estándar | Producción |
-|----------|---------|-------------|
+|----------|---------|----------|------------|
-| CPU | 2 núcleos | 4+ núcleos |
+| CPU | 2 núcleos | 4 núcleos | 8+ núcleos |
-| RAM | 12 GB | 16+ GB |
+| RAM | 12 GB | 16 GB | 32+ GB |
-| Almacenamiento | 20 GB SSD | 50+ GB SSD |
+| Almacenamiento | 20 GB SSD | 50 GB SSD | 100+ GB SSD |
-| Docker | 20.10+ | Última versión |
+| Docker | 20.10+ | 20.10+ | Última versión |
+
+**Pequeño**: Desarrollo, pruebas, usuario único (1-5 usuarios)
+**Estándar**: Equipos (5-50 usuarios), cargas de trabajo moderadas
+**Producción**: Equipos grandes (50+ usuarios), alta disponibilidad, ejecución intensiva de workflows
+
+<Callout type="info">
+Los requisitos de recursos están determinados por la ejecución de workflows (sandboxing isolated-vm), procesamiento de archivos (análisis de documentos en memoria) y operaciones vectoriales (pgvector). La memoria suele ser el factor limitante, no la CPU. La telemetría de producción muestra que la aplicación principal usa 4-8 GB en promedio con picos de hasta 12 GB bajo carga pesada.
+</Callout>
 
 ## Inicio rápido
 

apps/docs/content/docs/fr/self-hosting/index.mdx

@@ -10,12 +10,20 @@ Déployez Sim sur votre propre infrastructure avec Docker ou Kubernetes.
 
 ## Prérequis
 
-| Ressource | Minimum | Recommandé |
+| Ressource | Petit | Standard | Production |
-|----------|---------|-------------|
+|----------|-------|----------|------------|
-| CPU | 2 cœurs | 4+ cœurs |
+| CPU | 2 cœurs | 4 cœurs | 8+ cœurs |
-| RAM | 12 Go | 16+ Go |
+| RAM | 12 Go | 16 Go | 32+ Go |
-| Stockage | 20 Go SSD | 50+ Go SSD |
+| Stockage | 20 Go SSD | 50 Go SSD | 100+ Go SSD |
-| Docker | 20.10+ | Dernière version |
+| Docker | 20.10+ | 20.10+ | Dernière version |
+
+**Petit** : Développement, tests, utilisateur unique (1-5 utilisateurs)
+**Standard** : Équipes (5-50 utilisateurs), charges de travail modérées
+**Production** : Grandes équipes (50+ utilisateurs), haute disponibilité, exécution intensive de workflows
+
+<Callout type="info">
+Les besoins en ressources sont déterminés par l'exécution des workflows (sandboxing isolated-vm), le traitement des fichiers (analyse de documents en mémoire) et les opérations vectorielles (pgvector). La mémoire est généralement le facteur limitant, pas le CPU. La télémétrie de production montre que l'application principale utilise 4-8 Go en moyenne avec des pics jusqu'à 12 Go sous forte charge.
+</Callout>
 
 ## Démarrage rapide
 

apps/docs/content/docs/ja/self-hosting/index.mdx

@@ -10,12 +10,20 @@ DockerまたはKubernetesを使用して、自社のインフラストラクチ
 
 ## 要件
 
-| リソース | 最小 | 推奨 |
+| リソース | スモール | スタンダード | プロダクション |
-|----------|---------|-------------|
+|----------|---------|-------------|----------------|
-| CPU | 2コア | 4+コア |
+| CPU | 2コア | 4コア | 8+コア |
-| RAM | 12 GB | 16+ GB |
+| RAM | 12 GB | 16 GB | 32+ GB |
-| ストレージ | 20 GB SSD | 50+ GB SSD |
+| ストレージ | 20 GB SSD | 50 GB SSD | 100+ GB SSD |
-| Docker | 20.10+ | 最新版 |
+| Docker | 20.10+ | 20.10+ | 最新版 |
+
+**スモール**: 開発、テスト、シングルユーザー（1-5ユーザー）
+**スタンダード**: チーム（5-50ユーザー）、中程度のワークロード
+**プロダクション**: 大規模チーム（50+ユーザー）、高可用性、高負荷ワークフロー実行
+
+<Callout type="info">
+リソース要件は、ワークフロー実行（isolated-vmサンドボックス）、ファイル処理（メモリ内ドキュメント解析）、ベクトル演算（pgvector）によって決まります。CPUよりもメモリが制約要因となることが多いです。本番環境のテレメトリによると、メインアプリは平均4-8 GB、高負荷時は最大12 GBを使用します。
+</Callout>
 
 ## クイックスタート
 

apps/docs/content/docs/zh/self-hosting/index.mdx

@@ -10,12 +10,20 @@ import { Callout } from 'fumadocs-ui/components/callout'
 
 ## 要求
 
-| 资源 | 最低要求 | 推荐配置 |
+| 资源 | 小型 | 标准 | 生产环境 |
-|----------|---------|-------------|
+|----------|------|------|----------|
-| CPU | 2 核 | 4 核及以上 |
+| CPU | 2 核 | 4 核 | 8+ 核 |
-| 内存 | 12 GB | 16 GB 及以上 |
+| 内存 | 12 GB | 16 GB | 32+ GB |
-| 存储 | 20 GB SSD | 50 GB 及以上 SSD |
+| 存储 | 20 GB SSD | 50 GB SSD | 100+ GB SSD |
-| Docker | 20.10+ | 最新版本 |
+| Docker | 20.10+ | 20.10+ | 最新版本 |
+
+**小型**: 开发、测试、单用户（1-5 用户）
+**标准**: 团队（5-50 用户）、中等工作负载
+**生产环境**: 大型团队（50+ 用户）、高可用性、密集工作流执行
+
+<Callout type="info">
+资源需求由工作流执行（isolated-vm 沙箱）、文件处理（内存中文档解析）和向量运算（pgvector）决定。内存通常是限制因素，而不是 CPU。生产遥测数据显示，主应用平均使用 4-8 GB，高负载时峰值可达 12 GB。
+</Callout>
 
 ## 快速开始
 

apps/sim/app/api/knowledge/search/utils.test.ts

@@ -408,6 +408,7 @@ describe('Knowledge Search Utils', () => {
             input: ['test query'],
             model: 'text-embedding-3-small',
             encoding_format: 'float',
+            dimensions: 1536,
           }),
         })
       )

apps/sim/blocks/blocks/agent.ts

@@ -513,6 +513,12 @@ Return ONLY the JSON array.`,
         })(),
       }),
     },
+    {
+      id: 'maxTokens',
+      title: 'Max Output Tokens',
+      type: 'short-input',
+      placeholder: 'Enter max tokens (e.g., 4096)...',
+    },
     {
       id: 'responseFormat',
       title: 'Response Format',
@@ -754,6 +760,7 @@ Example 3 (Array Input):
       },
     },
     temperature: { type: 'number', description: 'Response randomness level' },
+    maxTokens: { type: 'number', description: 'Maximum number of tokens in the response' },
     reasoningEffort: { type: 'string', description: 'Reasoning effort level for GPT-5 models' },
     verbosity: { type: 'string', description: 'Verbosity level for GPT-5 models' },
     thinkingLevel: { type: 'string', description: 'Thinking level for Gemini 3 models' },

apps/sim/lib/knowledge/embeddings.ts

@@ -8,6 +8,17 @@ const logger = createLogger('EmbeddingUtils')
 
 const MAX_TOKENS_PER_REQUEST = 8000
 const MAX_CONCURRENT_BATCHES = env.KB_CONFIG_CONCURRENCY_LIMIT || 50
+const EMBEDDING_DIMENSIONS = 1536
+
+/**
+ * Check if the model supports custom dimensions.
+ * text-embedding-3-* models support the dimensions parameter.
+ * Checks for 'embedding-3' to handle Azure deployments with custom naming conventions.
+ */
+function supportsCustomDimensions(modelName: string): boolean {
+  const name = modelName.toLowerCase()
+  return name.includes('embedding-3') && !name.includes('ada')
+}
 
 export class EmbeddingAPIError extends Error {
   public status: number
@@ -93,15 +104,19 @@ async function getEmbeddingConfig(
 async function callEmbeddingAPI(inputs: string[], config: EmbeddingConfig): Promise<number[][]> {
   return retryWithExponentialBackoff(
     async () => {
+      const useDimensions = supportsCustomDimensions(config.modelName)
+
       const requestBody = config.useAzure
         ? {
             input: inputs,
             encoding_format: 'float',
+            ...(useDimensions && { dimensions: EMBEDDING_DIMENSIONS }),
           }
         : {
             input: inputs,
             model: config.modelName,
             encoding_format: 'float',
+            ...(useDimensions && { dimensions: EMBEDDING_DIMENSIONS }),
           }
 
       const response = await fetch(config.apiUrl, {

apps/sim/lib/uploads/providers/blob/client.ts

@@ -18,6 +18,52 @@ const logger = createLogger('BlobClient')
 
 let _blobServiceClient: BlobServiceClientInstance | null = null
 
+interface ParsedCredentials {
+  accountName: string
+  accountKey: string
+}
+
+/**
+ * Extract account name and key from an Azure connection string.
+ * Connection strings have the format: DefaultEndpointsProtocol=https;AccountName=...;AccountKey=...;EndpointSuffix=...
+ */
+function parseConnectionString(connectionString: string): ParsedCredentials {
+  const accountNameMatch = connectionString.match(/AccountName=([^;]+)/)
+  if (!accountNameMatch) {
+    throw new Error('Cannot extract account name from connection string')
+  }
+
+  const accountKeyMatch = connectionString.match(/AccountKey=([^;]+)/)
+  if (!accountKeyMatch) {
+    throw new Error('Cannot extract account key from connection string')
+  }
+
+  return {
+    accountName: accountNameMatch[1],
+    accountKey: accountKeyMatch[1],
+  }
+}
+
+/**
+ * Get account credentials from BLOB_CONFIG, extracting from connection string if necessary.
+ */
+function getAccountCredentials(): ParsedCredentials {
+  if (BLOB_CONFIG.connectionString) {
+    return parseConnectionString(BLOB_CONFIG.connectionString)
+  }
+
+  if (BLOB_CONFIG.accountName && BLOB_CONFIG.accountKey) {
+    return {
+      accountName: BLOB_CONFIG.accountName,
+      accountKey: BLOB_CONFIG.accountKey,
+    }
+  }
+
+  throw new Error(
+    'Azure Blob Storage credentials are missing – set AZURE_CONNECTION_STRING or both AZURE_ACCOUNT_NAME and AZURE_ACCOUNT_KEY'
+  )
+}
+
 export async function getBlobServiceClient(): Promise<BlobServiceClientInstance> {
   if (_blobServiceClient) return _blobServiceClient
 
@@ -127,6 +173,8 @@ export async function getPresignedUrl(key: string, expiresIn = 3600) {
   const containerClient = blobServiceClient.getContainerClient(BLOB_CONFIG.containerName)
   const blockBlobClient = containerClient.getBlockBlobClient(key)
 
+  const { accountName, accountKey } = getAccountCredentials()
+
   const sasOptions = {
     containerName: BLOB_CONFIG.containerName,
     blobName: key,
@@ -137,13 +185,7 @@ export async function getPresignedUrl(key: string, expiresIn = 3600) {
 
   const sasToken = generateBlobSASQueryParameters(
     sasOptions,
-    new StorageSharedKeyCredential(
+    new StorageSharedKeyCredential(accountName, accountKey)
-      BLOB_CONFIG.accountName,
-      BLOB_CONFIG.accountKey ??
-        (() => {
-          throw new Error('AZURE_ACCOUNT_KEY is required when using account name authentication')
-        })()
-    )
   ).toString()
 
   return `${blockBlobClient.url}?${sasToken}`
@@ -168,9 +210,14 @@ export async function getPresignedUrlWithConfig(
     StorageSharedKeyCredential,
   } = await import('@azure/storage-blob')
   let tempBlobServiceClient: BlobServiceClientInstance
+  let accountName: string
+  let accountKey: string
 
   if (customConfig.connectionString) {
     tempBlobServiceClient = BlobServiceClient.fromConnectionString(customConfig.connectionString)
+    const credentials = parseConnectionString(customConfig.connectionString)
+    accountName = credentials.accountName
+    accountKey = credentials.accountKey
   } else if (customConfig.accountName && customConfig.accountKey) {
     const sharedKeyCredential = new StorageSharedKeyCredential(
       customConfig.accountName,
@@ -180,6 +227,8 @@ export async function getPresignedUrlWithConfig(
       `https://${customConfig.accountName}.blob.core.windows.net`,
       sharedKeyCredential
     )
+    accountName = customConfig.accountName
+    accountKey = customConfig.accountKey
   } else {
     throw new Error(
       'Custom blob config must include either connectionString or accountName + accountKey'
@@ -199,13 +248,7 @@ export async function getPresignedUrlWithConfig(
 
   const sasToken = generateBlobSASQueryParameters(
     sasOptions,
-    new StorageSharedKeyCredential(
+    new StorageSharedKeyCredential(accountName, accountKey)
-      customConfig.accountName,
-      customConfig.accountKey ??
-        (() => {
-          throw new Error('Account key is required when using account name authentication')
-        })()
-    )
   ).toString()
 
   return `${blockBlobClient.url}?${sasToken}`
@@ -403,13 +446,9 @@ export async function getMultipartPartUrls(
   if (customConfig) {
     if (customConfig.connectionString) {
       blobServiceClient = BlobServiceClient.fromConnectionString(customConfig.connectionString)
-      const match = customConfig.connectionString.match(/AccountName=([^;]+)/)
+      const credentials = parseConnectionString(customConfig.connectionString)
-      if (!match) throw new Error('Cannot extract account name from connection string')
+      accountName = credentials.accountName
-      accountName = match[1]
+      accountKey = credentials.accountKey
-
-      const keyMatch = customConfig.connectionString.match(/AccountKey=([^;]+)/)
-      if (!keyMatch) throw new Error('Cannot extract account key from connection string')
-      accountKey = keyMatch[1]
     } else if (customConfig.accountName && customConfig.accountKey) {
       const credential = new StorageSharedKeyCredential(
         customConfig.accountName,
@@ -428,12 +467,9 @@ export async function getMultipartPartUrls(
   } else {
     blobServiceClient = await getBlobServiceClient()
     containerName = BLOB_CONFIG.containerName
-    accountName = BLOB_CONFIG.accountName
+    const credentials = getAccountCredentials()
-    accountKey =
+    accountName = credentials.accountName
-      BLOB_CONFIG.accountKey ||
+    accountKey = credentials.accountKey
-      (() => {
-        throw new Error('AZURE_ACCOUNT_KEY is required')
-      })()
   }
 
   const containerClient = blobServiceClient.getContainerClient(containerName)
@@ -501,12 +537,10 @@ export async function completeMultipartUpload(
   const containerClient = blobServiceClient.getContainerClient(containerName)
   const blockBlobClient = containerClient.getBlockBlobClient(key)
 
-  // Sort parts by part number and extract block IDs
   const sortedBlockIds = parts
     .sort((a, b) => a.partNumber - b.partNumber)
     .map((part) => part.blockId)
 
-  // Commit the block list to create the final blob
   await blockBlobClient.commitBlockList(sortedBlockIds, {
     metadata: {
       multipartUpload: 'completed',
@@ -557,10 +591,8 @@ export async function abortMultipartUpload(key: string, customConfig?: BlobConfi
   const blockBlobClient = containerClient.getBlockBlobClient(key)
 
   try {
-    // Delete the blob if it exists (this also cleans up any uncommitted blocks)
     await blockBlobClient.deleteIfExists()
   } catch (error) {
-    // Ignore errors since we're just cleaning up
     logger.warn('Error cleaning up multipart upload:', error)
   }
 }

apps/sim/providers/anthropic/index.ts

@@ -9,6 +9,7 @@ import {
   generateToolUseId,
 } from '@/providers/anthropic/utils'
 import {
+  getMaxOutputTokensForModel,
   getProviderDefaultModel,
   getProviderModels,
   supportsNativeStructuredOutputs,
@@ -178,7 +179,9 @@ export const anthropicProvider: ProviderConfig = {
       model: request.model,
       messages,
       system: systemPrompt,
-      max_tokens: Number.parseInt(String(request.maxTokens)) || 1024,
+      max_tokens:
+        Number.parseInt(String(request.maxTokens)) ||
+        getMaxOutputTokensForModel(request.model, request.stream ?? false),
       temperature: Number.parseFloat(String(request.temperature ?? 0.7)),
     }
 

apps/sim/providers/bedrock/index.ts

@@ -20,7 +20,11 @@ import {
   generateToolUseId,
   getBedrockInferenceProfileId,
 } from '@/providers/bedrock/utils'
-import { getProviderDefaultModel, getProviderModels } from '@/providers/models'
+import {
+  getMaxOutputTokensForModel,
+  getProviderDefaultModel,
+  getProviderModels,
+} from '@/providers/models'
 import type {
   ProviderConfig,
   ProviderRequest,
@@ -259,7 +263,9 @@ export const bedrockProvider: ProviderConfig = {
 
     const inferenceConfig = {
       temperature: Number.parseFloat(String(request.temperature ?? 0.7)),
-      maxTokens: Number.parseInt(String(request.maxTokens)) || 4096,
+      maxTokens:
+        Number.parseInt(String(request.maxTokens)) ||
+        getMaxOutputTokensForModel(request.model, request.stream ?? false),
     }
 
     const shouldStreamToolCalls = request.streamToolCalls ?? false

apps/sim/providers/models.ts

@@ -34,6 +34,12 @@ export interface ModelCapabilities {
   toolUsageControl?: boolean
   computerUse?: boolean
   nativeStructuredOutputs?: boolean
+  maxOutputTokens?: {
+    /** Maximum tokens for streaming requests */
+    max: number
+    /** Safe default for non-streaming requests (to avoid timeout issues) */
+    default: number
+  }
   reasoningEffort?: {
     values: string[]
   }
@@ -613,6 +619,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         capabilities: {
           temperature: { min: 0, max: 1 },
           nativeStructuredOutputs: true,
+          maxOutputTokens: { max: 64000, default: 4096 },
         },
         contextWindow: 200000,
       },
@@ -627,6 +634,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         capabilities: {
           temperature: { min: 0, max: 1 },
           nativeStructuredOutputs: true,
+          maxOutputTokens: { max: 64000, default: 4096 },
         },
         contextWindow: 200000,
       },
@@ -640,6 +648,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 1 },
+          maxOutputTokens: { max: 64000, default: 4096 },
         },
         contextWindow: 200000,
       },
@@ -654,6 +663,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         capabilities: {
           temperature: { min: 0, max: 1 },
           nativeStructuredOutputs: true,
+          maxOutputTokens: { max: 64000, default: 4096 },
         },
         contextWindow: 200000,
       },
@@ -668,6 +678,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         capabilities: {
           temperature: { min: 0, max: 1 },
           nativeStructuredOutputs: true,
+          maxOutputTokens: { max: 64000, default: 4096 },
         },
         contextWindow: 200000,
       },
@@ -681,6 +692,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         },
         capabilities: {
           temperature: { min: 0, max: 1 },
+          maxOutputTokens: { max: 64000, default: 4096 },
         },
         contextWindow: 200000,
       },
@@ -695,6 +707,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         capabilities: {
           temperature: { min: 0, max: 1 },
           computerUse: true,
+          maxOutputTokens: { max: 8192, default: 8192 },
         },
         contextWindow: 200000,
       },
@@ -709,6 +722,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         capabilities: {
           temperature: { min: 0, max: 1 },
           computerUse: true,
+          maxOutputTokens: { max: 8192, default: 8192 },
         },
         contextWindow: 200000,
       },
@@ -1655,6 +1669,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         capabilities: {
           temperature: { min: 0, max: 1 },
           nativeStructuredOutputs: true,
+          maxOutputTokens: { max: 64000, default: 4096 },
         },
         contextWindow: 200000,
       },
@@ -1668,6 +1683,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         capabilities: {
           temperature: { min: 0, max: 1 },
           nativeStructuredOutputs: true,
+          maxOutputTokens: { max: 64000, default: 4096 },
         },
         contextWindow: 200000,
       },
@@ -1681,6 +1697,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         capabilities: {
           temperature: { min: 0, max: 1 },
           nativeStructuredOutputs: true,
+          maxOutputTokens: { max: 64000, default: 4096 },
         },
         contextWindow: 200000,
       },
@@ -1694,6 +1711,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
         capabilities: {
           temperature: { min: 0, max: 1 },
           nativeStructuredOutputs: true,
+          maxOutputTokens: { max: 64000, default: 4096 },
         },
         contextWindow: 200000,
       },
@@ -2333,3 +2351,31 @@ export function getThinkingLevelsForModel(modelId: string): string[] | null {
   const capability = getThinkingCapability(modelId)
   return capability?.levels ?? null
 }
+
+/**
+ * Get the max output tokens for a specific model
+ * Returns the model's max capacity for streaming requests,
+ * or the model's safe default for non-streaming requests to avoid timeout issues.
+ *
+ * @param modelId - The model ID
+ * @param streaming - Whether the request is streaming (default: false)
+ */
+export function getMaxOutputTokensForModel(modelId: string, streaming = false): number {
+  const normalizedModelId = modelId.toLowerCase()
+  const STANDARD_MAX_OUTPUT_TOKENS = 4096
+
+  for (const provider of Object.values(PROVIDER_DEFINITIONS)) {
+    for (const model of provider.models) {
+      const baseModelId = model.id.toLowerCase()
+      if (normalizedModelId === baseModelId || normalizedModelId.startsWith(`${baseModelId}-`)) {
+        const outputTokens = model.capabilities.maxOutputTokens
+        if (outputTokens) {
+          return streaming ? outputTokens.max : outputTokens.default
+        }
+        return STANDARD_MAX_OUTPUT_TOKENS
+      }
+    }
+  }
+
+  return STANDARD_MAX_OUTPUT_TOKENS
+}

apps/sim/providers/utils.ts

@@ -8,6 +8,7 @@ import {
   getComputerUseModels,
   getEmbeddingModelPricing,
   getHostedModels as getHostedModelsFromDefinitions,
+  getMaxOutputTokensForModel as getMaxOutputTokensForModelFromDefinitions,
   getMaxTemperature as getMaxTempFromDefinitions,
   getModelPricing as getModelPricingFromDefinitions,
   getModelsWithReasoningEffort,
@@ -992,6 +993,18 @@ export function getThinkingLevelsForModel(model: string): string[] | null {
   return getThinkingLevelsForModelFromDefinitions(model)
 }
 
+/**
+ * Get max output tokens for a specific model
+ * Returns the model's maxOutputTokens capability for streaming requests,
+ * or a conservative default (8192) for non-streaming requests to avoid timeout issues.
+ *
+ * @param model - The model ID
+ * @param streaming - Whether the request is streaming (default: false)
+ */
+export function getMaxOutputTokensForModel(model: string, streaming = false): number {
+  return getMaxOutputTokensForModelFromDefinitions(model, streaming)
+}
+
 /**
  * Prepare tool execution parameters, separating tool parameters from system parameters
  */

apps/sim/tools/browser_use/run_task.ts

@@ -5,7 +5,7 @@ import type { ToolConfig, ToolResponse } from '@/tools/types'
 const logger = createLogger('BrowserUseTool')
 
 const POLL_INTERVAL_MS = 5000
-const MAX_POLL_TIME_MS = 180000
+const MAX_POLL_TIME_MS = 600000 // 10 minutes
 const MAX_CONSECUTIVE_ERRORS = 3
 
 async function createSessionWithProfile(

docker-compose.local.yml

@@ -52,7 +52,7 @@ services:
     deploy:
       resources:
         limits:
-          memory: 8G
+          memory: 1G
     healthcheck:
       test: ['CMD', 'wget', '--spider', '--quiet', 'http://127.0.0.1:3002/health']
       interval: 90s

docker-compose.ollama.yml

@@ -56,7 +56,7 @@ services:
     deploy:
       resources:
         limits:
-          memory: 8G
+          memory: 1G
     healthcheck:
       test: ['CMD', 'wget', '--spider', '--quiet', 'http://127.0.0.1:3002/health']
       interval: 90s

docker-compose.prod.yml

@@ -42,7 +42,7 @@ services:
     deploy:
       resources:
         limits:
-          memory: 4G
+          memory: 1G
     environment:
       - DATABASE_URL=postgresql://${POSTGRES_USER:-postgres}:${POSTGRES_PASSWORD:-postgres}@db:5432/${POSTGRES_DB:-simstudio}
       - NEXT_PUBLIC_APP_URL=${NEXT_PUBLIC_APP_URL:-http://localhost:3000}

helm/sim/examples/values-production.yaml

@@ -13,10 +13,10 @@ app:
 
   resources:
     limits:
-      memory: "6Gi"
+      memory: "8Gi"
       cpu: "2000m"
     requests:
-      memory: "4Gi"
+      memory: "6Gi"
       cpu: "1000m"
   
   # Production URLs (REQUIRED - update with your actual domain names)
@@ -52,11 +52,11 @@ realtime:
 
   resources:
     limits:
-      memory: "4Gi"
+      memory: "1Gi"
-      cpu: "1000m"
-    requests:
-      memory: "2Gi"
       cpu: "500m"
+    requests:
+      memory: "512Mi"
+      cpu: "250m"
   
   env:
     NEXT_PUBLIC_APP_URL: "https://sim.acme.ai"

helm/sim/values.yaml

@@ -29,10 +29,10 @@ app:
   # Resource limits and requests
   resources:
     limits:
-      memory: "4Gi"
+      memory: "8Gi"
       cpu: "2000m"
     requests:
-      memory: "2Gi"
+      memory: "4Gi"
       cpu: "1000m"
 
   # Node selector for pod scheduling (leave empty to allow scheduling on any node)
@@ -245,11 +245,11 @@ realtime:
   # Resource limits and requests
   resources:
     limits:
-      memory: "2Gi"
-      cpu: "1000m"
-    requests:
       memory: "1Gi"
       cpu: "500m"
+    requests:
+      memory: "512Mi"
+      cpu: "250m"
   
   # Node selector for pod scheduling (leave empty to allow scheduling on any node)
   nodeSelector: {}