diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml
index f9111a4258..fe07e409fc 100644
--- a/.github/workflows/test.yml
+++ b/.github/workflows/test.yml
@@ -105,7 +105,7 @@ jobs:
         run: |
           sleep 3
           pnpm i --filter root --filter inference... --filter hub... --filter tasks-gen --frozen-lockfile
-          pnpm --filter inference --filter hub --filter tasks publish --force --no-git-checks --registry http://localhost:4874/
+          pnpm --filter inference --filter hub --filter tasks --filter jinja publish --force --no-git-checks --registry http://localhost:4874/
 
       - name: E2E test - test yarn install
         working-directory: e2e/ts
@@ -136,7 +136,7 @@ jobs:
           deno-version: vx.x.x
       - name: E2E test - deno import from npm
         working-directory: e2e/deno
-        run: deno run --allow-net --allow-env=HF_TOKEN index.ts
+        run: deno run --allow-read --allow-net --allow-env=HF_TOKEN index.ts
         env:
           NPM_CONFIG_REGISTRY: http://localhost:4874/
           HF_TOKEN: ${{ secrets.HF_TOKEN }}
diff --git a/.vscode/settings.json b/.vscode/settings.json
index 9d7878528a..e5c0fb1c32 100644
--- a/.vscode/settings.json
+++ b/.vscode/settings.json
@@ -17,5 +17,5 @@
 	"search.exclude": {
 		"**/dist": true
 	},
-	"typescript.tsdk": "node_modules/typescript/lib"
+	"typescript.tsdk": "node_modules/typescript/lib",
 }
diff --git a/packages/inference/package.json b/packages/inference/package.json
index 466a8093ef..dfe93e8d64 100644
--- a/packages/inference/package.json
+++ b/packages/inference/package.json
@@ -33,9 +33,16 @@
 	"main": "./dist/index.cjs",
 	"module": "./dist/index.js",
 	"exports": {
-		"types": "./dist/src/index.d.ts",
-		"require": "./dist/index.cjs",
-		"import": "./dist/index.js"
+		".": {
+			"types": "./dist/src/index.d.ts",
+			"require": "./dist/index.cjs",
+			"import": "./dist/index.js"
+		}
+	},
+	"browser": {
+		"./src/snippets/index.js": false,
+		"./dist/index.js": "./dist/browser/index.js",
+		"./dist/index.mjs": "./dist/browser/index.mjs"
 	},
 	"type": "module",
 	"scripts": {
@@ -52,7 +59,8 @@
 		"check": "tsc"
 	},
 	"dependencies": {
-		"@huggingface/tasks": "workspace:^"
+		"@huggingface/tasks": "workspace:^",
+		"@huggingface/jinja": "workspace:^"
 	},
 	"devDependencies": {
 		"@types/node": "18.13.0"
diff --git a/packages/inference/pnpm-lock.yaml b/packages/inference/pnpm-lock.yaml
index 5bb56da2d1..082c8ed752 100644
--- a/packages/inference/pnpm-lock.yaml
+++ b/packages/inference/pnpm-lock.yaml
@@ -5,6 +5,9 @@ settings:
   excludeLinksFromLockfile: false
 
 dependencies:
+  '@huggingface/jinja':
+    specifier: workspace:^
+    version: link:../jinja
   '@huggingface/tasks':
     specifier: workspace:^
     version: link:../tasks
diff --git a/packages/inference/src/index.ts b/packages/inference/src/index.ts
index b21526a5ea..aa16be8e63 100644
--- a/packages/inference/src/index.ts
+++ b/packages/inference/src/index.ts
@@ -2,6 +2,6 @@ export { InferenceClient, InferenceClientEndpoint, HfInference } from "./Inferen
 export { InferenceOutputError } from "./lib/InferenceOutputError";
 export * from "./types";
 export * from "./tasks";
-
 import * as snippets from "./snippets/index.js";
+
 export { snippets };
diff --git a/packages/inference/src/lib/makeRequestOptions.ts b/packages/inference/src/lib/makeRequestOptions.ts
index a6472d99db..061fce5637 100644
--- a/packages/inference/src/lib/makeRequestOptions.ts
+++ b/packages/inference/src/lib/makeRequestOptions.ts
@@ -56,13 +56,15 @@ export async function makeRequestOptions(
 		/** In most cases (unless we pass a endpointUrl) we know the task */
 		task?: InferenceTask;
 		chatCompletion?: boolean;
+		/* Used internally to generate inference snippets (in which case model mapping is done separately) */
+		skipModelIdResolution?: boolean;
 	}
 ): Promise<{ url: string; info: RequestInit }> {
 	const { accessToken, endpointUrl, provider: maybeProvider, model: maybeModel, ...remainingArgs } = args;
 	const provider = maybeProvider ?? "hf-inference";
 	const providerConfig = providerConfigs[provider];
 
-	const { includeCredentials, task, chatCompletion, signal } = options ?? {};
+	const { includeCredentials, task, chatCompletion, signal, skipModelIdResolution } = options ?? {};
 
 	if (endpointUrl && provider !== "hf-inference") {
 		throw new Error(`Cannot use endpointUrl with a third-party provider.`);
@@ -81,15 +83,17 @@ export async function makeRequestOptions(
 	}
 	// eslint-disable-next-line @typescript-eslint/no-non-null-assertion
 	const hfModel = maybeModel ?? (await loadDefaultModel(task!));
-	const model = providerConfig.clientSideRoutingOnly
-		? // eslint-disable-next-line @typescript-eslint/no-non-null-assertion
-		  removeProviderPrefix(maybeModel!, provider)
-		: // For closed-models API providers, one needs to pass the model ID directly (e.g. "gpt-3.5-turbo")
-		  await getProviderModelId({ model: hfModel, provider }, args, {
-				task,
-				chatCompletion,
-				fetch: options?.fetch,
-		  });
+	const model = skipModelIdResolution
+		? hfModel
+		: providerConfig.clientSideRoutingOnly
+		  ? // eslint-disable-next-line @typescript-eslint/no-non-null-assertion
+		    removeProviderPrefix(maybeModel!, provider)
+		  : // For closed-models API providers, one needs to pass the model ID directly (e.g. "gpt-3.5-turbo")
+		    await getProviderModelId({ model: hfModel, provider }, args, {
+					task,
+					chatCompletion,
+					fetch: options?.fetch,
+		    });
 
 	const authMethod = (() => {
 		if (providerConfig.clientSideRoutingOnly) {
diff --git a/packages/inference/src/snippets/python.ts b/packages/inference/src/snippets/python.ts
index ef3df2564f..f685f8b9ae 100644
--- a/packages/inference/src/snippets/python.ts
+++ b/packages/inference/src/snippets/python.ts
@@ -1,58 +1,137 @@
-import { openAIbaseUrl, type SnippetInferenceProvider } from "@huggingface/tasks";
 import type { PipelineType, WidgetType } from "@huggingface/tasks/src/pipelines.js";
 import type { ChatCompletionInputMessage, GenerationParameters } from "@huggingface/tasks/src/tasks/index.js";
 import {
+	openAIbaseUrl,
 	type InferenceSnippet,
 	type ModelDataMinimal,
 	getModelInputSnippet,
 	stringifyGenerationConfig,
 	stringifyMessages,
 } from "@huggingface/tasks";
+import type { InferenceProvider } from "../types";
+import { Template } from "@huggingface/jinja";
+import fs from "fs";
+import path from "path";
+import { existsSync as pathExists } from "node:fs";
+
+const TOOLS = ["huggingface_hub", "requests", "fal_client", "openai"];
+
+type InputPreparationFn = (model: ModelDataMinimal, opts?: Record<string, unknown>) => string | object;
+interface TemplateParams {
+	accessToken?: string;
+	baseUrl?: string;
+	inputs?: string | object;
+	model?: ModelDataMinimal;
+	provider?: InferenceProvider;
+	providerModelId?: string;
+	methodName?: string; // specific to snippetBasic
+	importBase64?: boolean; // specific to snippetImportRequests
+}
+
+// Helpers to find + load templates
+
+const rootDirFinder = (): string => {
+	let currentPath =
+		typeof import.meta !== "undefined" && import.meta.url
+			? path.normalize(new URL(import.meta.url).pathname) /// for ESM
+			: __dirname; /// for CJS
+
+	while (currentPath !== "/") {
+		if (pathExists(path.join(currentPath, "package.json"))) {
+			return currentPath;
+		}
+
+		currentPath = path.normalize(path.join(currentPath, ".."));
+	}
+
+	return "/";
+};
+
+const templatePath = (tool: string, templateName: string): string =>
+	path.join(rootDirFinder(), "src", "snippets", "templates", "python", tool, `${templateName}.jinja`);
+const hasTemplate = (tool: string, templateName: string): boolean => pathExists(templatePath(tool, templateName));
+
+const loadTemplate = (tool: string, templateName: string): ((data: TemplateParams) => string) => {
+	const template = fs.readFileSync(templatePath(tool, templateName), "utf8");
+	return (data: TemplateParams) => new Template(template).render({ ...data });
+};
+
+const snippetImportInferenceClient = loadTemplate("huggingface_hub", "importInferenceClient");
+const snippetImportRequests = loadTemplate("requests", "importRequests");
+
+// Needed for huggingface_hub basic snippets
 
 const HFH_INFERENCE_CLIENT_METHODS: Partial<Record<WidgetType, string>> = {
 	"audio-classification": "audio_classification",
 	"audio-to-audio": "audio_to_audio",
 	"automatic-speech-recognition": "automatic_speech_recognition",
-	"text-to-speech": "text_to_speech",
+	"document-question-answering": "document_question_answering",
+	"feature-extraction": "feature_extraction",
+	"fill-mask": "fill_mask",
 	"image-classification": "image_classification",
 	"image-segmentation": "image_segmentation",
 	"image-to-image": "image_to_image",
 	"image-to-text": "image_to_text",
 	"object-detection": "object_detection",
-	"text-to-image": "text_to_image",
-	"text-to-video": "text_to_video",
-	"zero-shot-image-classification": "zero_shot_image_classification",
-	"document-question-answering": "document_question_answering",
-	"visual-question-answering": "visual_question_answering",
-	"feature-extraction": "feature_extraction",
-	"fill-mask": "fill_mask",
 	"question-answering": "question_answering",
 	"sentence-similarity": "sentence_similarity",
 	summarization: "summarization",
 	"table-question-answering": "table_question_answering",
+	"tabular-classification": "tabular_classification",
+	"tabular-regression": "tabular_regression",
 	"text-classification": "text_classification",
 	"text-generation": "text_generation",
+	"text-to-image": "text_to_image",
+	"text-to-speech": "text_to_speech",
+	"text-to-video": "text_to_video",
 	"token-classification": "token_classification",
 	translation: "translation",
+	"visual-question-answering": "visual_question_answering",
 	"zero-shot-classification": "zero_shot_classification",
-	"tabular-classification": "tabular_classification",
-	"tabular-regression": "tabular_regression",
+	"zero-shot-image-classification": "zero_shot_image_classification",
 };
 
-const snippetImportInferenceClient = (accessToken: string, provider: SnippetInferenceProvider): string =>
-	`\
-from huggingface_hub import InferenceClient
-
-client = InferenceClient(
-    provider="${provider}",
-    api_key="${accessToken || "{API_TOKEN}"}",
-)`;
+// Snippet generators
+const snippetGenerator = (templateName: string, inputPreparationFn?: InputPreparationFn) => {
+	return (
+		model: ModelDataMinimal,
+		accessToken: string,
+		provider: InferenceProvider,
+		providerModelId?: string,
+		opts?: Record<string, unknown>
+	): InferenceSnippet[] => {
+		const params: TemplateParams = {
+			accessToken,
+			baseUrl: templateName.includes("conversational") ? openAIbaseUrl(provider) : undefined,
+			inputs: inputPreparationFn ? inputPreparationFn(model, opts) : getModelInputSnippet(model),
+			model,
+			provider,
+			providerModelId: providerModelId ?? model.id,
+		};
+
+		// Iterate over tools => check if a snippet exists => generate
+		return TOOLS.map((tool) => {
+			if (!hasTemplate(tool, templateName)) {
+				return;
+			}
+			const template = loadTemplate(tool, templateName);
+			if (tool === "huggingface_hub" && templateName === "basic") {
+				if (!(model.pipeline_tag && model.pipeline_tag in HFH_INFERENCE_CLIENT_METHODS)) {
+					return;
+				}
+				params["methodName"] = HFH_INFERENCE_CLIENT_METHODS[model.pipeline_tag];
+			}
+			return {
+				client: tool,
+				content: template(params).trim(),
+			};
+		}).filter((snippet) => snippet !== undefined && snippet.content) as InferenceSnippet[];
+	};
+};
 
-const snippetConversational = (
+// Input preparation functions (required for a few tasks)
+const prepareConversationalInput = (
 	model: ModelDataMinimal,
-	accessToken: string,
-	provider: SnippetInferenceProvider,
-	providerModelId?: string,
 	opts?: {
 		streaming?: boolean;
 		messages?: ChatCompletionInputMessage[];
@@ -60,421 +139,53 @@ const snippetConversational = (
 		max_tokens?: GenerationParameters["max_tokens"];
 		top_p?: GenerationParameters["top_p"];
 	}
-): InferenceSnippet[] => {
-	const streaming = opts?.streaming ?? true;
+): object => {
 	const exampleMessages = getModelInputSnippet(model) as ChatCompletionInputMessage[];
 	const messages = opts?.messages ?? exampleMessages;
-	const messagesStr = stringifyMessages(messages, { attributeKeyQuotes: true });
-
 	const config = {
-		...(opts?.temperature ? { temperature: opts.temperature } : undefined),
+		...(opts?.temperature ? { temperature: opts?.temperature } : undefined),
 		max_tokens: opts?.max_tokens ?? 500,
-		...(opts?.top_p ? { top_p: opts.top_p } : undefined),
+		...(opts?.top_p ? { top_p: opts?.top_p } : undefined),
 	};
-	const configStr = stringifyGenerationConfig(config, {
-		indent: "\n\t",
-		attributeValueConnector: "=",
-	});
-
-	if (streaming) {
-		return [
-			{
-				client: "huggingface_hub",
-				content: `\
-${snippetImportInferenceClient(accessToken, provider)}
-
-messages = ${messagesStr}
-
-stream = client.chat.completions.create(
-	model="${model.id}", 
-	messages=messages, 
-	${configStr}
-	stream=True,
-)
-
-for chunk in stream:
-    print(chunk.choices[0].delta.content, end="")`,
-			},
-			{
-				client: "openai",
-				content: `\
-from openai import OpenAI
-
-client = OpenAI(
-	base_url="${openAIbaseUrl(provider)}",
-	api_key="${accessToken || "{API_TOKEN}"}"
-)
-
-messages = ${messagesStr}
-
-stream = client.chat.completions.create(
-    model="${providerModelId ?? model.id}", 
-	messages=messages, 
-	${configStr}
-	stream=True
-)
-
-for chunk in stream:
-	print(chunk.choices[0].delta.content, end="")`,
-			},
-		];
-	} else {
-		return [
-			{
-				client: "huggingface_hub",
-				content: `\
-${snippetImportInferenceClient(accessToken, provider)}
-
-messages = ${messagesStr}
-
-completion = client.chat.completions.create(
-    model="${model.id}", 
-	messages=messages, 
-	${configStr}
-)
-
-print(completion.choices[0].message)`,
-			},
-			{
-				client: "openai",
-				content: `\
-from openai import OpenAI
-
-client = OpenAI(
-	base_url="${openAIbaseUrl(provider)}",
-	api_key="${accessToken || "{API_TOKEN}"}"
-)
-
-messages = ${messagesStr}
-
-completion = client.chat.completions.create(
-	model="${providerModelId ?? model.id}", 
-	messages=messages, 
-	${configStr}
-)
-
-print(completion.choices[0].message)`,
-			},
-		];
-	}
-};
-
-const snippetZeroShotClassification = (model: ModelDataMinimal): InferenceSnippet[] => {
-	return [
-		{
-			client: "requests",
-			content: `\
-def query(payload):
-	response = requests.post(API_URL, headers=headers, json=payload)
-	return response.json()
-
-output = query({
-    "inputs": ${getModelInputSnippet(model)},
-    "parameters": {"candidate_labels": ["refund", "legal", "faq"]},
-})`,
-		},
-	];
-};
-
-const snippetZeroShotImageClassification = (model: ModelDataMinimal): InferenceSnippet[] => {
-	return [
-		{
-			client: "requests",
-			content: `def query(data):
-	with open(data["image_path"], "rb") as f:
-		img = f.read()
-	payload={
-		"parameters": data["parameters"],
-		"inputs": base64.b64encode(img).decode("utf-8")
-	}
-	response = requests.post(API_URL, headers=headers, json=payload)
-	return response.json()
-
-output = query({
-	"image_path": ${getModelInputSnippet(model)},
-	"parameters": {"candidate_labels": ["cat", "dog", "llama"]},
-})`,
-		},
-	];
-};
 
-const snippetBasic = (
-	model: ModelDataMinimal,
-	accessToken: string,
-	provider: SnippetInferenceProvider
-): InferenceSnippet[] => {
-	return [
-		...(model.pipeline_tag && model.pipeline_tag in HFH_INFERENCE_CLIENT_METHODS
-			? [
-					{
-						client: "huggingface_hub",
-						content: `\
-${snippetImportInferenceClient(accessToken, provider)}
-
-result = client.${HFH_INFERENCE_CLIENT_METHODS[model.pipeline_tag]}(
-	inputs=${getModelInputSnippet(model)},
-	model="${model.id}",
-)
-
-print(result)
-`,
-					},
-			  ]
-			: []),
-		{
-			client: "requests",
-			content: `\
-def query(payload):
-	response = requests.post(API_URL, headers=headers, json=payload)
-	return response.json()
-	
-output = query({
-	"inputs": ${getModelInputSnippet(model)},
-})`,
-		},
-	];
-};
-
-const snippetFile = (model: ModelDataMinimal): InferenceSnippet[] => {
-	return [
-		{
-			client: "requests",
-			content: `\
-def query(filename):
-	with open(filename, "rb") as f:
-		data = f.read()
-	response = requests.post(API_URL, headers=headers, data=data)
-	return response.json()
-
-output = query(${getModelInputSnippet(model)})`,
-		},
-	];
-};
-
-const snippetTextToImage = (
-	model: ModelDataMinimal,
-	accessToken: string,
-	provider: SnippetInferenceProvider,
-	providerModelId?: string
-): InferenceSnippet[] => {
-	return [
-		{
-			client: "huggingface_hub",
-			content: `\
-${snippetImportInferenceClient(accessToken, provider)}
-
-# output is a PIL.Image object
-image = client.text_to_image(
-	${getModelInputSnippet(model)},
-	model="${model.id}",
-)`,
-		},
-		...(provider === "fal-ai"
-			? [
-					{
-						client: "fal-client",
-						content: `\
-import fal_client
-
-result = fal_client.subscribe(
-	"${providerModelId ?? model.id}",
-	arguments={
-		"prompt": ${getModelInputSnippet(model)},
-	},
-)
-print(result)
-`,
-					},
-			  ]
-			: []),
-		...(provider === "hf-inference"
-			? [
-					{
-						client: "requests",
-						content: `\
-def query(payload):
-	response = requests.post(API_URL, headers=headers, json=payload)
-	return response.content
-
-image_bytes = query({
-	"inputs": ${getModelInputSnippet(model)},
-})
-
-# You can access the image with PIL.Image for example
-import io
-from PIL import Image
-image = Image.open(io.BytesIO(image_bytes))`,
-					},
-			  ]
-			: []),
-	];
-};
-
-const snippetTextToVideo = (
-	model: ModelDataMinimal,
-	accessToken: string,
-	provider: SnippetInferenceProvider
-): InferenceSnippet[] => {
-	return ["fal-ai", "replicate"].includes(provider)
-		? [
-				{
-					client: "huggingface_hub",
-					content: `\
-${snippetImportInferenceClient(accessToken, provider)}
-
-video = client.text_to_video(
-	${getModelInputSnippet(model)},
-	model="${model.id}",
-)`,
-				},
-		  ]
-		: [];
-};
-
-const snippetTabular = (model: ModelDataMinimal): InferenceSnippet[] => {
-	return [
-		{
-			client: "requests",
-			content: `\
-def query(payload):
-	response = requests.post(API_URL, headers=headers, json=payload)
-	return response.content
-
-response = query({
-	"inputs": {"data": ${getModelInputSnippet(model)}},
-})`,
-		},
-	];
-};
-
-const snippetTextToAudio = (model: ModelDataMinimal): InferenceSnippet[] => {
-	// Transformers TTS pipeline and api-inference-community (AIC) pipeline outputs are diverged
-	// with the latest update to inference-api (IA).
-	// Transformers IA returns a byte object (wav file), whereas AIC returns wav and sampling_rate.
-	if (model.library_name === "transformers") {
-		return [
-			{
-				client: "requests",
-				content: `\
-def query(payload):
-	response = requests.post(API_URL, headers=headers, json=payload)
-	return response.content
-
-audio_bytes = query({
-	"inputs": ${getModelInputSnippet(model)},
-})
-# You can access the audio with IPython.display for example
-from IPython.display import Audio
-Audio(audio_bytes)`,
-			},
-		];
-	} else {
-		return [
-			{
-				client: "requests",
-				content: `def query(payload):
-	response = requests.post(API_URL, headers=headers, json=payload)
-	return response.json()
-	
-audio, sampling_rate = query({
-	"inputs": ${getModelInputSnippet(model)},
-})
-# You can access the audio with IPython.display for example
-from IPython.display import Audio
-Audio(audio, rate=sampling_rate)`,
-			},
-		];
-	}
-};
-
-const snippetAutomaticSpeechRecognition = (
-	model: ModelDataMinimal,
-	accessToken: string,
-	provider: SnippetInferenceProvider
-): InferenceSnippet[] => {
-	return [
-		{
-			client: "huggingface_hub",
-			content: `${snippetImportInferenceClient(accessToken, provider)}
-output = client.automatic_speech_recognition(${getModelInputSnippet(model)}, model="${model.id}")`,
-		},
-		snippetFile(model)[0],
-	];
+	return {
+		messagesStr: stringifyMessages(messages, { attributeKeyQuotes: true }),
+		configStr: stringifyGenerationConfig(config, {
+			indent: "\n\t",
+			attributeValueConnector: "=",
+		}),
+	};
 };
 
-const snippetDocumentQuestionAnswering = (
-	model: ModelDataMinimal,
-	accessToken: string,
-	provider: SnippetInferenceProvider
-): InferenceSnippet[] => {
+const prepareDocumentQuestionAnsweringInput = (model: ModelDataMinimal): object => {
 	const inputsAsStr = getModelInputSnippet(model) as string;
 	const inputsAsObj = JSON.parse(inputsAsStr);
+	return { asObj: inputsAsObj, asStr: inputsAsStr };
+};
 
-	return [
-		{
-			client: "huggingface_hub",
-			content: `${snippetImportInferenceClient(accessToken, provider)}
-output = client.document_question_answering(
-    "${inputsAsObj.image}",
-	question="${inputsAsObj.question}",
-	model="${model.id}",
-)`,
-		},
-		{
-			client: "requests",
-			content: `def query(payload):
-	with open(payload["image"], "rb") as f:
-		img = f.read()
-		payload["image"] = base64.b64encode(img).decode("utf-8")
-	response = requests.post(API_URL, headers=headers, json=payload)
-	return response.json()
-
-output = query({
-    "inputs": ${inputsAsStr},
-})`,
-		},
-	];
+const prepareImageToImageInput = (model: ModelDataMinimal): object => {
+	return JSON.parse(getModelInputSnippet(model) as string);
 };
 
-const snippetImageToImage = (
+const snippetConversational = (
 	model: ModelDataMinimal,
 	accessToken: string,
-	provider: SnippetInferenceProvider
+	provider: InferenceProvider,
+	providerModelId?: string,
+	opts?: {
+		streaming?: boolean;
+		messages?: ChatCompletionInputMessage[];
+		temperature?: GenerationParameters["temperature"];
+		max_tokens?: GenerationParameters["max_tokens"];
+		top_p?: GenerationParameters["top_p"];
+	}
 ): InferenceSnippet[] => {
-	const inputsAsStr = getModelInputSnippet(model) as string;
-	const inputsAsObj = JSON.parse(inputsAsStr);
-
-	return [
-		{
-			client: "huggingface_hub",
-			content: `${snippetImportInferenceClient(accessToken, provider)}
-# output is a PIL.Image object
-image = client.image_to_image(
-    "${inputsAsObj.image}",
-    prompt="${inputsAsObj.prompt}",
-    model="${model.id}",
-)`,
-		},
-		{
-			client: "requests",
-			content: `def query(payload):
-	with open(payload["inputs"], "rb") as f:
-		img = f.read()
-		payload["inputs"] = base64.b64encode(img).decode("utf-8")
-	response = requests.post(API_URL, headers=headers, json=payload)
-	return response.content
-
-image_bytes = query({
-	"inputs": "${inputsAsObj.image}",
-	"parameters": {"prompt": "${inputsAsObj.prompt}"},
-})
-
-# You can access the image with PIL.Image for example
-import io
-from PIL import Image
-image = Image.open(io.BytesIO(image_bytes))`,
-		},
-	];
+	const streaming = opts?.streaming ?? true;
+	const templateName = streaming ? "conversationalStream" : "conversational";
+	return snippetGenerator(templateName, prepareConversationalInput)(model, accessToken, provider, providerModelId, {
+		...opts,
+		streaming,
+	});
 };
 
 const pythonSnippets: Partial<
@@ -483,81 +194,77 @@ const pythonSnippets: Partial<
 		(
 			model: ModelDataMinimal,
 			accessToken: string,
-			provider: SnippetInferenceProvider,
+			provider: InferenceProvider,
 			providerModelId?: string,
 			opts?: Record<string, unknown>
 		) => InferenceSnippet[]
 	>
 > = {
-	// Same order as in tasks/src/pipelines.ts
-	"text-classification": snippetBasic,
-	"token-classification": snippetBasic,
-	"table-question-answering": snippetBasic,
-	"question-answering": snippetBasic,
-	"zero-shot-classification": snippetZeroShotClassification,
-	translation: snippetBasic,
-	summarization: snippetBasic,
-	"feature-extraction": snippetBasic,
-	"text-generation": snippetBasic,
-	"text2text-generation": snippetBasic,
+	"audio-classification": snippetGenerator("basicFile"),
+	"audio-to-audio": snippetGenerator("basicFile"),
+	"automatic-speech-recognition": snippetGenerator("automaticSpeechRecognition"),
+	"document-question-answering": snippetGenerator("documentQuestionAnswering", prepareDocumentQuestionAnsweringInput),
+	"feature-extraction": snippetGenerator("basic"),
+	"fill-mask": snippetGenerator("basic"),
+	"image-classification": snippetGenerator("basicFile"),
+	"image-segmentation": snippetGenerator("basicFile"),
 	"image-text-to-text": snippetConversational,
-	"fill-mask": snippetBasic,
-	"sentence-similarity": snippetBasic,
-	"automatic-speech-recognition": snippetAutomaticSpeechRecognition,
-	"text-to-image": snippetTextToImage,
-	"text-to-video": snippetTextToVideo,
-	"text-to-speech": snippetTextToAudio,
-	"text-to-audio": snippetTextToAudio,
-	"audio-to-audio": snippetFile,
-	"audio-classification": snippetFile,
-	"image-classification": snippetFile,
-	"tabular-regression": snippetTabular,
-	"tabular-classification": snippetTabular,
-	"object-detection": snippetFile,
-	"image-segmentation": snippetFile,
-	"document-question-answering": snippetDocumentQuestionAnswering,
-	"image-to-text": snippetFile,
-	"image-to-image": snippetImageToImage,
-	"zero-shot-image-classification": snippetZeroShotImageClassification,
+	"image-to-image": snippetGenerator("imageToImage", prepareImageToImageInput),
+	"image-to-text": snippetGenerator("basicFile"),
+	"object-detection": snippetGenerator("basicFile"),
+	"question-answering": snippetGenerator("basic"),
+	"sentence-similarity": snippetGenerator("basic"),
+	summarization: snippetGenerator("basic"),
+	"tabular-classification": snippetGenerator("tabular"),
+	"tabular-regression": snippetGenerator("tabular"),
+	"table-question-answering": snippetGenerator("basic"),
+	"text-classification": snippetGenerator("basic"),
+	"text-generation": snippetGenerator("basic"),
+	"text-to-audio": snippetGenerator("textToAudio"),
+	"text-to-image": snippetGenerator("textToImage"),
+	"text-to-speech": snippetGenerator("textToAudio"),
+	"text-to-video": snippetGenerator("textToVideo"),
+	"text2text-generation": snippetGenerator("basic"),
+	"token-classification": snippetGenerator("basic"),
+	translation: snippetGenerator("basic"),
+	"zero-shot-classification": snippetGenerator("zeroShotClassification"),
+	"zero-shot-image-classification": snippetGenerator("zeroShotImageClassification"),
 };
-
 export function getPythonInferenceSnippet(
 	model: ModelDataMinimal,
 	accessToken: string,
-	provider: SnippetInferenceProvider,
+	provider: InferenceProvider,
 	providerModelId?: string,
 	opts?: Record<string, unknown>
 ): InferenceSnippet[] {
-	if (model.tags.includes("conversational")) {
-		// Conversational model detected, so we display a code snippet that features the Messages API
-		return snippetConversational(model, accessToken, provider, providerModelId, opts);
-	} else {
-		const snippets =
-			model.pipeline_tag && model.pipeline_tag in pythonSnippets
-				? pythonSnippets[model.pipeline_tag]?.(model, accessToken, provider, providerModelId) ?? []
-				: [];
+	const snippets = model.tags.includes("conversational")
+		? snippetConversational(model, accessToken, provider, providerModelId, opts)
+		: model.pipeline_tag && model.pipeline_tag in pythonSnippets
+		  ? pythonSnippets[model.pipeline_tag]?.(model, accessToken, provider, providerModelId) ?? []
+		  : [];
 
-		return snippets.map((snippet) => {
-			return {
-				...snippet,
-				content: addImportsToSnippet(snippet.content, model, accessToken),
-			};
-		});
-	}
+	return snippets.map((snippet) => addImportsToSnippet(snippet, model, accessToken, provider));
 }
 
-const addImportsToSnippet = (snippet: string, model: ModelDataMinimal, accessToken: string): string => {
-	if (snippet.includes("requests")) {
-		snippet = `import requests
-
-API_URL = "https://router.huggingface.co/hf-inference/models/${model.id}"
-headers = {"Authorization": ${accessToken ? `"Bearer ${accessToken}"` : `f"Bearer {API_TOKEN}"`}}
-
-${snippet}`;
-	}
-	if (snippet.includes("base64")) {
-		snippet = `import base64
-${snippet}`;
+const addImportsToSnippet = (
+	snippet: InferenceSnippet,
+	model: ModelDataMinimal,
+	accessToken: string,
+	provider: InferenceProvider
+): InferenceSnippet => {
+	let importSection: string | undefined = undefined;
+	if (snippet.client === "huggingface_hub") {
+		importSection = snippetImportInferenceClient({ accessToken, provider });
+	} else if (snippet.content.includes("requests")) {
+		importSection = snippetImportRequests({
+			accessToken,
+			model: model,
+			provider,
+			importBase64: snippet.content.includes("base64"),
+		});
 	}
-	return snippet;
+	return {
+		...snippet,
+		content: importSection ? `${importSection}\n\n${snippet.content}` : snippet.content,
+	};
 };
diff --git a/packages/inference/src/snippets/templates/python/fal_client/textToImage.jinja b/packages/inference/src/snippets/templates/python/fal_client/textToImage.jinja
new file mode 100644
index 0000000000..db0c94dd09
--- /dev/null
+++ b/packages/inference/src/snippets/templates/python/fal_client/textToImage.jinja
@@ -0,0 +1,11 @@
+{% if provider == "fal-ai" %}
+import fal_client
+
+result = fal_client.subscribe(
+    "{{ providerModelId }}",
+    arguments={
+        "prompt": {{ inputs }},
+    },
+)
+print(result)
+{% endif %} 
\ No newline at end of file
diff --git a/packages/inference/src/snippets/templates/python/huggingface_hub/automaticSpeechRecognition.jinja b/packages/inference/src/snippets/templates/python/huggingface_hub/automaticSpeechRecognition.jinja
new file mode 100644
index 0000000000..2a4f0f7769
--- /dev/null
+++ b/packages/inference/src/snippets/templates/python/huggingface_hub/automaticSpeechRecognition.jinja
@@ -0,0 +1 @@
+output = client.automatic_speech_recognition({{ inputs }}, model="{{ model.id }}") 
\ No newline at end of file
diff --git a/packages/inference/src/snippets/templates/python/huggingface_hub/basic.jinja b/packages/inference/src/snippets/templates/python/huggingface_hub/basic.jinja
new file mode 100644
index 0000000000..f2d22d6778
--- /dev/null
+++ b/packages/inference/src/snippets/templates/python/huggingface_hub/basic.jinja
@@ -0,0 +1,6 @@
+result = client.{{ methodName }}(
+    inputs={{ inputs }},
+    model="{{ model.id }}",
+)
+
+print(result) 
\ No newline at end of file
diff --git a/packages/inference/src/snippets/templates/python/huggingface_hub/conversational.jinja b/packages/inference/src/snippets/templates/python/huggingface_hub/conversational.jinja
new file mode 100644
index 0000000000..a3ee06e8a5
--- /dev/null
+++ b/packages/inference/src/snippets/templates/python/huggingface_hub/conversational.jinja
@@ -0,0 +1,9 @@
+messages = {{ inputs.messagesStr }}
+
+completion = client.chat.completions.create(
+    model="{{ model.id }}",
+    messages=messages,
+    {{ inputs.configStr }}
+)
+
+print(completion.choices[0].message) 
\ No newline at end of file
diff --git a/packages/inference/src/snippets/templates/python/huggingface_hub/conversationalStream.jinja b/packages/inference/src/snippets/templates/python/huggingface_hub/conversationalStream.jinja
new file mode 100644
index 0000000000..013e513c8a
--- /dev/null
+++ b/packages/inference/src/snippets/templates/python/huggingface_hub/conversationalStream.jinja
@@ -0,0 +1,11 @@
+messages = {{ inputs.messagesStr }}
+
+stream = client.chat.completions.create(
+    model="{{ model.id }}",
+    messages=messages,
+    {{ inputs.configStr }}
+    stream=True,
+)
+
+for chunk in stream:
+    print(chunk.choices[0].delta.content, end="") 
\ No newline at end of file
diff --git a/packages/inference/src/snippets/templates/python/huggingface_hub/documentQuestionAnswering.jinja b/packages/inference/src/snippets/templates/python/huggingface_hub/documentQuestionAnswering.jinja
new file mode 100644
index 0000000000..7643cfed0d
--- /dev/null
+++ b/packages/inference/src/snippets/templates/python/huggingface_hub/documentQuestionAnswering.jinja
@@ -0,0 +1,5 @@
+output = client.document_question_answering(
+    "{{ inputs.asObj.image }}",
+    question="{{ inputs.asObj.question }}",
+    model="{{ model.id }}",
+) 
\ No newline at end of file
diff --git a/packages/inference/src/snippets/templates/python/huggingface_hub/imageToImage.jinja b/packages/inference/src/snippets/templates/python/huggingface_hub/imageToImage.jinja
new file mode 100644
index 0000000000..001d807daf
--- /dev/null
+++ b/packages/inference/src/snippets/templates/python/huggingface_hub/imageToImage.jinja
@@ -0,0 +1,6 @@
+# output is a PIL.Image object
+image = client.image_to_image(
+    "{{ inputs.image }}",
+    prompt="{{ inputs.prompt }}",
+    model="{{ model.id }}",
+) 
\ No newline at end of file
diff --git a/packages/inference/src/snippets/templates/python/huggingface_hub/importInferenceClient.jinja b/packages/inference/src/snippets/templates/python/huggingface_hub/importInferenceClient.jinja
new file mode 100644
index 0000000000..9a6514f267
--- /dev/null
+++ b/packages/inference/src/snippets/templates/python/huggingface_hub/importInferenceClient.jinja
@@ -0,0 +1,6 @@
+from huggingface_hub import InferenceClient
+
+client = InferenceClient(
+    provider="{{ provider }}",
+    api_key="{{ accessToken }}",
+)
\ No newline at end of file
diff --git a/packages/inference/src/snippets/templates/python/huggingface_hub/textToImage.jinja b/packages/inference/src/snippets/templates/python/huggingface_hub/textToImage.jinja
new file mode 100644
index 0000000000..5379d68b11
--- /dev/null
+++ b/packages/inference/src/snippets/templates/python/huggingface_hub/textToImage.jinja
@@ -0,0 +1,5 @@
+# output is a PIL.Image object
+image = client.text_to_image(
+    {{ inputs }},
+    model="{{ model.id }}",
+) 
\ No newline at end of file
diff --git a/packages/inference/src/snippets/templates/python/huggingface_hub/textToVideo.jinja b/packages/inference/src/snippets/templates/python/huggingface_hub/textToVideo.jinja
new file mode 100644
index 0000000000..f1d5e8d8a8
--- /dev/null
+++ b/packages/inference/src/snippets/templates/python/huggingface_hub/textToVideo.jinja
@@ -0,0 +1,4 @@
+video = client.text_to_video(
+    {{ inputs }},
+    model="{{ model.id }}",
+) 
\ No newline at end of file
diff --git a/packages/inference/src/snippets/templates/python/openai/conversational.jinja b/packages/inference/src/snippets/templates/python/openai/conversational.jinja
new file mode 100644
index 0000000000..4d827f1788
--- /dev/null
+++ b/packages/inference/src/snippets/templates/python/openai/conversational.jinja
@@ -0,0 +1,16 @@
+from openai import OpenAI
+
+client = OpenAI(
+    base_url="{{ baseUrl }}",
+    api_key="{{ accessToken }}"
+)
+
+messages = {{ inputs.messagesStr }}
+
+completion = client.chat.completions.create(
+    model="{{ providerModelId }}",
+    messages=messages,
+    {{ inputs.configStr }}
+)
+
+print(completion.choices[0].message) 
\ No newline at end of file
diff --git a/packages/inference/src/snippets/templates/python/openai/conversationalStream.jinja b/packages/inference/src/snippets/templates/python/openai/conversationalStream.jinja
new file mode 100644
index 0000000000..8c71bd745a
--- /dev/null
+++ b/packages/inference/src/snippets/templates/python/openai/conversationalStream.jinja
@@ -0,0 +1,18 @@
+from openai import OpenAI
+
+client = OpenAI(
+    base_url="{{ baseUrl }}",
+    api_key="{{ accessToken }}"
+)
+
+messages = {{ inputs.messagesStr }}
+
+stream = client.chat.completions.create(
+    model="{{ providerModelId }}",
+    messages=messages,
+    {{ inputs.configStr }}
+    stream=True
+)
+
+for chunk in stream:
+    print(chunk.choices[0].delta.content, end="") 
\ No newline at end of file
diff --git a/packages/inference/src/snippets/templates/python/requests/automaticSpeechRecognition.jinja b/packages/inference/src/snippets/templates/python/requests/automaticSpeechRecognition.jinja
new file mode 100644
index 0000000000..e73c2ce79a
--- /dev/null
+++ b/packages/inference/src/snippets/templates/python/requests/automaticSpeechRecognition.jinja
@@ -0,0 +1,7 @@
+def query(filename):
+    with open(filename, "rb") as f:
+        data = f.read()
+    response = requests.post(API_URL, headers=headers, data=data)
+    return response.json()
+
+output = query({{ inputs }}) 
\ No newline at end of file
diff --git a/packages/inference/src/snippets/templates/python/requests/basic.jinja b/packages/inference/src/snippets/templates/python/requests/basic.jinja
new file mode 100644
index 0000000000..489218820d
--- /dev/null
+++ b/packages/inference/src/snippets/templates/python/requests/basic.jinja
@@ -0,0 +1,7 @@
+def query(payload):
+    response = requests.post(API_URL, headers=headers, json=payload)
+    return response.json()
+
+output = query({
+    "inputs": {{ inputs }},
+}) 
\ No newline at end of file
diff --git a/packages/inference/src/snippets/templates/python/requests/basicFile.jinja b/packages/inference/src/snippets/templates/python/requests/basicFile.jinja
new file mode 100644
index 0000000000..e73c2ce79a
--- /dev/null
+++ b/packages/inference/src/snippets/templates/python/requests/basicFile.jinja
@@ -0,0 +1,7 @@
+def query(filename):
+    with open(filename, "rb") as f:
+        data = f.read()
+    response = requests.post(API_URL, headers=headers, data=data)
+    return response.json()
+
+output = query({{ inputs }}) 
\ No newline at end of file
diff --git a/packages/inference/src/snippets/templates/python/requests/documentQuestionAnswering.jinja b/packages/inference/src/snippets/templates/python/requests/documentQuestionAnswering.jinja
new file mode 100644
index 0000000000..76072952f5
--- /dev/null
+++ b/packages/inference/src/snippets/templates/python/requests/documentQuestionAnswering.jinja
@@ -0,0 +1,10 @@
+def query(payload):
+    with open(payload["image"], "rb") as f:
+        img = f.read()
+        payload["image"] = base64.b64encode(img).decode("utf-8")
+    response = requests.post(API_URL, headers=headers, json=payload)
+    return response.json()
+
+output = query({
+    "inputs": {{ inputs.asStr }},
+}) 
\ No newline at end of file
diff --git a/packages/inference/src/snippets/templates/python/requests/imageToImage.jinja b/packages/inference/src/snippets/templates/python/requests/imageToImage.jinja
new file mode 100644
index 0000000000..4c05cc307a
--- /dev/null
+++ b/packages/inference/src/snippets/templates/python/requests/imageToImage.jinja
@@ -0,0 +1,16 @@
+def query(payload):
+    with open(payload["inputs"], "rb") as f:
+        img = f.read()
+        payload["inputs"] = base64.b64encode(img).decode("utf-8")
+    response = requests.post(API_URL, headers=headers, json=payload)
+    return response.content
+
+image_bytes = query({
+    "inputs": "{{ inputs.image }}",
+    "parameters": {"prompt": "{{ inputs.prompt }}"},
+})
+
+# You can access the image with PIL.Image for example
+import io
+from PIL import Image
+image = Image.open(io.BytesIO(image_bytes)) 
\ No newline at end of file
diff --git a/packages/inference/src/snippets/templates/python/requests/importRequests.jinja b/packages/inference/src/snippets/templates/python/requests/importRequests.jinja
new file mode 100644
index 0000000000..f0486a32d4
--- /dev/null
+++ b/packages/inference/src/snippets/templates/python/requests/importRequests.jinja
@@ -0,0 +1,7 @@
+{% if importBase64 %}
+import base64
+{% endif %}
+import requests
+
+API_URL = "https://router.huggingface.co/{{ provider }}/models/{{ model.id }}"
+headers = {"Authorization": "Bearer {{ accessToken }}"}
\ No newline at end of file
diff --git a/packages/inference/src/snippets/templates/python/requests/tabular.jinja b/packages/inference/src/snippets/templates/python/requests/tabular.jinja
new file mode 100644
index 0000000000..76914d4c5e
--- /dev/null
+++ b/packages/inference/src/snippets/templates/python/requests/tabular.jinja
@@ -0,0 +1,9 @@
+def query(payload):
+    response = requests.post(API_URL, headers=headers, json=payload)
+    return response.content
+
+response = query({
+    "inputs": {
+        "data": {{ inputs }}
+    },
+}) 
\ No newline at end of file
diff --git a/packages/inference/src/snippets/templates/python/requests/textToAudio.jinja b/packages/inference/src/snippets/templates/python/requests/textToAudio.jinja
new file mode 100644
index 0000000000..ecb2cf23ad
--- /dev/null
+++ b/packages/inference/src/snippets/templates/python/requests/textToAudio.jinja
@@ -0,0 +1,23 @@
+{% if model.library_name == "transformers" %}
+def query(payload):
+    response = requests.post(API_URL, headers=headers, json=payload)
+    return response.content
+
+audio_bytes = query({
+    "inputs": {{ inputs }},
+})
+# You can access the audio with IPython.display for example
+from IPython.display import Audio
+Audio(audio_bytes)
+{% else %}
+def query(payload):
+    response = requests.post(API_URL, headers=headers, json=payload)
+    return response.json()
+
+audio, sampling_rate = query({
+    "inputs": {{ inputs }},
+})
+# You can access the audio with IPython.display for example
+from IPython.display import Audio
+Audio(audio, rate=sampling_rate)
+{% endif %} 
\ No newline at end of file
diff --git a/packages/inference/src/snippets/templates/python/requests/textToImage.jinja b/packages/inference/src/snippets/templates/python/requests/textToImage.jinja
new file mode 100644
index 0000000000..1c30ac461c
--- /dev/null
+++ b/packages/inference/src/snippets/templates/python/requests/textToImage.jinja
@@ -0,0 +1,14 @@
+{% if provider == "hf-inference" %}
+def query(payload):
+    response = requests.post(API_URL, headers=headers, json=payload)
+    return response.content
+
+image_bytes = query({
+    "inputs": {{ inputs }},
+})
+
+# You can access the image with PIL.Image for example
+import io
+from PIL import Image
+image = Image.open(io.BytesIO(image_bytes))
+{% endif %} 
\ No newline at end of file
diff --git a/packages/inference/src/snippets/templates/python/requests/zeroShotClassification.jinja b/packages/inference/src/snippets/templates/python/requests/zeroShotClassification.jinja
new file mode 100644
index 0000000000..664ec00b5f
--- /dev/null
+++ b/packages/inference/src/snippets/templates/python/requests/zeroShotClassification.jinja
@@ -0,0 +1,8 @@
+def query(payload):
+    response = requests.post(API_URL, headers=headers, json=payload)
+    return response.json()
+
+output = query({
+    "inputs": {{ inputs }},
+    "parameters": {"candidate_labels": ["refund", "legal", "faq"]},
+}) 
\ No newline at end of file
diff --git a/packages/inference/src/snippets/templates/python/requests/zeroShotImageClassification.jinja b/packages/inference/src/snippets/templates/python/requests/zeroShotImageClassification.jinja
new file mode 100644
index 0000000000..769aac400e
--- /dev/null
+++ b/packages/inference/src/snippets/templates/python/requests/zeroShotImageClassification.jinja
@@ -0,0 +1,14 @@
+def query(data):
+    with open(data["image_path"], "rb") as f:
+        img = f.read()
+    payload={
+        "parameters": data["parameters"],
+        "inputs": base64.b64encode(img).decode("utf-8")
+    }
+    response = requests.post(API_URL, headers=headers, json=payload)
+    return response.json()
+
+output = query({
+    "image_path": {{ inputs }},
+    "parameters": {"candidate_labels": ["cat", "dog", "llama"]},
+}) 
\ No newline at end of file
diff --git a/packages/inference/tsup.config.ts b/packages/inference/tsup.config.ts
new file mode 100644
index 0000000000..6be4e128a6
--- /dev/null
+++ b/packages/inference/tsup.config.ts
@@ -0,0 +1,23 @@
+import type { Options } from "tsup";
+
+const baseConfig: Options = {
+	entry: ["./index.ts"],
+	format: ["cjs", "esm"],
+	outDir: "dist",
+	clean: true,
+};
+
+const nodeConfig: Options = {
+	...baseConfig,
+	platform: "node",
+};
+
+const browserConfig: Options = {
+	...baseConfig,
+	platform: "browser",
+	target: "es2018",
+	splitting: true,
+	outDir: "dist/browser",
+};
+
+export default [nodeConfig, browserConfig];
diff --git a/packages/tasks-gen/scripts/generate-snippets-fixtures.ts b/packages/tasks-gen/scripts/generate-snippets-fixtures.ts
index b22d7551e4..c9051a2c57 100644
--- a/packages/tasks-gen/scripts/generate-snippets-fixtures.ts
+++ b/packages/tasks-gen/scripts/generate-snippets-fixtures.ts
@@ -112,6 +112,28 @@ const TEST_CASES: {
 		languages: ["py"],
 		providers: ["hf-inference"],
 	},
+	{
+		testName: "tabular",
+		model: {
+			id: "templates/tabular-classification",
+			pipeline_tag: "tabular-classification",
+			tags: [],
+			inference: "",
+		},
+		providers: ["hf-inference"],
+		languages: ["py"],
+	},
+	{
+		testName: "text-to-audio-transformers",
+		model: {
+			id: "facebook/musicgen-small",
+			pipeline_tag: "text-to-audio",
+			tags: ["transformers"],
+			inference: "",
+		},
+		providers: ["hf-inference"],
+		languages: ["py"],
+	},
 	{
 		testName: "text-to-image",
 		model: {
@@ -145,6 +167,39 @@ const TEST_CASES: {
 		providers: ["hf-inference"],
 		languages: ["sh", "js", "py"],
 	},
+	{
+		testName: "basic-snippet--token-classification",
+		model: {
+			id: "FacebookAI/xlm-roberta-large-finetuned-conll03-english",
+			pipeline_tag: "token-classification",
+			tags: [],
+			inference: "",
+		},
+		providers: ["hf-inference"],
+		languages: ["py"],
+	},
+	{
+		testName: "zero-shot-classification",
+		model: {
+			id: "facebook/bart-large-mnli",
+			pipeline_tag: "zero-shot-classification",
+			tags: [],
+			inference: "",
+		},
+		providers: ["hf-inference"],
+		languages: ["py"],
+	},
+	{
+		testName: "zero-shot-image-classification",
+		model: {
+			id: "openai/clip-vit-large-patch14",
+			pipeline_tag: "zero-shot-image-classification",
+			tags: [],
+			inference: "",
+		},
+		providers: ["hf-inference"],
+		languages: ["py"],
+	},
 ] as const;
 
 const GET_SNIPPET_FN = {
diff --git a/packages/tasks-gen/snippets-fixtures/automatic-speech-recognition/0.huggingface_hub.hf-inference.py b/packages/tasks-gen/snippets-fixtures/automatic-speech-recognition/0.huggingface_hub.hf-inference.py
index 9a945fcf65..fa9955a05b 100644
--- a/packages/tasks-gen/snippets-fixtures/automatic-speech-recognition/0.huggingface_hub.hf-inference.py
+++ b/packages/tasks-gen/snippets-fixtures/automatic-speech-recognition/0.huggingface_hub.hf-inference.py
@@ -4,4 +4,5 @@
     provider="hf-inference",
     api_key="api_token",
 )
+
 output = client.automatic_speech_recognition("sample1.flac", model="openai/whisper-large-v3-turbo")
\ No newline at end of file
diff --git a/packages/tasks-gen/snippets-fixtures/automatic-speech-recognition/1.requests.hf-inference.py b/packages/tasks-gen/snippets-fixtures/automatic-speech-recognition/1.requests.hf-inference.py
index 4acaa0ed5b..665b620769 100644
--- a/packages/tasks-gen/snippets-fixtures/automatic-speech-recognition/1.requests.hf-inference.py
+++ b/packages/tasks-gen/snippets-fixtures/automatic-speech-recognition/1.requests.hf-inference.py
@@ -4,9 +4,9 @@
 headers = {"Authorization": "Bearer api_token"}
 
 def query(filename):
-	with open(filename, "rb") as f:
-		data = f.read()
-	response = requests.post(API_URL, headers=headers, data=data)
-	return response.json()
+    with open(filename, "rb") as f:
+        data = f.read()
+    response = requests.post(API_URL, headers=headers, data=data)
+    return response.json()
 
 output = query("sample1.flac")
\ No newline at end of file
diff --git a/packages/tasks-gen/snippets-fixtures/basic-snippet--token-classification/0.huggingface_hub.hf-inference.py b/packages/tasks-gen/snippets-fixtures/basic-snippet--token-classification/0.huggingface_hub.hf-inference.py
new file mode 100644
index 0000000000..60eb648eba
--- /dev/null
+++ b/packages/tasks-gen/snippets-fixtures/basic-snippet--token-classification/0.huggingface_hub.hf-inference.py
@@ -0,0 +1,13 @@
+from huggingface_hub import InferenceClient
+
+client = InferenceClient(
+    provider="hf-inference",
+    api_key="api_token",
+)
+
+result = client.token_classification(
+    inputs="My name is Sarah Jessica Parker but you can call me Jessica",
+    model="FacebookAI/xlm-roberta-large-finetuned-conll03-english",
+)
+
+print(result)
\ No newline at end of file
diff --git a/packages/tasks-gen/snippets-fixtures/basic-snippet--token-classification/1.requests.hf-inference.py b/packages/tasks-gen/snippets-fixtures/basic-snippet--token-classification/1.requests.hf-inference.py
new file mode 100644
index 0000000000..52a5ca4217
--- /dev/null
+++ b/packages/tasks-gen/snippets-fixtures/basic-snippet--token-classification/1.requests.hf-inference.py
@@ -0,0 +1,12 @@
+import requests
+
+API_URL = "https://router.huggingface.co/hf-inference/models/FacebookAI/xlm-roberta-large-finetuned-conll03-english"
+headers = {"Authorization": "Bearer api_token"}
+
+def query(payload):
+    response = requests.post(API_URL, headers=headers, json=payload)
+    return response.json()
+
+output = query({
+    "inputs": "My name is Sarah Jessica Parker but you can call me Jessica",
+})
\ No newline at end of file
diff --git a/packages/tasks-gen/snippets-fixtures/conversational-llm-non-stream/0.huggingface_hub.hf-inference.py b/packages/tasks-gen/snippets-fixtures/conversational-llm-non-stream/0.huggingface_hub.hf-inference.py
index 44e29d2a0b..4e53fde9ac 100644
--- a/packages/tasks-gen/snippets-fixtures/conversational-llm-non-stream/0.huggingface_hub.hf-inference.py
+++ b/packages/tasks-gen/snippets-fixtures/conversational-llm-non-stream/0.huggingface_hub.hf-inference.py
@@ -13,9 +13,9 @@
 ]
 
 completion = client.chat.completions.create(
-    model="meta-llama/Llama-3.1-8B-Instruct", 
-	messages=messages, 
-	max_tokens=500,
+    model="meta-llama/Llama-3.1-8B-Instruct",
+    messages=messages,
+    max_tokens=500,
 )
 
 print(completion.choices[0].message)
\ No newline at end of file
diff --git a/packages/tasks-gen/snippets-fixtures/conversational-llm-non-stream/0.huggingface_hub.together.py b/packages/tasks-gen/snippets-fixtures/conversational-llm-non-stream/0.huggingface_hub.together.py
index ee348d6987..0e5231e4dc 100644
--- a/packages/tasks-gen/snippets-fixtures/conversational-llm-non-stream/0.huggingface_hub.together.py
+++ b/packages/tasks-gen/snippets-fixtures/conversational-llm-non-stream/0.huggingface_hub.together.py
@@ -13,9 +13,9 @@
 ]
 
 completion = client.chat.completions.create(
-    model="meta-llama/Llama-3.1-8B-Instruct", 
-	messages=messages, 
-	max_tokens=500,
+    model="meta-llama/Llama-3.1-8B-Instruct",
+    messages=messages,
+    max_tokens=500,
 )
 
 print(completion.choices[0].message)
\ No newline at end of file
diff --git a/packages/tasks-gen/snippets-fixtures/conversational-llm-non-stream/1.openai.hf-inference.py b/packages/tasks-gen/snippets-fixtures/conversational-llm-non-stream/1.openai.hf-inference.py
index 4d557b289e..bed410d47b 100644
--- a/packages/tasks-gen/snippets-fixtures/conversational-llm-non-stream/1.openai.hf-inference.py
+++ b/packages/tasks-gen/snippets-fixtures/conversational-llm-non-stream/1.openai.hf-inference.py
@@ -1,8 +1,8 @@
 from openai import OpenAI
 
 client = OpenAI(
-	base_url="https://router.huggingface.co/hf-inference/v1",
-	api_key="api_token"
+    base_url="https://router.huggingface.co/hf-inference/v1",
+    api_key="api_token"
 )
 
 messages = [
@@ -13,9 +13,9 @@
 ]
 
 completion = client.chat.completions.create(
-	model="meta-llama/Llama-3.1-8B-Instruct", 
-	messages=messages, 
-	max_tokens=500,
+    model="meta-llama/Llama-3.1-8B-Instruct",
+    messages=messages,
+    max_tokens=500,
 )
 
 print(completion.choices[0].message)
\ No newline at end of file
diff --git a/packages/tasks-gen/snippets-fixtures/conversational-llm-non-stream/1.openai.together.py b/packages/tasks-gen/snippets-fixtures/conversational-llm-non-stream/1.openai.together.py
index d4fb0f865c..ea0f8683bc 100644
--- a/packages/tasks-gen/snippets-fixtures/conversational-llm-non-stream/1.openai.together.py
+++ b/packages/tasks-gen/snippets-fixtures/conversational-llm-non-stream/1.openai.together.py
@@ -1,8 +1,8 @@
 from openai import OpenAI
 
 client = OpenAI(
-	base_url="https://router.huggingface.co/together",
-	api_key="api_token"
+    base_url="https://router.huggingface.co/together",
+    api_key="api_token"
 )
 
 messages = [
@@ -13,9 +13,9 @@
 ]
 
 completion = client.chat.completions.create(
-	model="<together alias for meta-llama/Llama-3.1-8B-Instruct>", 
-	messages=messages, 
-	max_tokens=500,
+    model="<together alias for meta-llama/Llama-3.1-8B-Instruct>",
+    messages=messages,
+    max_tokens=500,
 )
 
 print(completion.choices[0].message)
\ No newline at end of file
diff --git a/packages/tasks-gen/snippets-fixtures/conversational-llm-stream/0.huggingface_hub.hf-inference.py b/packages/tasks-gen/snippets-fixtures/conversational-llm-stream/0.huggingface_hub.hf-inference.py
index 0f1fd0dbdc..a99e90b97c 100644
--- a/packages/tasks-gen/snippets-fixtures/conversational-llm-stream/0.huggingface_hub.hf-inference.py
+++ b/packages/tasks-gen/snippets-fixtures/conversational-llm-stream/0.huggingface_hub.hf-inference.py
@@ -13,10 +13,10 @@
 ]
 
 stream = client.chat.completions.create(
-	model="meta-llama/Llama-3.1-8B-Instruct", 
-	messages=messages, 
-	max_tokens=500,
-	stream=True,
+    model="meta-llama/Llama-3.1-8B-Instruct",
+    messages=messages,
+    max_tokens=500,
+    stream=True,
 )
 
 for chunk in stream:
diff --git a/packages/tasks-gen/snippets-fixtures/conversational-llm-stream/0.huggingface_hub.together.py b/packages/tasks-gen/snippets-fixtures/conversational-llm-stream/0.huggingface_hub.together.py
index 31d00d979c..d4bb30cda6 100644
--- a/packages/tasks-gen/snippets-fixtures/conversational-llm-stream/0.huggingface_hub.together.py
+++ b/packages/tasks-gen/snippets-fixtures/conversational-llm-stream/0.huggingface_hub.together.py
@@ -13,10 +13,10 @@
 ]
 
 stream = client.chat.completions.create(
-	model="meta-llama/Llama-3.1-8B-Instruct", 
-	messages=messages, 
-	max_tokens=500,
-	stream=True,
+    model="meta-llama/Llama-3.1-8B-Instruct",
+    messages=messages,
+    max_tokens=500,
+    stream=True,
 )
 
 for chunk in stream:
diff --git a/packages/tasks-gen/snippets-fixtures/conversational-llm-stream/1.openai.hf-inference.py b/packages/tasks-gen/snippets-fixtures/conversational-llm-stream/1.openai.hf-inference.py
index 12167b83bc..1476285a3b 100644
--- a/packages/tasks-gen/snippets-fixtures/conversational-llm-stream/1.openai.hf-inference.py
+++ b/packages/tasks-gen/snippets-fixtures/conversational-llm-stream/1.openai.hf-inference.py
@@ -1,8 +1,8 @@
 from openai import OpenAI
 
 client = OpenAI(
-	base_url="https://router.huggingface.co/hf-inference/v1",
-	api_key="api_token"
+    base_url="https://router.huggingface.co/hf-inference/v1",
+    api_key="api_token"
 )
 
 messages = [
@@ -13,11 +13,11 @@
 ]
 
 stream = client.chat.completions.create(
-    model="meta-llama/Llama-3.1-8B-Instruct", 
-	messages=messages, 
-	max_tokens=500,
-	stream=True
+    model="meta-llama/Llama-3.1-8B-Instruct",
+    messages=messages,
+    max_tokens=500,
+    stream=True
 )
 
 for chunk in stream:
-	print(chunk.choices[0].delta.content, end="")
\ No newline at end of file
+    print(chunk.choices[0].delta.content, end="")
\ No newline at end of file
diff --git a/packages/tasks-gen/snippets-fixtures/conversational-llm-stream/1.openai.together.py b/packages/tasks-gen/snippets-fixtures/conversational-llm-stream/1.openai.together.py
index 67d0bde845..e04dfcb7ed 100644
--- a/packages/tasks-gen/snippets-fixtures/conversational-llm-stream/1.openai.together.py
+++ b/packages/tasks-gen/snippets-fixtures/conversational-llm-stream/1.openai.together.py
@@ -1,8 +1,8 @@
 from openai import OpenAI
 
 client = OpenAI(
-	base_url="https://router.huggingface.co/together",
-	api_key="api_token"
+    base_url="https://router.huggingface.co/together",
+    api_key="api_token"
 )
 
 messages = [
@@ -13,11 +13,11 @@
 ]
 
 stream = client.chat.completions.create(
-    model="<together alias for meta-llama/Llama-3.1-8B-Instruct>", 
-	messages=messages, 
-	max_tokens=500,
-	stream=True
+    model="<together alias for meta-llama/Llama-3.1-8B-Instruct>",
+    messages=messages,
+    max_tokens=500,
+    stream=True
 )
 
 for chunk in stream:
-	print(chunk.choices[0].delta.content, end="")
\ No newline at end of file
+    print(chunk.choices[0].delta.content, end="")
\ No newline at end of file
diff --git a/packages/tasks-gen/snippets-fixtures/conversational-vlm-non-stream/0.huggingface_hub.fireworks-ai.py b/packages/tasks-gen/snippets-fixtures/conversational-vlm-non-stream/0.huggingface_hub.fireworks-ai.py
index 41f26d30a7..27fbac0fba 100644
--- a/packages/tasks-gen/snippets-fixtures/conversational-vlm-non-stream/0.huggingface_hub.fireworks-ai.py
+++ b/packages/tasks-gen/snippets-fixtures/conversational-vlm-non-stream/0.huggingface_hub.fireworks-ai.py
@@ -24,9 +24,9 @@
 ]
 
 completion = client.chat.completions.create(
-    model="meta-llama/Llama-3.2-11B-Vision-Instruct", 
-	messages=messages, 
-	max_tokens=500,
+    model="meta-llama/Llama-3.2-11B-Vision-Instruct",
+    messages=messages,
+    max_tokens=500,
 )
 
 print(completion.choices[0].message)
\ No newline at end of file
diff --git a/packages/tasks-gen/snippets-fixtures/conversational-vlm-non-stream/0.huggingface_hub.hf-inference.py b/packages/tasks-gen/snippets-fixtures/conversational-vlm-non-stream/0.huggingface_hub.hf-inference.py
index 507809756d..02201bfa11 100644
--- a/packages/tasks-gen/snippets-fixtures/conversational-vlm-non-stream/0.huggingface_hub.hf-inference.py
+++ b/packages/tasks-gen/snippets-fixtures/conversational-vlm-non-stream/0.huggingface_hub.hf-inference.py
@@ -24,9 +24,9 @@
 ]
 
 completion = client.chat.completions.create(
-    model="meta-llama/Llama-3.2-11B-Vision-Instruct", 
-	messages=messages, 
-	max_tokens=500,
+    model="meta-llama/Llama-3.2-11B-Vision-Instruct",
+    messages=messages,
+    max_tokens=500,
 )
 
 print(completion.choices[0].message)
\ No newline at end of file
diff --git a/packages/tasks-gen/snippets-fixtures/conversational-vlm-non-stream/1.openai.fireworks-ai.py b/packages/tasks-gen/snippets-fixtures/conversational-vlm-non-stream/1.openai.fireworks-ai.py
index 19a126c63d..25490a6a9c 100644
--- a/packages/tasks-gen/snippets-fixtures/conversational-vlm-non-stream/1.openai.fireworks-ai.py
+++ b/packages/tasks-gen/snippets-fixtures/conversational-vlm-non-stream/1.openai.fireworks-ai.py
@@ -1,8 +1,8 @@
 from openai import OpenAI
 
 client = OpenAI(
-	base_url="https://router.huggingface.co/fireworks-ai",
-	api_key="api_token"
+    base_url="https://router.huggingface.co/fireworks-ai",
+    api_key="api_token"
 )
 
 messages = [
@@ -24,9 +24,9 @@
 ]
 
 completion = client.chat.completions.create(
-	model="<fireworks-ai alias for meta-llama/Llama-3.2-11B-Vision-Instruct>", 
-	messages=messages, 
-	max_tokens=500,
+    model="<fireworks-ai alias for meta-llama/Llama-3.2-11B-Vision-Instruct>",
+    messages=messages,
+    max_tokens=500,
 )
 
 print(completion.choices[0].message)
\ No newline at end of file
diff --git a/packages/tasks-gen/snippets-fixtures/conversational-vlm-non-stream/1.openai.hf-inference.py b/packages/tasks-gen/snippets-fixtures/conversational-vlm-non-stream/1.openai.hf-inference.py
index 517466190f..dab0422b2d 100644
--- a/packages/tasks-gen/snippets-fixtures/conversational-vlm-non-stream/1.openai.hf-inference.py
+++ b/packages/tasks-gen/snippets-fixtures/conversational-vlm-non-stream/1.openai.hf-inference.py
@@ -1,8 +1,8 @@
 from openai import OpenAI
 
 client = OpenAI(
-	base_url="https://router.huggingface.co/hf-inference/v1",
-	api_key="api_token"
+    base_url="https://router.huggingface.co/hf-inference/v1",
+    api_key="api_token"
 )
 
 messages = [
@@ -24,9 +24,9 @@
 ]
 
 completion = client.chat.completions.create(
-	model="meta-llama/Llama-3.2-11B-Vision-Instruct", 
-	messages=messages, 
-	max_tokens=500,
+    model="meta-llama/Llama-3.2-11B-Vision-Instruct",
+    messages=messages,
+    max_tokens=500,
 )
 
 print(completion.choices[0].message)
\ No newline at end of file
diff --git a/packages/tasks-gen/snippets-fixtures/conversational-vlm-stream/0.huggingface_hub.fireworks-ai.py b/packages/tasks-gen/snippets-fixtures/conversational-vlm-stream/0.huggingface_hub.fireworks-ai.py
index ad86364e4d..6395669c4c 100644
--- a/packages/tasks-gen/snippets-fixtures/conversational-vlm-stream/0.huggingface_hub.fireworks-ai.py
+++ b/packages/tasks-gen/snippets-fixtures/conversational-vlm-stream/0.huggingface_hub.fireworks-ai.py
@@ -24,10 +24,10 @@
 ]
 
 stream = client.chat.completions.create(
-	model="meta-llama/Llama-3.2-11B-Vision-Instruct", 
-	messages=messages, 
-	max_tokens=500,
-	stream=True,
+    model="meta-llama/Llama-3.2-11B-Vision-Instruct",
+    messages=messages,
+    max_tokens=500,
+    stream=True,
 )
 
 for chunk in stream:
diff --git a/packages/tasks-gen/snippets-fixtures/conversational-vlm-stream/0.huggingface_hub.hf-inference.py b/packages/tasks-gen/snippets-fixtures/conversational-vlm-stream/0.huggingface_hub.hf-inference.py
index 7d58f8020a..dc31c62ae2 100644
--- a/packages/tasks-gen/snippets-fixtures/conversational-vlm-stream/0.huggingface_hub.hf-inference.py
+++ b/packages/tasks-gen/snippets-fixtures/conversational-vlm-stream/0.huggingface_hub.hf-inference.py
@@ -24,10 +24,10 @@
 ]
 
 stream = client.chat.completions.create(
-	model="meta-llama/Llama-3.2-11B-Vision-Instruct", 
-	messages=messages, 
-	max_tokens=500,
-	stream=True,
+    model="meta-llama/Llama-3.2-11B-Vision-Instruct",
+    messages=messages,
+    max_tokens=500,
+    stream=True,
 )
 
 for chunk in stream:
diff --git a/packages/tasks-gen/snippets-fixtures/conversational-vlm-stream/1.openai.fireworks-ai.py b/packages/tasks-gen/snippets-fixtures/conversational-vlm-stream/1.openai.fireworks-ai.py
index 33acb547ad..81bede83e9 100644
--- a/packages/tasks-gen/snippets-fixtures/conversational-vlm-stream/1.openai.fireworks-ai.py
+++ b/packages/tasks-gen/snippets-fixtures/conversational-vlm-stream/1.openai.fireworks-ai.py
@@ -1,8 +1,8 @@
 from openai import OpenAI
 
 client = OpenAI(
-	base_url="https://router.huggingface.co/fireworks-ai",
-	api_key="api_token"
+    base_url="https://router.huggingface.co/fireworks-ai",
+    api_key="api_token"
 )
 
 messages = [
@@ -24,11 +24,11 @@
 ]
 
 stream = client.chat.completions.create(
-    model="<fireworks-ai alias for meta-llama/Llama-3.2-11B-Vision-Instruct>", 
-	messages=messages, 
-	max_tokens=500,
-	stream=True
+    model="<fireworks-ai alias for meta-llama/Llama-3.2-11B-Vision-Instruct>",
+    messages=messages,
+    max_tokens=500,
+    stream=True
 )
 
 for chunk in stream:
-	print(chunk.choices[0].delta.content, end="")
\ No newline at end of file
+    print(chunk.choices[0].delta.content, end="")
\ No newline at end of file
diff --git a/packages/tasks-gen/snippets-fixtures/conversational-vlm-stream/1.openai.hf-inference.py b/packages/tasks-gen/snippets-fixtures/conversational-vlm-stream/1.openai.hf-inference.py
index 6c788816ee..8ceb00bfe3 100644
--- a/packages/tasks-gen/snippets-fixtures/conversational-vlm-stream/1.openai.hf-inference.py
+++ b/packages/tasks-gen/snippets-fixtures/conversational-vlm-stream/1.openai.hf-inference.py
@@ -1,8 +1,8 @@
 from openai import OpenAI
 
 client = OpenAI(
-	base_url="https://router.huggingface.co/hf-inference/v1",
-	api_key="api_token"
+    base_url="https://router.huggingface.co/hf-inference/v1",
+    api_key="api_token"
 )
 
 messages = [
@@ -24,11 +24,11 @@
 ]
 
 stream = client.chat.completions.create(
-    model="meta-llama/Llama-3.2-11B-Vision-Instruct", 
-	messages=messages, 
-	max_tokens=500,
-	stream=True
+    model="meta-llama/Llama-3.2-11B-Vision-Instruct",
+    messages=messages,
+    max_tokens=500,
+    stream=True
 )
 
 for chunk in stream:
-	print(chunk.choices[0].delta.content, end="")
\ No newline at end of file
+    print(chunk.choices[0].delta.content, end="")
\ No newline at end of file
diff --git a/packages/tasks-gen/snippets-fixtures/document-question-answering/0.huggingface_hub.hf-inference.py b/packages/tasks-gen/snippets-fixtures/document-question-answering/0.huggingface_hub.hf-inference.py
index 46e437c63a..738921cc13 100644
--- a/packages/tasks-gen/snippets-fixtures/document-question-answering/0.huggingface_hub.hf-inference.py
+++ b/packages/tasks-gen/snippets-fixtures/document-question-answering/0.huggingface_hub.hf-inference.py
@@ -4,8 +4,9 @@
     provider="hf-inference",
     api_key="api_token",
 )
+
 output = client.document_question_answering(
     "cat.png",
-	question="What is in this image?",
-	model="impira/layoutlm-invoices",
+    question="What is in this image?",
+    model="impira/layoutlm-invoices",
 )
\ No newline at end of file
diff --git a/packages/tasks-gen/snippets-fixtures/document-question-answering/1.requests.hf-inference.py b/packages/tasks-gen/snippets-fixtures/document-question-answering/1.requests.hf-inference.py
index 51fed678d8..d8ab49b354 100644
--- a/packages/tasks-gen/snippets-fixtures/document-question-answering/1.requests.hf-inference.py
+++ b/packages/tasks-gen/snippets-fixtures/document-question-answering/1.requests.hf-inference.py
@@ -5,11 +5,11 @@
 headers = {"Authorization": "Bearer api_token"}
 
 def query(payload):
-	with open(payload["image"], "rb") as f:
-		img = f.read()
-		payload["image"] = base64.b64encode(img).decode("utf-8")
-	response = requests.post(API_URL, headers=headers, json=payload)
-	return response.json()
+    with open(payload["image"], "rb") as f:
+        img = f.read()
+        payload["image"] = base64.b64encode(img).decode("utf-8")
+    response = requests.post(API_URL, headers=headers, json=payload)
+    return response.json()
 
 output = query({
     "inputs": {
diff --git a/packages/tasks-gen/snippets-fixtures/image-to-image/0.huggingface_hub.hf-inference.py b/packages/tasks-gen/snippets-fixtures/image-to-image/0.huggingface_hub.hf-inference.py
index c21949966a..dc89746110 100644
--- a/packages/tasks-gen/snippets-fixtures/image-to-image/0.huggingface_hub.hf-inference.py
+++ b/packages/tasks-gen/snippets-fixtures/image-to-image/0.huggingface_hub.hf-inference.py
@@ -4,6 +4,7 @@
     provider="hf-inference",
     api_key="api_token",
 )
+
 # output is a PIL.Image object
 image = client.image_to_image(
     "cat.png",
diff --git a/packages/tasks-gen/snippets-fixtures/image-to-image/1.requests.hf-inference.py b/packages/tasks-gen/snippets-fixtures/image-to-image/1.requests.hf-inference.py
index 2d16ba05f5..d411492847 100644
--- a/packages/tasks-gen/snippets-fixtures/image-to-image/1.requests.hf-inference.py
+++ b/packages/tasks-gen/snippets-fixtures/image-to-image/1.requests.hf-inference.py
@@ -5,15 +5,15 @@
 headers = {"Authorization": "Bearer api_token"}
 
 def query(payload):
-	with open(payload["inputs"], "rb") as f:
-		img = f.read()
-		payload["inputs"] = base64.b64encode(img).decode("utf-8")
-	response = requests.post(API_URL, headers=headers, json=payload)
-	return response.content
+    with open(payload["inputs"], "rb") as f:
+        img = f.read()
+        payload["inputs"] = base64.b64encode(img).decode("utf-8")
+    response = requests.post(API_URL, headers=headers, json=payload)
+    return response.content
 
 image_bytes = query({
-	"inputs": "cat.png",
-	"parameters": {"prompt": "Turn the cat into a tiger."},
+    "inputs": "cat.png",
+    "parameters": {"prompt": "Turn the cat into a tiger."},
 })
 
 # You can access the image with PIL.Image for example
diff --git a/packages/tasks-gen/snippets-fixtures/tabular/0.requests.hf-inference.py b/packages/tasks-gen/snippets-fixtures/tabular/0.requests.hf-inference.py
new file mode 100644
index 0000000000..e2714639b4
--- /dev/null
+++ b/packages/tasks-gen/snippets-fixtures/tabular/0.requests.hf-inference.py
@@ -0,0 +1,14 @@
+import requests
+
+API_URL = "https://router.huggingface.co/hf-inference/models/templates/tabular-classification"
+headers = {"Authorization": "Bearer api_token"}
+
+def query(payload):
+    response = requests.post(API_URL, headers=headers, json=payload)
+    return response.content
+
+response = query({
+    "inputs": {
+        "data": '{"Height":[11.52,12.48],"Length1":[23.2,24.0],"Length2":[25.4,26.3],"Species": ["Bream","Bream"]}'
+    },
+})
\ No newline at end of file
diff --git a/packages/tasks-gen/snippets-fixtures/text-classification/0.huggingface_hub.hf-inference.py b/packages/tasks-gen/snippets-fixtures/text-classification/0.huggingface_hub.hf-inference.py
index d6b9b9aa45..3150e44217 100644
--- a/packages/tasks-gen/snippets-fixtures/text-classification/0.huggingface_hub.hf-inference.py
+++ b/packages/tasks-gen/snippets-fixtures/text-classification/0.huggingface_hub.hf-inference.py
@@ -6,8 +6,8 @@
 )
 
 result = client.text_classification(
-	inputs="I like you. I love you",
-	model="distilbert/distilbert-base-uncased-finetuned-sst-2-english",
+    inputs="I like you. I love you",
+    model="distilbert/distilbert-base-uncased-finetuned-sst-2-english",
 )
 
-print(result)
+print(result)
\ No newline at end of file
diff --git a/packages/tasks-gen/snippets-fixtures/text-classification/1.requests.hf-inference.py b/packages/tasks-gen/snippets-fixtures/text-classification/1.requests.hf-inference.py
index 281fca0df2..cb900c6b72 100644
--- a/packages/tasks-gen/snippets-fixtures/text-classification/1.requests.hf-inference.py
+++ b/packages/tasks-gen/snippets-fixtures/text-classification/1.requests.hf-inference.py
@@ -4,9 +4,9 @@
 headers = {"Authorization": "Bearer api_token"}
 
 def query(payload):
-	response = requests.post(API_URL, headers=headers, json=payload)
-	return response.json()
-	
+    response = requests.post(API_URL, headers=headers, json=payload)
+    return response.json()
+
 output = query({
-	"inputs": "I like you. I love you",
+    "inputs": "I like you. I love you",
 })
\ No newline at end of file
diff --git a/packages/tasks-gen/snippets-fixtures/text-to-audio-transformers/0.requests.hf-inference.py b/packages/tasks-gen/snippets-fixtures/text-to-audio-transformers/0.requests.hf-inference.py
new file mode 100644
index 0000000000..ad492aee35
--- /dev/null
+++ b/packages/tasks-gen/snippets-fixtures/text-to-audio-transformers/0.requests.hf-inference.py
@@ -0,0 +1,15 @@
+import requests
+
+API_URL = "https://router.huggingface.co/hf-inference/models/facebook/musicgen-small"
+headers = {"Authorization": "Bearer api_token"}
+
+def query(payload):
+    response = requests.post(API_URL, headers=headers, json=payload)
+    return response.json()
+
+audio, sampling_rate = query({
+    "inputs": "liquid drum and bass, atmospheric synths, airy sounds",
+})
+# You can access the audio with IPython.display for example
+from IPython.display import Audio
+Audio(audio, rate=sampling_rate)
\ No newline at end of file
diff --git a/packages/tasks-gen/snippets-fixtures/text-to-image/0.huggingface_hub.fal-ai.py b/packages/tasks-gen/snippets-fixtures/text-to-image/0.huggingface_hub.fal-ai.py
index 8f8a6b1021..f70ff34f3a 100644
--- a/packages/tasks-gen/snippets-fixtures/text-to-image/0.huggingface_hub.fal-ai.py
+++ b/packages/tasks-gen/snippets-fixtures/text-to-image/0.huggingface_hub.fal-ai.py
@@ -7,6 +7,6 @@
 
 # output is a PIL.Image object
 image = client.text_to_image(
-	"Astronaut riding a horse",
-	model="black-forest-labs/FLUX.1-schnell",
+    "Astronaut riding a horse",
+    model="black-forest-labs/FLUX.1-schnell",
 )
\ No newline at end of file
diff --git a/packages/tasks-gen/snippets-fixtures/text-to-image/0.huggingface_hub.hf-inference.py b/packages/tasks-gen/snippets-fixtures/text-to-image/0.huggingface_hub.hf-inference.py
index b4c48cec43..84ce49c29c 100644
--- a/packages/tasks-gen/snippets-fixtures/text-to-image/0.huggingface_hub.hf-inference.py
+++ b/packages/tasks-gen/snippets-fixtures/text-to-image/0.huggingface_hub.hf-inference.py
@@ -7,6 +7,6 @@
 
 # output is a PIL.Image object
 image = client.text_to_image(
-	"Astronaut riding a horse",
-	model="black-forest-labs/FLUX.1-schnell",
+    "Astronaut riding a horse",
+    model="black-forest-labs/FLUX.1-schnell",
 )
\ No newline at end of file
diff --git a/packages/tasks-gen/snippets-fixtures/text-to-image/1.fal-client.fal-ai.py b/packages/tasks-gen/snippets-fixtures/text-to-image/1.fal-client.fal-ai.py
deleted file mode 100644
index 36220f1057..0000000000
--- a/packages/tasks-gen/snippets-fixtures/text-to-image/1.fal-client.fal-ai.py
+++ /dev/null
@@ -1,9 +0,0 @@
-import fal_client
-
-result = fal_client.subscribe(
-	"<fal-ai alias for black-forest-labs/FLUX.1-schnell>",
-	arguments={
-		"prompt": "Astronaut riding a horse",
-	},
-)
-print(result)
diff --git a/packages/tasks-gen/snippets-fixtures/text-to-image/1.fal_client.fal-ai.py b/packages/tasks-gen/snippets-fixtures/text-to-image/1.fal_client.fal-ai.py
new file mode 100644
index 0000000000..461dbb2271
--- /dev/null
+++ b/packages/tasks-gen/snippets-fixtures/text-to-image/1.fal_client.fal-ai.py
@@ -0,0 +1,9 @@
+import fal_client
+
+result = fal_client.subscribe(
+    "<fal-ai alias for black-forest-labs/FLUX.1-schnell>",
+    arguments={
+        "prompt": "Astronaut riding a horse",
+    },
+)
+print(result)
\ No newline at end of file
diff --git a/packages/tasks-gen/snippets-fixtures/text-to-image/1.requests.hf-inference.py b/packages/tasks-gen/snippets-fixtures/text-to-image/1.requests.hf-inference.py
index a8e1ad9de0..568c788dc7 100644
--- a/packages/tasks-gen/snippets-fixtures/text-to-image/1.requests.hf-inference.py
+++ b/packages/tasks-gen/snippets-fixtures/text-to-image/1.requests.hf-inference.py
@@ -4,11 +4,11 @@
 headers = {"Authorization": "Bearer api_token"}
 
 def query(payload):
-	response = requests.post(API_URL, headers=headers, json=payload)
-	return response.content
+    response = requests.post(API_URL, headers=headers, json=payload)
+    return response.content
 
 image_bytes = query({
-	"inputs": "Astronaut riding a horse",
+    "inputs": "Astronaut riding a horse",
 })
 
 # You can access the image with PIL.Image for example
diff --git a/packages/tasks-gen/snippets-fixtures/text-to-video/0.huggingface_hub.fal-ai.py b/packages/tasks-gen/snippets-fixtures/text-to-video/0.huggingface_hub.fal-ai.py
index fc96fd381b..240c59184d 100644
--- a/packages/tasks-gen/snippets-fixtures/text-to-video/0.huggingface_hub.fal-ai.py
+++ b/packages/tasks-gen/snippets-fixtures/text-to-video/0.huggingface_hub.fal-ai.py
@@ -6,6 +6,6 @@
 )
 
 video = client.text_to_video(
-	"A young man walking on the street",
-	model="tencent/HunyuanVideo",
+    "A young man walking on the street",
+    model="tencent/HunyuanVideo",
 )
\ No newline at end of file
diff --git a/packages/tasks-gen/snippets-fixtures/text-to-video/0.huggingface_hub.replicate.py b/packages/tasks-gen/snippets-fixtures/text-to-video/0.huggingface_hub.replicate.py
index 208c583119..6299fe51d5 100644
--- a/packages/tasks-gen/snippets-fixtures/text-to-video/0.huggingface_hub.replicate.py
+++ b/packages/tasks-gen/snippets-fixtures/text-to-video/0.huggingface_hub.replicate.py
@@ -6,6 +6,6 @@
 )
 
 video = client.text_to_video(
-	"A young man walking on the street",
-	model="tencent/HunyuanVideo",
+    "A young man walking on the street",
+    model="tencent/HunyuanVideo",
 )
\ No newline at end of file
diff --git a/packages/tasks-gen/snippets-fixtures/zero-shot-classification/0.requests.hf-inference.py b/packages/tasks-gen/snippets-fixtures/zero-shot-classification/0.requests.hf-inference.py
new file mode 100644
index 0000000000..3054fe8de8
--- /dev/null
+++ b/packages/tasks-gen/snippets-fixtures/zero-shot-classification/0.requests.hf-inference.py
@@ -0,0 +1,13 @@
+import requests
+
+API_URL = "https://router.huggingface.co/hf-inference/models/facebook/bart-large-mnli"
+headers = {"Authorization": "Bearer api_token"}
+
+def query(payload):
+    response = requests.post(API_URL, headers=headers, json=payload)
+    return response.json()
+
+output = query({
+    "inputs": "Hi, I recently bought a device from your company but it is not working as advertised and I would like to get reimbursed!",
+    "parameters": {"candidate_labels": ["refund", "legal", "faq"]},
+})
\ No newline at end of file
diff --git a/packages/tasks-gen/snippets-fixtures/zero-shot-image-classification/0.requests.hf-inference.py b/packages/tasks-gen/snippets-fixtures/zero-shot-image-classification/0.requests.hf-inference.py
new file mode 100644
index 0000000000..a6cfd80da1
--- /dev/null
+++ b/packages/tasks-gen/snippets-fixtures/zero-shot-image-classification/0.requests.hf-inference.py
@@ -0,0 +1,20 @@
+import base64
+import requests
+
+API_URL = "https://router.huggingface.co/hf-inference/models/openai/clip-vit-large-patch14"
+headers = {"Authorization": "Bearer api_token"}
+
+def query(data):
+    with open(data["image_path"], "rb") as f:
+        img = f.read()
+    payload={
+        "parameters": data["parameters"],
+        "inputs": base64.b64encode(img).decode("utf-8")
+    }
+    response = requests.post(API_URL, headers=headers, json=payload)
+    return response.json()
+
+output = query({
+    "image_path": "cats.jpg",
+    "parameters": {"candidate_labels": ["cat", "dog", "llama"]},
+})
\ No newline at end of file