@huggingface/inference 3.7.1 → 3.8.0

package/src/providers/hf-inference.ts

@@ -87,6 +87,8 @@ interface AudioToAudioOutput {
 	label: string;
 }
 
+export const EQUIVALENT_SENTENCE_TRANSFORMERS_TASKS = ["feature-extraction", "sentence-similarity"] as const;
+
 export class HFInferenceTask extends TaskProviderHelper {
 	constructor() {
 		super("hf-inference", `${HF_ROUTER_URL}/hf-inference`);
@@ -385,13 +387,13 @@ export class HFInferenceQuestionAnsweringTask extends HFInferenceTask implements
 							typeof elem.end === "number" &&
 							typeof elem.score === "number" &&
 							typeof elem.start === "number"
-					)
+				  )
 				: typeof response === "object" &&
-					!!response &&
-					typeof response.answer === "string" &&
-					typeof response.end === "number" &&
-					typeof response.score === "number" &&
-					typeof response.start === "number"
+				  !!response &&
+				  typeof response.answer === "string" &&
+				  typeof response.end === "number" &&
+				  typeof response.score === "number" &&
+				  typeof response.start === "number"
 		) {
 			return Array.isArray(response) ? response[0] : response;
 		}

package/src/snippets/getInferenceSnippets.ts

@@ -11,6 +11,8 @@ import type { ChatCompletionInputMessage, GenerationParameters } from "@huggingf
 import { makeRequestOptionsFromResolvedModel } from "../lib/makeRequestOptions";
 import type { InferenceProvider, InferenceTask, RequestArgs } from "../types";
 import { templates } from "./templates.exported";
+import type { InferenceProviderModelMapping } from "../lib/getInferenceProviderMapping";
+import { getProviderHelper } from "../lib/getProviderHelper";
 
 const PYTHON_CLIENTS = ["huggingface_hub", "fal_client", "requests", "openai"] as const;
 const JS_CLIENTS = ["fetch", "huggingface.js", "openai"] as const;
@@ -35,6 +37,7 @@ interface TemplateParams {
 	model?: ModelDataMinimal;
 	provider?: InferenceProvider;
 	providerModelId?: string;
+	billTo?: string;
 	methodName?: string; // specific to snippetBasic
 	importBase64?: boolean; // specific to snippetImportRequests
 	importJson?: boolean; // specific to snippetImportRequests
@@ -116,9 +119,11 @@ const snippetGenerator = (templateName: string, inputPreparationFn?: InputPrepar
 		model: ModelDataMinimal,
 		accessToken: string,
 		provider: InferenceProvider,
-		providerModelId?: string,
+		inferenceProviderMapping?: InferenceProviderModelMapping,
+		billTo?: string,
 		opts?: Record<string, unknown>
 	): InferenceSnippet[] => {
+		const providerModelId = inferenceProviderMapping?.providerId ?? model.id;
 		/// Hacky: hard-code conversational templates here
 		let task = model.pipeline_tag as InferenceTask;
 		if (
@@ -130,17 +135,27 @@ const snippetGenerator = (templateName: string, inputPreparationFn?: InputPrepar
 			inputPreparationFn = prepareConversationalInput;
 			task = "conversational";
 		}
+		let providerHelper: ReturnType<typeof getProviderHelper>;
+		try {
+			providerHelper = getProviderHelper(provider, task);
+		} catch (e) {
+			console.error(`Failed to get provider helper for ${provider} (${task})`, e);
+			return [];
+		}
 		/// Prepare inputs + make request
 		const inputs = inputPreparationFn ? inputPreparationFn(model, opts) : { inputs: getModelInputSnippet(model) };
 		const request = makeRequestOptionsFromResolvedModel(
-			providerModelId ?? model.id,
+			providerModelId,
+			providerHelper,
 			{
-				accessToken: accessToken,
-				provider: provider,
+				accessToken,
+				provider,
 				...inputs,
 			} as RequestArgs,
+			inferenceProviderMapping,
 			{
-				task: task,
+				task,
+				billTo,
 			}
 		);
 
@@ -179,6 +194,7 @@ const snippetGenerator = (templateName: string, inputPreparationFn?: InputPrepar
 			model,
 			provider,
 			providerModelId: providerModelId ?? model.id,
+			billTo,
 		};
 
 		/// Iterate over clients => check if a snippet exists => generate
@@ -266,7 +282,8 @@ const snippets: Partial<
 			model: ModelDataMinimal,
 			accessToken: string,
 			provider: InferenceProvider,
-			providerModelId?: string,
+			inferenceProviderMapping?: InferenceProviderModelMapping,
+			billTo?: string,
 			opts?: Record<string, unknown>
 		) => InferenceSnippet[]
 	>
@@ -306,11 +323,12 @@ export function getInferenceSnippets(
 	model: ModelDataMinimal,
 	accessToken: string,
 	provider: InferenceProvider,
-	providerModelId?: string,
+	inferenceProviderMapping?: InferenceProviderModelMapping,
+	billTo?: string,
 	opts?: Record<string, unknown>
 ): InferenceSnippet[] {
 	return model.pipeline_tag && model.pipeline_tag in snippets
-		? snippets[model.pipeline_tag]?.(model, accessToken, provider, providerModelId, opts) ?? []
+		? snippets[model.pipeline_tag]?.(model, accessToken, provider, inferenceProviderMapping, billTo, opts) ?? []
 		: [];
 }
 

package/src/snippets/templates.exported.ts

@@ -2,25 +2,25 @@
 export const templates: Record<string, Record<string, Record<string, string>>> = {
   "js": {
     "fetch": {
-      "basic": "async function query(data) {\n\tconst response = await fetch(\n\t\t\"{{ fullUrl }}\",\n\t\t{\n\t\t\theaders: {\n\t\t\t\tAuthorization: \"{{ authorizationHeader }}\",\n\t\t\t\t\"Content-Type\": \"application/json\",\n\t\t\t},\n\t\t\tmethod: \"POST\",\n\t\t\tbody: JSON.stringify(data),\n\t\t}\n\t);\n\tconst result = await response.json();\n\treturn result;\n}\n\nquery({ inputs: {{ providerInputs.asObj.inputs }} }).then((response) => {\n    console.log(JSON.stringify(response));\n});",
-      "basicAudio": "async function query(data) {\n\tconst response = await fetch(\n\t\t\"{{ fullUrl }}\",\n\t\t{\n\t\t\theaders: {\n\t\t\t\tAuthorization: \"{{ authorizationHeader }}\",\n\t\t\t\t\"Content-Type\": \"audio/flac\"\n\t\t\t},\n\t\t\tmethod: \"POST\",\n\t\t\tbody: JSON.stringify(data),\n\t\t}\n\t);\n\tconst result = await response.json();\n\treturn result;\n}\n\nquery({ inputs: {{ providerInputs.asObj.inputs }} }).then((response) => {\n    console.log(JSON.stringify(response));\n});",
-      "basicImage": "async function query(data) {\n\tconst response = await fetch(\n\t\t\"{{ fullUrl }}\",\n\t\t{\n\t\t\theaders: {\n\t\t\t\tAuthorization: \"{{ authorizationHeader }}\",\n\t\t\t\t\"Content-Type\": \"image/jpeg\"\n\t\t\t},\n\t\t\tmethod: \"POST\",\n\t\t\tbody: JSON.stringify(data),\n\t\t}\n\t);\n\tconst result = await response.json();\n\treturn result;\n}\n\nquery({ inputs: {{ providerInputs.asObj.inputs }} }).then((response) => {\n    console.log(JSON.stringify(response));\n});",
-      "textToAudio": "{% if model.library_name == \"transformers\" %}\nasync function query(data) {\n\tconst response = await fetch(\n\t\t\"{{ fullUrl }}\",\n\t\t{\n\t\t\theaders: {\n\t\t\t\tAuthorization: \"{{ authorizationHeader }}\",\n\t\t\t\t\"Content-Type\": \"application/json\",\n\t\t\t},\n\t\t\tmethod: \"POST\",\n\t\t\tbody: JSON.stringify(data),\n\t\t}\n\t);\n\tconst result = await response.blob();\n    return result;\n}\n\nquery({ inputs: {{ providerInputs.asObj.inputs }} }).then((response) => {\n    // Returns a byte object of the Audio wavform. Use it directly!\n});\n{% else %}\nasync function query(data) {\n\tconst response = await fetch(\n\t\t\"{{ fullUrl }}\",\n\t\t{\n\t\t\theaders: {\n\t\t\t\tAuthorization: \"{{ authorizationHeader }}\",\n\t\t\t\t\"Content-Type\": \"application/json\",\n\t\t\t},\n\t\t\tmethod: \"POST\",\n\t\t\tbody: JSON.stringify(data),\n\t\t}\n\t);\n    const result = await response.json();\n    return result;\n}\n\nquery({ inputs: {{ providerInputs.asObj.inputs }} }).then((response) => {\n    console.log(JSON.stringify(response));\n});\n{% endif %} ",
-      "textToImage": "async function query(data) {\n\tconst response = await fetch(\n\t\t\"{{ fullUrl }}\",\n\t\t{\n\t\t\theaders: {\n\t\t\t\tAuthorization: \"{{ authorizationHeader }}\",\n\t\t\t\t\"Content-Type\": \"application/json\",\n\t\t\t},\n\t\t\tmethod: \"POST\",\n\t\t\tbody: JSON.stringify(data),\n\t\t}\n\t);\n\tconst result = await response.blob();\n\treturn result;\n}\n\n\nquery({ {{ providerInputs.asTsString }} }).then((response) => {\n    // Use image\n});",
-      "zeroShotClassification": "async function query(data) {\n    const response = await fetch(\n\t\t\"{{ fullUrl }}\",\n        {\n            headers: {\n\t\t\t\tAuthorization: \"{{ authorizationHeader }}\",\n                \"Content-Type\": \"application/json\",\n            },\n            method: \"POST\",\n            body: JSON.stringify(data),\n        }\n    );\n    const result = await response.json();\n    return result;\n}\n\nquery({\n    inputs: {{ providerInputs.asObj.inputs }},\n    parameters: { candidate_labels: [\"refund\", \"legal\", \"faq\"] }\n}).then((response) => {\n    console.log(JSON.stringify(response));\n});"
+      "basic": "async function query(data) {\n\tconst response = await fetch(\n\t\t\"{{ fullUrl }}\",\n\t\t{\n\t\t\theaders: {\n\t\t\t\tAuthorization: \"{{ authorizationHeader }}\",\n\t\t\t\t\"Content-Type\": \"application/json\",\n{% if billTo %}\n\t\t\t\t\"X-HF-Bill-To\": \"{{ billTo }}\",\n{% endif %}\t\t\t},\n\t\t\tmethod: \"POST\",\n\t\t\tbody: JSON.stringify(data),\n\t\t}\n\t);\n\tconst result = await response.json();\n\treturn result;\n}\n\nquery({ inputs: {{ providerInputs.asObj.inputs }} }).then((response) => {\n    console.log(JSON.stringify(response));\n});",
+      "basicAudio": "async function query(data) {\n\tconst response = await fetch(\n\t\t\"{{ fullUrl }}\",\n\t\t{\n\t\t\theaders: {\n\t\t\t\tAuthorization: \"{{ authorizationHeader }}\",\n\t\t\t\t\"Content-Type\": \"audio/flac\",\n{% if billTo %}\n\t\t\t\t\"X-HF-Bill-To\": \"{{ billTo }}\",\n{% endif %}\t\t\t},\n\t\t\tmethod: \"POST\",\n\t\t\tbody: JSON.stringify(data),\n\t\t}\n\t);\n\tconst result = await response.json();\n\treturn result;\n}\n\nquery({ inputs: {{ providerInputs.asObj.inputs }} }).then((response) => {\n    console.log(JSON.stringify(response));\n});",
+      "basicImage": "async function query(data) {\n\tconst response = await fetch(\n\t\t\"{{ fullUrl }}\",\n\t\t{\n\t\t\theaders: {\n\t\t\t\tAuthorization: \"{{ authorizationHeader }}\",\n\t\t\t\t\"Content-Type\": \"image/jpeg\",\n{% if billTo %}\n\t\t\t\t\"X-HF-Bill-To\": \"{{ billTo }}\",\n{% endif %}\t\t\t},\n\t\t\tmethod: \"POST\",\n\t\t\tbody: JSON.stringify(data),\n\t\t}\n\t);\n\tconst result = await response.json();\n\treturn result;\n}\n\nquery({ inputs: {{ providerInputs.asObj.inputs }} }).then((response) => {\n    console.log(JSON.stringify(response));\n});",
+      "textToAudio": "{% if model.library_name == \"transformers\" %}\nasync function query(data) {\n\tconst response = await fetch(\n\t\t\"{{ fullUrl }}\",\n\t\t{\n\t\t\theaders: {\n\t\t\t\tAuthorization: \"{{ authorizationHeader }}\",\n\t\t\t\t\"Content-Type\": \"application/json\",\n{% if billTo %}\n\t\t\t\t\"X-HF-Bill-To\": \"{{ billTo }}\",\n{% endif %}\t\t\t},\n\t\t\tmethod: \"POST\",\n\t\t\tbody: JSON.stringify(data),\n\t\t}\n\t);\n\tconst result = await response.blob();\n    return result;\n}\n\nquery({ inputs: {{ providerInputs.asObj.inputs }} }).then((response) => {\n    // Returns a byte object of the Audio wavform. Use it directly!\n});\n{% else %}\nasync function query(data) {\n\tconst response = await fetch(\n\t\t\"{{ fullUrl }}\",\n\t\t{\n\t\t\theaders: {\n\t\t\t\tAuthorization: \"{{ authorizationHeader }}\",\n\t\t\t\t\"Content-Type\": \"application/json\",\n\t\t\t},\n\t\t\tmethod: \"POST\",\n\t\t\tbody: JSON.stringify(data),\n\t\t}\n\t);\n    const result = await response.json();\n    return result;\n}\n\nquery({ inputs: {{ providerInputs.asObj.inputs }} }).then((response) => {\n    console.log(JSON.stringify(response));\n});\n{% endif %} ",
+      "textToImage": "async function query(data) {\n\tconst response = await fetch(\n\t\t\"{{ fullUrl }}\",\n\t\t{\n\t\t\theaders: {\n\t\t\t\tAuthorization: \"{{ authorizationHeader }}\",\n\t\t\t\t\"Content-Type\": \"application/json\",\n{% if billTo %}\n\t\t\t\t\"X-HF-Bill-To\": \"{{ billTo }}\",\n{% endif %}\t\t\t},\n\t\t\tmethod: \"POST\",\n\t\t\tbody: JSON.stringify(data),\n\t\t}\n\t);\n\tconst result = await response.blob();\n\treturn result;\n}\n\n\nquery({ {{ providerInputs.asTsString }} }).then((response) => {\n    // Use image\n});",
+      "zeroShotClassification": "async function query(data) {\n    const response = await fetch(\n\t\t\"{{ fullUrl }}\",\n        {\n            headers: {\n\t\t\t\tAuthorization: \"{{ authorizationHeader }}\",\n                \"Content-Type\": \"application/json\",\n{% if billTo %}\n                \"X-HF-Bill-To\": \"{{ billTo }}\",\n{% endif %}         },\n            method: \"POST\",\n            body: JSON.stringify(data),\n        }\n    );\n    const result = await response.json();\n    return result;\n}\n\nquery({\n    inputs: {{ providerInputs.asObj.inputs }},\n    parameters: { candidate_labels: [\"refund\", \"legal\", \"faq\"] }\n}).then((response) => {\n    console.log(JSON.stringify(response));\n});"
     },
     "huggingface.js": {
-      "basic": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst output = await client.{{ methodName }}({\n\tmodel: \"{{ model.id }}\",\n\tinputs: {{ inputs.asObj.inputs }},\n\tprovider: \"{{ provider }}\",\n});\n\nconsole.log(output);",
-      "basicAudio": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst data = fs.readFileSync({{inputs.asObj.inputs}});\n\nconst output = await client.{{ methodName }}({\n\tdata,\n\tmodel: \"{{ model.id }}\",\n\tprovider: \"{{ provider }}\",\n});\n\nconsole.log(output);",
-      "basicImage": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst data = fs.readFileSync({{inputs.asObj.inputs}});\n\nconst output = await client.{{ methodName }}({\n\tdata,\n\tmodel: \"{{ model.id }}\",\n\tprovider: \"{{ provider }}\",\n});\n\nconsole.log(output);",
-      "conversational": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst chatCompletion = await client.chatCompletion({\n    provider: \"{{ provider }}\",\n    model: \"{{ model.id }}\",\n{{ inputs.asTsString }}\n});\n\nconsole.log(chatCompletion.choices[0].message);",
-      "conversationalStream": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nlet out = \"\";\n\nconst stream = await client.chatCompletionStream({\n    provider: \"{{ provider }}\",\n    model: \"{{ model.id }}\",\n{{ inputs.asTsString }}\n});\n\nfor await (const chunk of stream) {\n\tif (chunk.choices && chunk.choices.length > 0) {\n\t\tconst newContent = chunk.choices[0].delta.content;\n\t\tout += newContent;\n\t\tconsole.log(newContent);\n\t}  \n}",
-      "textToImage": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst image = await client.textToImage({\n    provider: \"{{ provider }}\",\n    model: \"{{ model.id }}\",\n\tinputs: {{ inputs.asObj.inputs }},\n\tparameters: { num_inference_steps: 5 },\n});\n/// Use the generated image (it's a Blob)",
-      "textToVideo": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst image = await client.textToVideo({\n    provider: \"{{ provider }}\",\n    model: \"{{ model.id }}\",\n\tinputs: {{ inputs.asObj.inputs }},\n});\n// Use the generated video (it's a Blob)"
+      "basic": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst output = await client.{{ methodName }}({\n\tmodel: \"{{ model.id }}\",\n\tinputs: {{ inputs.asObj.inputs }},\n\tprovider: \"{{ provider }}\",\n}{% if billTo %}, {\n\tbillTo: \"{{ billTo }}\",\n}{% endif %});\n\nconsole.log(output);",
+      "basicAudio": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst data = fs.readFileSync({{inputs.asObj.inputs}});\n\nconst output = await client.{{ methodName }}({\n\tdata,\n\tmodel: \"{{ model.id }}\",\n\tprovider: \"{{ provider }}\",\n}{% if billTo %}, {\n\tbillTo: \"{{ billTo }}\",\n}{% endif %});\n\nconsole.log(output);",
+      "basicImage": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst data = fs.readFileSync({{inputs.asObj.inputs}});\n\nconst output = await client.{{ methodName }}({\n\tdata,\n\tmodel: \"{{ model.id }}\",\n\tprovider: \"{{ provider }}\",\n}{% if billTo %}, {\n\tbillTo: \"{{ billTo }}\",\n}{% endif %});\n\nconsole.log(output);",
+      "conversational": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst chatCompletion = await client.chatCompletion({\n    provider: \"{{ provider }}\",\n    model: \"{{ model.id }}\",\n{{ inputs.asTsString }}\n}{% if billTo %}, {\n    billTo: \"{{ billTo }}\",\n}{% endif %});\n\nconsole.log(chatCompletion.choices[0].message);",
+      "conversationalStream": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nlet out = \"\";\n\nconst stream = await client.chatCompletionStream({\n    provider: \"{{ provider }}\",\n    model: \"{{ model.id }}\",\n{{ inputs.asTsString }}\n}{% if billTo %}, {\n    billTo: \"{{ billTo }}\",\n}{% endif %});\n\nfor await (const chunk of stream) {\n\tif (chunk.choices && chunk.choices.length > 0) {\n\t\tconst newContent = chunk.choices[0].delta.content;\n\t\tout += newContent;\n\t\tconsole.log(newContent);\n\t}  \n}",
+      "textToImage": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst image = await client.textToImage({\n    provider: \"{{ provider }}\",\n    model: \"{{ model.id }}\",\n\tinputs: {{ inputs.asObj.inputs }},\n\tparameters: { num_inference_steps: 5 },\n}{% if billTo %}, {\n    billTo: \"{{ billTo }}\",\n}{% endif %});\n/// Use the generated image (it's a Blob)",
+      "textToVideo": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst image = await client.textToVideo({\n    provider: \"{{ provider }}\",\n    model: \"{{ model.id }}\",\n\tinputs: {{ inputs.asObj.inputs }},\n}{% if billTo %}, {\n    billTo: \"{{ billTo }}\",\n}{% endif %});\n// Use the generated video (it's a Blob)"
     },
     "openai": {
-      "conversational": "import { OpenAI } from \"openai\";\n\nconst client = new OpenAI({\n\tbaseURL: \"{{ baseUrl }}\",\n\tapiKey: \"{{ accessToken }}\",\n});\n\nconst chatCompletion = await client.chat.completions.create({\n\tmodel: \"{{ providerModelId }}\",\n{{ inputs.asTsString }}\n});\n\nconsole.log(chatCompletion.choices[0].message);",
-      "conversationalStream": "import { OpenAI } from \"openai\";\n\nconst client = new OpenAI({\n\tbaseURL: \"{{ baseUrl }}\",\n\tapiKey: \"{{ accessToken }}\",\n});\n\nconst stream = await client.chat.completions.create({\n    model: \"{{ providerModelId }}\",\n{{ inputs.asTsString }}\n    stream: true,\n});\n\nfor await (const chunk of stream) {\n    process.stdout.write(chunk.choices[0]?.delta?.content || \"\");\n}"
+      "conversational": "import { OpenAI } from \"openai\";\n\nconst client = new OpenAI({\n\tbaseURL: \"{{ baseUrl }}\",\n\tapiKey: \"{{ accessToken }}\",\n{% if billTo %}\n\tdefaultHeaders: {\n\t\t\"X-HF-Bill-To\": \"{{ billTo }}\" \n\t}\n{% endif %}\n});\n\nconst chatCompletion = await client.chat.completions.create({\n\tmodel: \"{{ providerModelId }}\",\n{{ inputs.asTsString }}\n});\n\nconsole.log(chatCompletion.choices[0].message);",
+      "conversationalStream": "import { OpenAI } from \"openai\";\n\nconst client = new OpenAI({\n\tbaseURL: \"{{ baseUrl }}\",\n\tapiKey: \"{{ accessToken }}\",\n{% if billTo %}\n    defaultHeaders: {\n\t\t\"X-HF-Bill-To\": \"{{ billTo }}\" \n\t}\n{% endif %}\n});\n\nconst stream = await client.chat.completions.create({\n    model: \"{{ providerModelId }}\",\n{{ inputs.asTsString }}\n    stream: true,\n});\n\nfor await (const chunk of stream) {\n    process.stdout.write(chunk.choices[0]?.delta?.content || \"\");\n}"
     }
   },
   "python": {
@@ -35,13 +35,13 @@ export const templates: Record<string, Record<string, Record<string, string>>> =
       "conversationalStream": "stream = client.chat.completions.create(\n    model=\"{{ model.id }}\",\n{{ inputs.asPythonString }}\n    stream=True,\n)\n\nfor chunk in stream:\n    print(chunk.choices[0].delta.content, end=\"\") ",
       "documentQuestionAnswering": "output = client.document_question_answering(\n    \"{{ inputs.asObj.image }}\",\n    question=\"{{ inputs.asObj.question }}\",\n    model=\"{{ model.id }}\",\n) ",
       "imageToImage": "# output is a PIL.Image object\nimage = client.image_to_image(\n    \"{{ inputs.asObj.inputs }}\",\n    prompt=\"{{ inputs.asObj.parameters.prompt }}\",\n    model=\"{{ model.id }}\",\n) ",
-      "importInferenceClient": "from huggingface_hub import InferenceClient\n\nclient = InferenceClient(\n    provider=\"{{ provider }}\",\n    api_key=\"{{ accessToken }}\",\n)",
+      "importInferenceClient": "from huggingface_hub import InferenceClient\n\nclient = InferenceClient(\n    provider=\"{{ provider }}\",\n    api_key=\"{{ accessToken }}\",\n{% if billTo %}\n    bill_to=\"{{ billTo }}\",\n{% endif %}\n)",
       "textToImage": "# output is a PIL.Image object\nimage = client.text_to_image(\n    {{ inputs.asObj.inputs }},\n    model=\"{{ model.id }}\",\n) ",
       "textToVideo": "video = client.text_to_video(\n    {{ inputs.asObj.inputs }},\n    model=\"{{ model.id }}\",\n) "
     },
     "openai": {
-      "conversational": "from openai import OpenAI\n\nclient = OpenAI(\n    base_url=\"{{ baseUrl }}\",\n    api_key=\"{{ accessToken }}\"\n)\n\ncompletion = client.chat.completions.create(\n    model=\"{{ providerModelId }}\",\n{{ inputs.asPythonString }}\n)\n\nprint(completion.choices[0].message) ",
-      "conversationalStream": "from openai import OpenAI\n\nclient = OpenAI(\n    base_url=\"{{ baseUrl }}\",\n    api_key=\"{{ accessToken }}\"\n)\n\nstream = client.chat.completions.create(\n    model=\"{{ providerModelId }}\",\n{{ inputs.asPythonString }}\n    stream=True,\n)\n\nfor chunk in stream:\n    print(chunk.choices[0].delta.content, end=\"\")"
+      "conversational": "from openai import OpenAI\n\nclient = OpenAI(\n    base_url=\"{{ baseUrl }}\",\n    api_key=\"{{ accessToken }}\",\n{% if billTo %}\n    default_headers={\n        \"X-HF-Bill-To\": \"{{ billTo }}\"\n    }\n{% endif %}\n)\n\ncompletion = client.chat.completions.create(\n    model=\"{{ providerModelId }}\",\n{{ inputs.asPythonString }}\n)\n\nprint(completion.choices[0].message) ",
+      "conversationalStream": "from openai import OpenAI\n\nclient = OpenAI(\n    base_url=\"{{ baseUrl }}\",\n    api_key=\"{{ accessToken }}\",\n{% if billTo %}\n    default_headers={\n        \"X-HF-Bill-To\": \"{{ billTo }}\"\n    }\n{% endif %}\n)\n\nstream = client.chat.completions.create(\n    model=\"{{ providerModelId }}\",\n{{ inputs.asPythonString }}\n    stream=True,\n)\n\nfor chunk in stream:\n    print(chunk.choices[0].delta.content, end=\"\")"
     },
     "requests": {
       "basic": "def query(payload):\n    response = requests.post(API_URL, headers=headers, json=payload)\n    return response.json()\n\noutput = query({\n    \"inputs\": {{ providerInputs.asObj.inputs }},\n}) ",
@@ -51,7 +51,7 @@ export const templates: Record<string, Record<string, Record<string, string>>> =
       "conversationalStream": "def query(payload):\n    response = requests.post(API_URL, headers=headers, json=payload, stream=True)\n    for line in response.iter_lines():\n        if not line.startswith(b\"data:\"):\n            continue\n        if line.strip() == b\"data: [DONE]\":\n            return\n        yield json.loads(line.decode(\"utf-8\").lstrip(\"data:\").rstrip(\"/n\"))\n\nchunks = query({\n{{ providerInputs.asJsonString }},\n    \"stream\": True,\n})\n\nfor chunk in chunks:\n    print(chunk[\"choices\"][0][\"delta\"][\"content\"], end=\"\")",
       "documentQuestionAnswering": "def query(payload):\n    with open(payload[\"image\"], \"rb\") as f:\n        img = f.read()\n        payload[\"image\"] = base64.b64encode(img).decode(\"utf-8\")\n    response = requests.post(API_URL, headers=headers, json=payload)\n    return response.json()\n\noutput = query({\n    \"inputs\": {\n        \"image\": \"{{ inputs.asObj.image }}\",\n        \"question\": \"{{ inputs.asObj.question }}\",\n    },\n}) ",
       "imageToImage": "def query(payload):\n    with open(payload[\"inputs\"], \"rb\") as f:\n        img = f.read()\n        payload[\"inputs\"] = base64.b64encode(img).decode(\"utf-8\")\n    response = requests.post(API_URL, headers=headers, json=payload)\n    return response.content\n\nimage_bytes = query({\n{{ providerInputs.asJsonString }}\n})\n\n# You can access the image with PIL.Image for example\nimport io\nfrom PIL import Image\nimage = Image.open(io.BytesIO(image_bytes)) ",
-      "importRequests": "{% if importBase64 %}\nimport base64\n{% endif %}\n{% if importJson %}\nimport json\n{% endif %}\nimport requests\n\nAPI_URL = \"{{ fullUrl }}\"\nheaders = {\"Authorization\": \"{{ authorizationHeader }}\"}",
+      "importRequests": "{% if importBase64 %}\nimport base64\n{% endif %}\n{% if importJson %}\nimport json\n{% endif %}\nimport requests\n\nAPI_URL = \"{{ fullUrl }}\"\nheaders = {\n    \"Authorization\": \"{{ authorizationHeader }}\",\n{% if billTo %}\n    \"X-HF-Bill-To\": \"{{ billTo }}\"\n{% endif %}\n}",
       "tabular": "def query(payload):\n    response = requests.post(API_URL, headers=headers, json=payload)\n    return response.content\n\nresponse = query({\n    \"inputs\": {\n        \"data\": {{ providerInputs.asObj.inputs }}\n    },\n}) ",
       "textToAudio": "{% if model.library_name == \"transformers\" %}\ndef query(payload):\n    response = requests.post(API_URL, headers=headers, json=payload)\n    return response.content\n\naudio_bytes = query({\n    \"inputs\": {{ providerInputs.asObj.inputs }},\n})\n# You can access the audio with IPython.display for example\nfrom IPython.display import Audio\nAudio(audio_bytes)\n{% else %}\ndef query(payload):\n    response = requests.post(API_URL, headers=headers, json=payload)\n    return response.json()\n\naudio, sampling_rate = query({\n    \"inputs\": {{ providerInputs.asObj.inputs }},\n})\n# You can access the audio with IPython.display for example\nfrom IPython.display import Audio\nAudio(audio, rate=sampling_rate)\n{% endif %} ",
       "textToImage": "{% if provider == \"hf-inference\" %}\ndef query(payload):\n    response = requests.post(API_URL, headers=headers, json=payload)\n    return response.content\n\nimage_bytes = query({\n    \"inputs\": {{ providerInputs.asObj.inputs }},\n})\n\n# You can access the image with PIL.Image for example\nimport io\nfrom PIL import Image\nimage = Image.open(io.BytesIO(image_bytes))\n{% endif %}",
@@ -61,12 +61,12 @@ export const templates: Record<string, Record<string, Record<string, string>>> =
   },
   "sh": {
     "curl": {
-      "basic": "curl {{ fullUrl }} \\\n    -X POST \\\n    -H 'Authorization: {{ authorizationHeader }}' \\\n    -H 'Content-Type: application/json' \\\n    -d '{\n{{ providerInputs.asCurlString }}\n    }'",
-      "basicAudio": "curl {{ fullUrl }} \\\n    -X POST \\\n    -H 'Authorization: {{ authorizationHeader }}' \\\n    -H 'Content-Type: audio/flac' \\\n    --data-binary @{{ providerInputs.asObj.inputs }}",
-      "basicImage": "curl {{ fullUrl }} \\\n    -X POST \\\n    -H 'Authorization: {{ authorizationHeader }}' \\\n    -H 'Content-Type: image/jpeg' \\\n    --data-binary @{{ providerInputs.asObj.inputs }}",
-      "conversational": "curl {{ fullUrl }} \\\n    -H 'Authorization: {{ authorizationHeader }}' \\\n    -H 'Content-Type: application/json' \\\n    -d '{\n{{ providerInputs.asCurlString }},\n        \"stream\": false\n    }'",
-      "conversationalStream": "curl {{ fullUrl }} \\\n    -H 'Authorization: {{ authorizationHeader }}' \\\n    -H 'Content-Type: application/json' \\\n    -d '{\n{{ providerInputs.asCurlString }},\n        \"stream\": true\n    }'",
-      "zeroShotClassification": "curl {{ fullUrl }} \\\n    -X POST \\\n    -d '{\"inputs\": {{ providerInputs.asObj.inputs }}, \"parameters\": {\"candidate_labels\": [\"refund\", \"legal\", \"faq\"]}}' \\\n    -H 'Content-Type: application/json' \\\n    -H 'Authorization: {{ authorizationHeader }}'"
+      "basic": "curl {{ fullUrl }} \\\n    -X POST \\\n    -H 'Authorization: {{ authorizationHeader }}' \\\n    -H 'Content-Type: application/json' \\\n{% if billTo %}\n    -H 'X-HF-Bill-To: {{ billTo }}' \\\n{% endif %}\n    -d '{\n{{ providerInputs.asCurlString }}\n    }'",
+      "basicAudio": "curl {{ fullUrl }} \\\n    -X POST \\\n    -H 'Authorization: {{ authorizationHeader }}' \\\n    -H 'Content-Type: audio/flac' \\\n{% if billTo %}\n    -H 'X-HF-Bill-To: {{ billTo }}' \\\n{% endif %}\n    --data-binary @{{ providerInputs.asObj.inputs }}",
+      "basicImage": "curl {{ fullUrl }} \\\n    -X POST \\\n    -H 'Authorization: {{ authorizationHeader }}' \\\n    -H 'Content-Type: image/jpeg' \\\n{% if billTo %}\n    -H 'X-HF-Bill-To: {{ billTo }}' \\\n{% endif %}\n    --data-binary @{{ providerInputs.asObj.inputs }}",
+      "conversational": "curl {{ fullUrl }} \\\n    -H 'Authorization: {{ authorizationHeader }}' \\\n    -H 'Content-Type: application/json' \\\n{% if billTo %}\n    -H 'X-HF-Bill-To: {{ billTo }}' \\\n{% endif %}\n    -d '{\n{{ providerInputs.asCurlString }},\n        \"stream\": false\n    }'",
+      "conversationalStream": "curl {{ fullUrl }} \\\n    -H 'Authorization: {{ authorizationHeader }}' \\\n    -H 'Content-Type: application/json' \\\n{% if billTo %}\n    -H 'X-HF-Bill-To: {{ billTo }}' \\\n{% endif %}\n    -d '{\n{{ providerInputs.asCurlString }},\n        \"stream\": true\n    }'",
+      "zeroShotClassification": "curl {{ fullUrl }} \\\n    -X POST \\\n    -d '{\"inputs\": {{ providerInputs.asObj.inputs }}, \"parameters\": {\"candidate_labels\": [\"refund\", \"legal\", \"faq\"]}}' \\\n    -H 'Content-Type: application/json' \\\n    -H 'Authorization: {{ authorizationHeader }}'\n{% if billTo %} \\\n    -H 'X-HF-Bill-To: {{ billTo }}'\n{% endif %}"
     }
   }
 } as const;

package/src/tasks/audio/audioClassification.ts

@@ -17,7 +17,7 @@ export async function audioClassification(
 ): Promise<AudioClassificationOutput> {
 	const providerHelper = getProviderHelper(args.provider ?? "hf-inference", "audio-classification");
 	const payload = preparePayload(args);
-	const { data: res } = await innerRequest<AudioClassificationOutput>(payload, {
+	const { data: res } = await innerRequest<AudioClassificationOutput>(payload, providerHelper, {
 		...options,
 		task: "audio-classification",
 	});

package/src/tasks/audio/audioToAudio.ts

@@ -38,7 +38,7 @@ export interface AudioToAudioOutput {
 export async function audioToAudio(args: AudioToAudioArgs, options?: Options): Promise<AudioToAudioOutput[]> {
 	const providerHelper = getProviderHelper(args.provider ?? "hf-inference", "audio-to-audio");
 	const payload = preparePayload(args);
-	const { data: res } = await innerRequest<AudioToAudioOutput>(payload, {
+	const { data: res } = await innerRequest<AudioToAudioOutput>(payload, providerHelper, {
 		...options,
 		task: "audio-to-audio",
 	});

package/src/tasks/audio/automaticSpeechRecognition.ts

@@ -20,7 +20,7 @@ export async function automaticSpeechRecognition(
 ): Promise<AutomaticSpeechRecognitionOutput> {
 	const providerHelper = getProviderHelper(args.provider ?? "hf-inference", "automatic-speech-recognition");
 	const payload = await buildPayload(args);
-	const { data: res } = await innerRequest<AutomaticSpeechRecognitionOutput>(payload, {
+	const { data: res } = await innerRequest<AutomaticSpeechRecognitionOutput>(payload, providerHelper, {
 		...options,
 		task: "automatic-speech-recognition",
 	});

package/src/tasks/audio/textToSpeech.ts

@@ -14,7 +14,7 @@ interface OutputUrlTextToSpeechGeneration {
 export async function textToSpeech(args: TextToSpeechArgs, options?: Options): Promise<Blob> {
 	const provider = args.provider ?? "hf-inference";
 	const providerHelper = getProviderHelper(provider, "text-to-speech");
-	const { data: res } = await innerRequest<Blob | OutputUrlTextToSpeechGeneration>(args, {
+	const { data: res } = await innerRequest<Blob | OutputUrlTextToSpeechGeneration>(args, providerHelper, {
 		...options,
 		task: "text-to-speech",
 	});

package/src/tasks/custom/request.ts

@@ -1,3 +1,4 @@
+import { getProviderHelper } from "../../lib/getProviderHelper";
 import type { InferenceTask, Options, RequestArgs } from "../../types";
 import { innerRequest } from "../../utils/request";
 
@@ -15,6 +16,7 @@ export async function request<T>(
 	console.warn(
 		"The request method is deprecated and will be removed in a future version of huggingface.js. Use specific task functions instead."
 	);
-	const result = await innerRequest<T>(args, options);
+	const providerHelper = getProviderHelper(args.provider ?? "hf-inference", options?.task);
+	const result = await innerRequest<T>(args, providerHelper, options);
 	return result.data;
 }

package/src/tasks/custom/streamingRequest.ts

@@ -1,5 +1,7 @@
+import { getProviderHelper } from "../../lib/getProviderHelper";
 import type { InferenceTask, Options, RequestArgs } from "../../types";
 import { innerStreamingRequest } from "../../utils/request";
+
 /**
  * Primitive to make custom inference calls that expect server-sent events, and returns the response through a generator
  * @deprecated Use specific task functions instead. This function will be removed in a future version.
@@ -14,5 +16,6 @@ export async function* streamingRequest<T>(
 	console.warn(
 		"The streamingRequest method is deprecated and will be removed in a future version of huggingface.js. Use specific task functions instead."
 	);
-	yield* innerStreamingRequest(args, options);
+	const providerHelper = getProviderHelper(args.provider ?? "hf-inference", options?.task);
+	yield* innerStreamingRequest(args, providerHelper, options);
 }

package/src/tasks/cv/imageClassification.ts

@@ -16,7 +16,7 @@ export async function imageClassification(
 ): Promise<ImageClassificationOutput> {
 	const providerHelper = getProviderHelper(args.provider ?? "hf-inference", "image-classification");
 	const payload = preparePayload(args);
-	const { data: res } = await innerRequest<ImageClassificationOutput>(payload, {
+	const { data: res } = await innerRequest<ImageClassificationOutput>(payload, providerHelper, {
 		...options,
 		task: "image-classification",
 	});

package/src/tasks/cv/imageSegmentation.ts

@@ -16,7 +16,7 @@ export async function imageSegmentation(
 ): Promise<ImageSegmentationOutput> {
 	const providerHelper = getProviderHelper(args.provider ?? "hf-inference", "image-segmentation");
 	const payload = preparePayload(args);
-	const { data: res } = await innerRequest<ImageSegmentationOutput>(payload, {
+	const { data: res } = await innerRequest<ImageSegmentationOutput>(payload, providerHelper, {
 		...options,
 		task: "image-segmentation",
 	});

package/src/tasks/cv/imageToImage.ts

@@ -27,7 +27,7 @@ export async function imageToImage(args: ImageToImageArgs, options?: Options): P
 			),
 		};
 	}
-	const { data: res } = await innerRequest<Blob>(reqArgs, {
+	const { data: res } = await innerRequest<Blob>(reqArgs, providerHelper, {
 		...options,
 		task: "image-to-image",
 	});

package/src/tasks/cv/imageToText.ts

@@ -12,7 +12,7 @@ export type ImageToTextArgs = BaseArgs & (ImageToTextInput | LegacyImageInput);
 export async function imageToText(args: ImageToTextArgs, options?: Options): Promise<ImageToTextOutput> {
 	const providerHelper = getProviderHelper(args.provider ?? "hf-inference", "image-to-text");
 	const payload = preparePayload(args);
-	const { data: res } = await innerRequest<[ImageToTextOutput]>(payload, {
+	const { data: res } = await innerRequest<[ImageToTextOutput]>(payload, providerHelper, {
 		...options,
 		task: "image-to-text",
 	});

package/src/tasks/cv/objectDetection.ts

@@ -13,7 +13,7 @@ export type ObjectDetectionArgs = BaseArgs & (ObjectDetectionInput | LegacyImage
 export async function objectDetection(args: ObjectDetectionArgs, options?: Options): Promise<ObjectDetectionOutput> {
 	const providerHelper = getProviderHelper(args.provider ?? "hf-inference", "object-detection");
 	const payload = preparePayload(args);
-	const { data: res } = await innerRequest<ObjectDetectionOutput>(payload, {
+	const { data: res } = await innerRequest<ObjectDetectionOutput>(payload, providerHelper, {
 		...options,
 		task: "object-detection",
 	});

package/src/tasks/cv/textToImage.ts

@@ -25,11 +25,11 @@ export async function textToImage(
 export async function textToImage(args: TextToImageArgs, options?: TextToImageOptions): Promise<Blob | string> {
 	const provider = args.provider ?? "hf-inference";
 	const providerHelper = getProviderHelper(provider, "text-to-image");
-	const { data: res } = await innerRequest<Record<string, unknown>>(args, {
+	const { data: res } = await innerRequest<Record<string, unknown>>(args, providerHelper, {
 		...options,
 		task: "text-to-image",
 	});
 
-	const { url, info } = await makeRequestOptions(args, { ...options, task: "text-to-image" });
+	const { url, info } = await makeRequestOptions(args, providerHelper, { ...options, task: "text-to-image" });
 	return providerHelper.getResponse(res, url, info.headers as Record<string, string>, options?.outputType);
 }

package/src/tasks/cv/textToVideo.ts

@@ -14,10 +14,14 @@ export type TextToVideoOutput = Blob;
 export async function textToVideo(args: TextToVideoArgs, options?: Options): Promise<TextToVideoOutput> {
 	const provider = args.provider ?? "hf-inference";
 	const providerHelper = getProviderHelper(provider, "text-to-video");
-	const { data: response } = await innerRequest<FalAiQueueOutput | ReplicateOutput | NovitaOutput>(args, {
-		...options,
-		task: "text-to-video",
-	});
-	const { url, info } = await makeRequestOptions(args, { ...options, task: "text-to-video" });
+	const { data: response } = await innerRequest<FalAiQueueOutput | ReplicateOutput | NovitaOutput>(
+		args,
+		providerHelper,
+		{
+			...options,
+			task: "text-to-video",
+		}
+	);
+	const { url, info } = await makeRequestOptions(args, providerHelper, { ...options, task: "text-to-video" });
 	return providerHelper.getResponse(response, url, info.headers as Record<string, string>);
 }

package/src/tasks/cv/zeroShotImageClassification.ts

@@ -46,7 +46,7 @@ export async function zeroShotImageClassification(
 ): Promise<ZeroShotImageClassificationOutput> {
 	const providerHelper = getProviderHelper(args.provider ?? "hf-inference", "zero-shot-image-classification");
 	const payload = await preparePayload(args);
-	const { data: res } = await innerRequest<ZeroShotImageClassificationOutput>(payload, {
+	const { data: res } = await innerRequest<ZeroShotImageClassificationOutput>(payload, providerHelper, {
 		...options,
 		task: "zero-shot-image-classification",
 	});

package/src/tasks/multimodal/documentQuestionAnswering.ts

@@ -30,6 +30,7 @@ export async function documentQuestionAnswering(
 	} as RequestArgs;
 	const { data: res } = await innerRequest<DocumentQuestionAnsweringOutput | DocumentQuestionAnsweringOutput[number]>(
 		reqArgs,
+		providerHelper,
 		{
 			...options,
 			task: "document-question-answering",

package/src/tasks/multimodal/visualQuestionAnswering.ts

@@ -29,7 +29,7 @@ export async function visualQuestionAnswering(
 		},
 	} as RequestArgs;
 
-	const { data: res } = await innerRequest<VisualQuestionAnsweringOutput>(reqArgs, {
+	const { data: res } = await innerRequest<VisualQuestionAnsweringOutput>(reqArgs, providerHelper, {
 		...options,
 		task: "visual-question-answering",
 	});

package/src/tasks/nlp/chatCompletion.ts

@@ -11,7 +11,7 @@ export async function chatCompletion(
 	options?: Options
 ): Promise<ChatCompletionOutput> {
 	const providerHelper = getProviderHelper(args.provider ?? "hf-inference", "conversational");
-	const { data: response } = await innerRequest<ChatCompletionOutput>(args, {
+	const { data: response } = await innerRequest<ChatCompletionOutput>(args, providerHelper, {
 		...options,
 		task: "conversational",
 	});

package/src/tasks/nlp/chatCompletionStream.ts

@@ -1,4 +1,5 @@
 import type { ChatCompletionInput, ChatCompletionStreamOutput } from "@huggingface/tasks";
+import { getProviderHelper } from "../../lib/getProviderHelper";
 import type { BaseArgs, Options } from "../../types";
 import { innerStreamingRequest } from "../../utils/request";
 
@@ -9,7 +10,8 @@ export async function* chatCompletionStream(
 	args: BaseArgs & ChatCompletionInput,
 	options?: Options
 ): AsyncGenerator<ChatCompletionStreamOutput> {
-	yield* innerStreamingRequest<ChatCompletionStreamOutput>(args, {
+	const providerHelper = getProviderHelper(args.provider ?? "hf-inference", "conversational");
+	yield* innerStreamingRequest<ChatCompletionStreamOutput>(args, providerHelper, {
 		...options,
 		task: "conversational",
 	});

package/src/tasks/nlp/featureExtraction.ts

@@ -18,7 +18,7 @@ export async function featureExtraction(
 	options?: Options
 ): Promise<FeatureExtractionOutput> {
 	const providerHelper = getProviderHelper(args.provider ?? "hf-inference", "feature-extraction");
-	const { data: res } = await innerRequest<FeatureExtractionOutput>(args, {
+	const { data: res } = await innerRequest<FeatureExtractionOutput>(args, providerHelper, {
 		...options,
 		task: "feature-extraction",
 	});

package/src/tasks/nlp/fillMask.ts

@@ -10,7 +10,7 @@ export type FillMaskArgs = BaseArgs & FillMaskInput;
  */
 export async function fillMask(args: FillMaskArgs, options?: Options): Promise<FillMaskOutput> {
 	const providerHelper = getProviderHelper(args.provider ?? "hf-inference", "fill-mask");
-	const { data: res } = await innerRequest<FillMaskOutput>(args, {
+	const { data: res } = await innerRequest<FillMaskOutput>(args, providerHelper, {
 		...options,
 		task: "fill-mask",
 	});

package/src/tasks/nlp/questionAnswering.ts

@@ -13,9 +13,13 @@ export async function questionAnswering(
 	options?: Options
 ): Promise<QuestionAnsweringOutput[number]> {
 	const providerHelper = getProviderHelper(args.provider ?? "hf-inference", "question-answering");
-	const { data: res } = await innerRequest<QuestionAnsweringOutput | QuestionAnsweringOutput[number]>(args, {
-		...options,
-		task: "question-answering",
-	});
+	const { data: res } = await innerRequest<QuestionAnsweringOutput | QuestionAnsweringOutput[number]>(
+		args,
+		providerHelper,
+		{
+			...options,
+			task: "question-answering",
+		}
+	);
 	return providerHelper.getResponse(res);
 }

package/src/tasks/nlp/sentenceSimilarity.ts

@@ -13,7 +13,7 @@ export async function sentenceSimilarity(
 	options?: Options
 ): Promise<SentenceSimilarityOutput> {
 	const providerHelper = getProviderHelper(args.provider ?? "hf-inference", "sentence-similarity");
-	const { data: res } = await innerRequest<SentenceSimilarityOutput>(args, {
+	const { data: res } = await innerRequest<SentenceSimilarityOutput>(args, providerHelper, {
 		...options,
 		task: "sentence-similarity",
 	});

package/src/tasks/nlp/summarization.ts

@@ -10,7 +10,7 @@ export type SummarizationArgs = BaseArgs & SummarizationInput;
  */
 export async function summarization(args: SummarizationArgs, options?: Options): Promise<SummarizationOutput> {
 	const providerHelper = getProviderHelper(args.provider ?? "hf-inference", "summarization");
-	const { data: res } = await innerRequest<SummarizationOutput[]>(args, {
+	const { data: res } = await innerRequest<SummarizationOutput[]>(args, providerHelper, {
 		...options,
 		task: "summarization",
 	});

package/src/tasks/nlp/tableQuestionAnswering.ts

@@ -13,9 +13,13 @@ export async function tableQuestionAnswering(
 	options?: Options
 ): Promise<TableQuestionAnsweringOutput[number]> {
 	const providerHelper = getProviderHelper(args.provider ?? "hf-inference", "table-question-answering");
-	const { data: res } = await innerRequest<TableQuestionAnsweringOutput | TableQuestionAnsweringOutput[number]>(args, {
-		...options,
-		task: "table-question-answering",
-	});
+	const { data: res } = await innerRequest<TableQuestionAnsweringOutput | TableQuestionAnsweringOutput[number]>(
+		args,
+		providerHelper,
+		{
+			...options,
+			task: "table-question-answering",
+		}
+	);
 	return providerHelper.getResponse(res);
 }

package/src/tasks/nlp/textClassification.ts

@@ -13,7 +13,7 @@ export async function textClassification(
 	options?: Options
 ): Promise<TextClassificationOutput> {
 	const providerHelper = getProviderHelper(args.provider ?? "hf-inference", "text-classification");
-	const { data: res } = await innerRequest<TextClassificationOutput>(args, {
+	const { data: res } = await innerRequest<TextClassificationOutput>(args, providerHelper, {
 		...options,
 		task: "text-classification",
 	});

package/src/tasks/nlp/textGeneration.ts

@@ -13,11 +13,10 @@ export async function textGeneration(
 	args: BaseArgs & TextGenerationInput,
 	options?: Options
 ): Promise<TextGenerationOutput> {
-	const provider = args.provider ?? "hf-inference";
-	const providerHelper = getProviderHelper(provider, "text-generation");
+	const providerHelper = getProviderHelper(args.provider ?? "hf-inference", "text-generation");
 	const { data: response } = await innerRequest<
 		HyperbolicTextCompletionOutput | TextGenerationOutput | TextGenerationOutput[]
-	>(args, {
+	>(args, providerHelper, {
 		...options,
 		task: "text-generation",
 	});

package/src/tasks/nlp/textGenerationStream.ts

@@ -1,4 +1,5 @@
 import type { TextGenerationInput } from "@huggingface/tasks";
+import { getProviderHelper } from "../../lib/getProviderHelper";
 import type { BaseArgs, Options } from "../../types";
 import { innerStreamingRequest } from "../../utils/request";
 
@@ -89,7 +90,8 @@ export async function* textGenerationStream(
 	args: BaseArgs & TextGenerationInput,
 	options?: Options
 ): AsyncGenerator<TextGenerationStreamOutput> {
-	yield* innerStreamingRequest<TextGenerationStreamOutput>(args, {
+	const providerHelper = getProviderHelper(args.provider ?? "hf-inference", "text-generation");
+	yield* innerStreamingRequest<TextGenerationStreamOutput>(args, providerHelper, {
 		...options,
 		task: "text-generation",
 	});

package/src/tasks/nlp/tokenClassification.ts

@@ -2,7 +2,6 @@ import type { TokenClassificationInput, TokenClassificationOutput } from "@huggi
 import { getProviderHelper } from "../../lib/getProviderHelper";
 import type { BaseArgs, Options } from "../../types";
 import { innerRequest } from "../../utils/request";
-import { toArray } from "../../utils/toArray";
 
 export type TokenClassificationArgs = BaseArgs & TokenClassificationInput;
 
@@ -14,9 +13,13 @@ export async function tokenClassification(
 	options?: Options
 ): Promise<TokenClassificationOutput> {
 	const providerHelper = getProviderHelper(args.provider ?? "hf-inference", "token-classification");
-	const { data: res } = await innerRequest<TokenClassificationOutput[number] | TokenClassificationOutput>(args, {
-		...options,
-		task: "token-classification",
-	});
+	const { data: res } = await innerRequest<TokenClassificationOutput[number] | TokenClassificationOutput>(
+		args,
+		providerHelper,
+		{
+			...options,
+			task: "token-classification",
+		}
+	);
 	return providerHelper.getResponse(res);
 }