fix(Hugging Face Inference Model Node, Embeddings Hugging Face Inference Node): Fix credential validation for HF nodes and add provider for (#18047)

packages/@n8n/nodes-langchain/credentials/HuggingFaceApi.credentials.ts

@@ -34,8 +34,8 @@ export class HuggingFaceApi implements ICredentialType {
 
 	test: ICredentialTestRequest = {
 		request: {
-			baseURL: 'https://api-inference.huggingface.co',
+			baseURL: 'https://huggingface.co',
-			url: '/models/gpt2',
+			url: '/api/whoami-v2',
 		},
 	};
 }

packages/@n8n/nodes-langchain/nodes/embeddings/EmbeddingsHuggingFaceInference/EmbeddingsHuggingFaceInference.node.ts

@@ -1,6 +1,9 @@
+import type { InferenceProviderOrPolicy } from '@huggingface/inference';
+import { PROVIDERS_OR_POLICIES } from '@huggingface/inference';
 import { HuggingFaceInferenceEmbeddings } from '@langchain/community/embeddings/hf';
 import {
 	NodeConnectionTypes,
+	NodeOperationError,
 	type INodeType,
 	type INodeTypeDescription,
 	type ISupplyDataFunctions,
@@ -76,6 +79,13 @@ export class EmbeddingsHuggingFaceInference implements INodeType {
 						description: 'Custom endpoint URL',
 						type: 'string',
 					},
+					{
+						displayName: 'Provider',
+						name: 'provider',
+						type: 'options',
+						options: PROVIDERS_OR_POLICIES.map((value) => ({ value, name: value })),
+						default: 'auto',
+					},
 				],
 			},
 		],
@@ -91,6 +101,10 @@ export class EmbeddingsHuggingFaceInference implements INodeType {
 		const credentials = await this.getCredentials('huggingFaceApi');
 		const options = this.getNodeParameter('options', itemIndex, {}) as object;
 
+		if ('provider' in options && !isValidHFProviderOrPolicy(options.provider)) {
+			throw new NodeOperationError(this.getNode(), 'Unsupported provider');
+		}
+
 		const embeddings = new HuggingFaceInferenceEmbeddings({
 			apiKey: credentials.apiKey as string,
 			model,
@@ -102,3 +116,9 @@ export class EmbeddingsHuggingFaceInference implements INodeType {
 		};
 	}
 }
+
+function isValidHFProviderOrPolicy(provider: unknown): provider is InferenceProviderOrPolicy {
+	return (
+		typeof provider === 'string' && (PROVIDERS_OR_POLICIES as readonly string[]).includes(provider)
+	);
+}

packages/@n8n/nodes-langchain/nodes/llms/LMOpenHuggingFaceInference/LmOpenHuggingFaceInference.node.ts

@@ -55,7 +55,7 @@ export class LmOpenHuggingFaceInference implements INodeType {
 				displayName: 'Model',
 				name: 'model',
 				type: 'string',
-				default: 'gpt2',
+				default: 'mistralai/Mistral-Nemo-Base-2407',
 			},
 			{
 				displayName: 'Options',
@@ -139,6 +139,11 @@ export class LmOpenHuggingFaceInference implements INodeType {
 		const modelName = this.getNodeParameter('model', itemIndex) as string;
 		const options = this.getNodeParameter('options', itemIndex, {}) as object;
 
+		// LangChain does not yet support specifying Provider
+		// That's why mistral's model is the default value
+		// It is one of the few models that seem to work out of the box
+		// Other models are returning "Model x/y is not supported for task text-generation and provider z. Supported task: conversational."
+		// https://github.com/langchain-ai/langchainjs/discussions/8434#discussioncomment-13603787
 		const model = new HuggingFaceInference({
 			model: modelName,
 			apiKey: credentials.apiKey as string,

packages/@n8n/nodes-langchain/package.json

@@ -171,7 +171,7 @@
     "@google-ai/generativelanguage": "2.6.0",
     "@google-cloud/resource-manager": "5.3.0",
     "@google/generative-ai": "0.21.0",
-    "@huggingface/inference": "2.8.0",
+    "@huggingface/inference": "4.0.5",
     "@langchain/anthropic": "catalog:",
     "@langchain/aws": "0.1.11",
     "@langchain/cohere": "0.3.4",

pnpm-lock.yaml

@@ -961,8 +961,8 @@ importers:
         specifier: 0.21.0
         version: 0.21.0
       '@huggingface/inference':
-        specifier: 2.8.0
+        specifier: 4.0.5
-        version: 2.8.0
+        version: 4.0.5
       '@langchain/anthropic':
         specifier: 'catalog:'
         version: 0.3.23(@langchain/core@0.3.61(@opentelemetry/api@1.9.0)(@opentelemetry/sdk-trace-base@1.30.1(@opentelemetry/api@1.9.0))(openai@5.8.1(ws@8.18.2)(zod@3.25.67)))
@@ -974,7 +974,7 @@ importers:
         version: 0.3.4(@langchain/core@0.3.61(@opentelemetry/api@1.9.0)(@opentelemetry/sdk-trace-base@1.30.1(@opentelemetry/api@1.9.0))(openai@5.8.1(ws@8.18.2)(zod@3.25.67)))(encoding@0.1.13)
       '@langchain/community':
         specifier: 'catalog:'
-        version: 0.3.47(438356ed01a7485a5354296e4de89428)
+        version: 0.3.47(2b92f6f7d0c28eaf0346fdac5730ff18)
       '@langchain/core':
         specifier: 'catalog:'
         version: 0.3.61(@opentelemetry/api@1.9.0)(@opentelemetry/sdk-trace-base@1.30.1(@opentelemetry/api@1.9.0))(openai@5.8.1(ws@8.18.2)(zod@3.25.67))
@@ -4693,12 +4693,16 @@ packages:
   '@hapi/topo@5.1.0':
     resolution: {integrity: sha512-foQZKJig7Ob0BMAYBfcJk8d77QtOe7Wo4ox7ff1lQYoNNAb6jwcY1ncdoy2e9wQZzvNy7ODZCYJkK8kzmcAnAg==}
 
-  '@huggingface/inference@2.8.0':
+  '@huggingface/inference@4.0.5':
-    resolution: {integrity: sha512-Ti681P1qckcCAqgzmL53jBnluPuZGelmMIuXNjgAwC5+RIjF4S0SDQu6oy44ZTwekwNp2ETaZ2sXsOk+45aC4w==}
+    resolution: {integrity: sha512-/Qc45BGrN+FBA3JfdeoHfafxfNShH/dxvOsXbBdcxyxIRIYOyefeiXSlShZGVCaiqYpm+10na28D0YtvjKPTlw==}
     engines: {node: '>=18'}
 
-  '@huggingface/tasks@0.11.13':
+  '@huggingface/jinja@0.5.1':
-    resolution: {integrity: sha512-TqFEyFtKYAYwDg9h4XQMzoSxN2NMpwSnerPBx7Y4RbM1nHLM+CTXAUHcDY+hydcA5CoDDaBGzkHw+mttY3AmFQ==}
+    resolution: {integrity: sha512-yUZLld4lrM9iFxHCwFQ7D1HW2MWMwSbeB7WzWqFYDWK+rEb+WldkLdAJxUPOmgICMHZLzZGVcVjFh3w/YGubng==}
+    engines: {node: '>=18'}
+
+  '@huggingface/tasks@0.19.34':
+    resolution: {integrity: sha512-dIl3jyeddCEFJeogJOcbhfIq1tlo3N9K4EAxG/MfkGL0l7hI2kfs91Ut+1h6i09TQM8A9XM91NV7Jz6PgfWE7Q==}
 
   '@humanfs/core@0.19.1':
     resolution: {integrity: sha512-5DyQ4+1JEUzejeK1JGICcideyfUbGixgS9jNgex5nqkW+cY7WZhxBigmieN5Qnw9ZosSNVC9KQKyb+GUaGyKUA==}
@@ -19071,11 +19075,14 @@ snapshots:
     dependencies:
       '@hapi/hoek': 9.3.0
 
-  '@huggingface/inference@2.8.0':
+  '@huggingface/inference@4.0.5':
     dependencies:
-      '@huggingface/tasks': 0.11.13
+      '@huggingface/jinja': 0.5.1
+      '@huggingface/tasks': 0.19.34
 
-  '@huggingface/tasks@0.11.13': {}
+  '@huggingface/jinja@0.5.1': {}
+
+  '@huggingface/tasks@0.19.34': {}
 
   '@humanfs/core@0.19.1': {}
 
@@ -19540,7 +19547,7 @@ snapshots:
       - aws-crt
       - encoding
 
-  '@langchain/community@0.3.47(438356ed01a7485a5354296e4de89428)':
+  '@langchain/community@0.3.47(2b92f6f7d0c28eaf0346fdac5730ff18)':
     dependencies:
       '@browserbasehq/stagehand': 1.9.0(@playwright/test@1.54.2)(deepmerge@4.3.1)(dotenv@16.5.0)(encoding@0.1.13)(openai@5.8.1(ws@8.18.2)(zod@3.25.67))(zod@3.25.67)
       '@ibm-cloud/watsonx-ai': 1.1.2
@@ -19570,7 +19577,7 @@ snapshots:
       '@getzep/zep-js': 0.9.0
       '@google-ai/generativelanguage': 2.6.0(encoding@0.1.13)
       '@google-cloud/storage': 7.12.1(encoding@0.1.13)
-      '@huggingface/inference': 2.8.0
+      '@huggingface/inference': 4.0.5
       '@mozilla/readability': 0.6.0
       '@pinecone-database/pinecone': 5.1.2
       '@qdrant/js-client-rest': 1.14.1(typescript@5.9.2)