feature(models): added vllm provider (#2103)

* Add vLLM self-hosted provider

* updated vllm to have pull parity with openai, dynamically fetch models

---------

Co-authored-by: MagellaX <[email protected]>

Author: waleedlatif1

apps/sim/app/(landing)/components/integrations/integrations.tsx

@@ -13,6 +13,7 @@ const modelProviderIcons = [
   { icon: Icons.OllamaIcon, label: 'Ollama' },
   { icon: Icons.DeepseekIcon, label: 'Deepseek' },
   { icon: Icons.ElevenLabsIcon, label: 'ElevenLabs' },
+  { icon: Icons.VllmIcon, label: 'vLLM' },
 ]
 
 const communicationIcons = [
@@ -88,7 +89,6 @@ interface TickerRowProps {
 }
 
 function TickerRow({ direction, offset, showOdd, icons }: TickerRowProps) {
-  // Create multiple copies of the icons array for seamless looping
   const extendedIcons = [...icons, ...icons, ...icons, ...icons]
 
   return (

apps/sim/app/.env.example

@@ -20,4 +20,6 @@ INTERNAL_API_SECRET=your_internal_api_secret # Use `openssl rand -hex 32` to gen
                    # If left commented out, emails will be logged to console instead
 
 # Local AI Models (Optional)
-# OLLAMA_URL=http://localhost:11434  # URL for local Ollama server - uncomment if using local models
+# OLLAMA_URL=http://localhost:11434  # URL for local Ollama server - uncomment if using local models
+# VLLM_BASE_URL=http://localhost:8000    # Base URL for your self-hosted vLLM (OpenAI-compatible)
+# VLLM_API_KEY=                          # Optional bearer token if your vLLM instance requires auth

apps/sim/app/api/providers/vllm/models/route.ts

@@ -0,0 +1,56 @@
+import { type NextRequest, NextResponse } from 'next/server'
+import { env } from '@/lib/env'
+import { createLogger } from '@/lib/logs/console/logger'
+
+const logger = createLogger('VLLMModelsAPI')
+
+/**
+ * Get available vLLM models
+ */
+export async function GET(request: NextRequest) {
+  const baseUrl = (env.VLLM_BASE_URL || '').replace(/\/$/, '')
+
+  if (!baseUrl) {
+    logger.info('VLLM_BASE_URL not configured')
+    return NextResponse.json({ models: [] })
+  }
+
+  try {
+    logger.info('Fetching vLLM models', {
+      baseUrl,
+    })
+
+    const response = await fetch(`${baseUrl}/v1/models`, {
+      headers: {
+        'Content-Type': 'application/json',
+      },
+      next: { revalidate: 60 },
+    })
+
+    if (!response.ok) {
+      logger.warn('vLLM service is not available', {
+        status: response.status,
+        statusText: response.statusText,
+      })
+      return NextResponse.json({ models: [] })
+    }
+
+    const data = (await response.json()) as { data: Array<{ id: string }> }
+    const models = data.data.map((model) => `vllm/${model.id}`)
+
+    logger.info('Successfully fetched vLLM models', {
+      count: models.length,
+      models,
+    })
+
+    return NextResponse.json({ models })
+  } catch (error) {
+    logger.error('Failed to fetch vLLM models', {
+      error: error instanceof Error ? error.message : 'Unknown error',
+      baseUrl,
+    })
+
+    // Return empty array instead of error to avoid breaking the UI
+    return NextResponse.json({ models: [] })
+  }
+}

apps/sim/app/workspace/[workspaceId]/providers/provider-models-loader.tsx

@@ -3,7 +3,11 @@
 import { useEffect } from 'react'
 import { createLogger } from '@/lib/logs/console/logger'
 import { useProviderModels } from '@/hooks/queries/providers'
-import { updateOllamaProviderModels, updateOpenRouterProviderModels } from '@/providers/utils'
+import {
+  updateOllamaProviderModels,
+  updateOpenRouterProviderModels,
+  updateVLLMProviderModels,
+} from '@/providers/utils'
 import { useProvidersStore } from '@/stores/providers/store'
 import type { ProviderName } from '@/stores/providers/types'
 
@@ -24,6 +28,8 @@ function useSyncProvider(provider: ProviderName) {
     try {
       if (provider === 'ollama') {
         updateOllamaProviderModels(data)
+      } else if (provider === 'vllm') {
+        updateVLLMProviderModels(data)
       } else if (provider === 'openrouter') {
         void updateOpenRouterProviderModels(data)
       }
@@ -44,6 +50,7 @@ function useSyncProvider(provider: ProviderName) {
 export function ProviderModelsLoader() {
   useSyncProvider('base')
   useSyncProvider('ollama')
+  useSyncProvider('vllm')
   useSyncProvider('openrouter')
   return null
 }

apps/sim/blocks/blocks/agent.ts

@@ -18,6 +18,10 @@ const getCurrentOllamaModels = () => {
   return useProvidersStore.getState().providers.ollama.models
 }
 
+const getCurrentVLLMModels = () => {
+  return useProvidersStore.getState().providers.vllm.models
+}
+
 import { useProvidersStore } from '@/stores/providers/store'
 import type { ToolResponse } from '@/tools/types'
 
@@ -90,8 +94,11 @@ export const AgentBlock: BlockConfig<AgentResponse> = {
         const providersState = useProvidersStore.getState()
         const baseModels = providersState.providers.base.models
         const ollamaModels = providersState.providers.ollama.models
+        const vllmModels = providersState.providers.vllm.models
         const openrouterModels = providersState.providers.openrouter.models
-        const allModels = Array.from(new Set([...baseModels, ...ollamaModels, ...openrouterModels]))
+        const allModels = Array.from(
+          new Set([...baseModels, ...ollamaModels, ...vllmModels, ...openrouterModels])
+        )
 
         return allModels.map((model) => {
           const icon = getProviderIcon(model)
@@ -172,7 +179,7 @@ export const AgentBlock: BlockConfig<AgentResponse> = {
       password: true,
       connectionDroppable: false,
       required: true,
-      // Hide API key for hosted models and Ollama models
+      // Hide API key for hosted models, Ollama models, and vLLM models
       condition: isHosted
         ? {
             field: 'model',
@@ -181,8 +188,8 @@ export const AgentBlock: BlockConfig<AgentResponse> = {
           }
         : () => ({
             field: 'model',
-            value: getCurrentOllamaModels(),
-            not: true, // Show for all models EXCEPT Ollama models
+            value: [...getCurrentOllamaModels(), ...getCurrentVLLMModels()],
+            not: true, // Show for all models EXCEPT Ollama and vLLM models
           }),
     },
     {

apps/sim/components/icons.tsx

@@ -4150,3 +4150,13 @@ export function VideoIcon(props: SVGProps<SVGSVGElement>) {
     </svg>
   )
 }
+
+export function VllmIcon(props: SVGProps<SVGSVGElement>) {
+  return (
+    <svg {...props} fill='currentColor' viewBox='0 0 24 24' xmlns='http://www.w3.org/2000/svg'>
+      <title>vLLM</title>
+      <path d='M0 4.973h9.324V23L0 4.973z' fill='#FDB515' />
+      <path d='M13.986 4.351L22.378 0l-6.216 23H9.324l4.662-18.649z' fill='#30A2FF' />
+    </svg>
+  )
+}

apps/sim/hooks/queries/providers.ts

@@ -7,6 +7,7 @@ const logger = createLogger('ProviderModelsQuery')
 const providerEndpoints: Record<ProviderName, string> = {
   base: '/api/providers/base/models',
   ollama: '/api/providers/ollama/models',
+  vllm: '/api/providers/vllm/models',
   openrouter: '/api/providers/openrouter/models',
 }
 

apps/sim/lib/env.ts

@@ -77,6 +77,8 @@ export const env = createEnv({
     ANTHROPIC_API_KEY_2:                   z.string().min(1).optional(),           // Additional Anthropic API key for load balancing
     ANTHROPIC_API_KEY_3:                   z.string().min(1).optional(),           // Additional Anthropic API key for load balancing
     OLLAMA_URL:                            z.string().url().optional(),            // Ollama local LLM server URL
+    VLLM_BASE_URL:                         z.string().url().optional(),            // vLLM self-hosted base URL (OpenAI-compatible)
+    VLLM_API_KEY:                          z.string().optional(),                  // Optional bearer token for vLLM
     ELEVENLABS_API_KEY:                    z.string().min(1).optional(),           // ElevenLabs API key for text-to-speech in deployed chat
     SERPER_API_KEY:                        z.string().min(1).optional(),           // Serper API key for online search
     EXA_API_KEY:                           z.string().min(1).optional(),           // Exa AI API key for enhanced online search

apps/sim/providers/models.ts

@@ -19,6 +19,7 @@ import {
   OllamaIcon,
   OpenAIIcon,
   OpenRouterIcon,
+  VllmIcon,
   xAIIcon,
 } from '@/components/icons'
 
@@ -82,6 +83,19 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
     contextInformationAvailable: false,
     models: [],
   },
+  vllm: {
+    id: 'vllm',
+    name: 'vLLM',
+    icon: VllmIcon,
+    description: 'Self-hosted vLLM with an OpenAI-compatible API',
+    defaultModel: 'vllm/generic',
+    modelPatterns: [/^vllm\//],
+    capabilities: {
+      temperature: { min: 0, max: 2 },
+      toolUsageControl: true,
+    },
+    models: [],
+  },
   openai: {
     id: 'openai',
     name: 'OpenAI',
@@ -1366,6 +1380,21 @@ export function updateOllamaModels(models: string[]): void {
   }))
 }
 
+/**
+ * Update vLLM models dynamically
+ */
+export function updateVLLMModels(models: string[]): void {
+  PROVIDER_DEFINITIONS.vllm.models = models.map((modelId) => ({
+    id: modelId,
+    pricing: {
+      input: 0,
+      output: 0,
+      updatedAt: new Date().toISOString().split('T')[0],
+    },
+    capabilities: {},
+  }))
+}
+
 /**
  * Update OpenRouter models dynamically
  */

apps/sim/providers/types.ts

@@ -12,6 +12,7 @@ export type ProviderId =
   | 'mistral'
   | 'ollama'
   | 'openrouter'
+  | 'vllm'
 
 /**
  * Model pricing information per million tokens