@huggingface/tasks 0.12.22 → 0.12.24

package/src/snippets/curl.ts

@@ -1,36 +1,73 @@
 import type { PipelineType } from "../pipelines.js";
+import type { ChatCompletionInputMessage, GenerationParameters } from "../tasks/index.js";
+import { stringifyGenerationConfig, stringifyMessages } from "./common.js";
 import { getModelInputSnippet } from "./inputs.js";
-import type { ModelDataMinimal } from "./types.js";
+import type { InferenceSnippet, ModelDataMinimal } from "./types.js";
 
-export const snippetBasic = (model: ModelDataMinimal, accessToken: string): string =>
-	`curl https://api-inference.huggingface.co/models/${model.id} \\
+export const snippetBasic = (model: ModelDataMinimal, accessToken: string): InferenceSnippet => ({
+	content: `curl https://api-inference.huggingface.co/models/${model.id} \\
 	-X POST \\
 	-d '{"inputs": ${getModelInputSnippet(model, true)}}' \\
 	-H 'Content-Type: application/json' \\
-	-H "Authorization: Bearer ${accessToken || `{API_TOKEN}`}"`;
+	-H "Authorization: Bearer ${accessToken || `{API_TOKEN}`}"`,
+});
 
-export const snippetTextGeneration = (model: ModelDataMinimal, accessToken: string): string => {
+export const snippetTextGeneration = (
+	model: ModelDataMinimal,
+	accessToken: string,
+	opts?: {
+		streaming?: boolean;
+		messages?: ChatCompletionInputMessage[];
+		temperature?: GenerationParameters["temperature"];
+		max_tokens?: GenerationParameters["max_tokens"];
+		top_p?: GenerationParameters["top_p"];
+	}
+): InferenceSnippet => {
 	if (model.tags.includes("conversational")) {
 		// Conversational model detected, so we display a code snippet that features the Messages API
-		return `curl 'https://api-inference.huggingface.co/models/${model.id}/v1/chat/completions' \\
+		const streaming = opts?.streaming ?? true;
+		const messages: ChatCompletionInputMessage[] = opts?.messages ?? [
+			{ role: "user", content: "What is the capital of France?" },
+		];
+
+		const config = {
+			...(opts?.temperature ? { temperature: opts.temperature } : undefined),
+			max_tokens: opts?.max_tokens ?? 500,
+			...(opts?.top_p ? { top_p: opts.top_p } : undefined),
+		};
+		return {
+			content: `curl 'https://api-inference.huggingface.co/models/${model.id}/v1/chat/completions' \\
 -H "Authorization: Bearer ${accessToken || `{API_TOKEN}`}" \\
 -H 'Content-Type: application/json' \\
--d '{
-	"model": "${model.id}",
-	"messages": [{"role": "user", "content": "What is the capital of France?"}],
-	"max_tokens": 500,
-	"stream": false
-}'
-`;
+--data '{
+    "model": "${model.id}",
+    "messages": ${stringifyMessages(messages, {
+			sep: ",\n\t\t",
+			start: `[\n\t\t`,
+			end: `\n\t]`,
+			attributeKeyQuotes: true,
+			customContentEscaper: (str) => str.replace(/'/g, "'\\''"),
+		})},
+    ${stringifyGenerationConfig(config, {
+			sep: ",\n    ",
+			start: "",
+			end: "",
+			attributeKeyQuotes: true,
+			attributeValueConnector: ": ",
+		})},
+    "stream": ${!!streaming}
+}'`,
+		};
 	} else {
 		return snippetBasic(model, accessToken);
 	}
 };
 
-export const snippetImageTextToTextGeneration = (model: ModelDataMinimal, accessToken: string): string => {
+export const snippetImageTextToTextGeneration = (model: ModelDataMinimal, accessToken: string): InferenceSnippet => {
 	if (model.tags.includes("conversational")) {
 		// Conversational model detected, so we display a code snippet that features the Messages API
-		return `curl 'https://api-inference.huggingface.co/models/${model.id}/v1/chat/completions' \\
+		return {
+			content: `curl 'https://api-inference.huggingface.co/models/${model.id}/v1/chat/completions' \\
 -H "Authorization: Bearer ${accessToken || `{API_TOKEN}`}" \\
 -H 'Content-Type: application/json' \\
 -d '{
@@ -47,26 +84,34 @@ export const snippetImageTextToTextGeneration = (model: ModelDataMinimal, access
 	"max_tokens": 500,
 	"stream": false
 }'
-`;
+`,
+		};
 	} else {
 		return snippetBasic(model, accessToken);
 	}
 };
 
-export const snippetZeroShotClassification = (model: ModelDataMinimal, accessToken: string): string =>
-	`curl https://api-inference.huggingface.co/models/${model.id} \\
+export const snippetZeroShotClassification = (model: ModelDataMinimal, accessToken: string): InferenceSnippet => ({
+	content: `curl https://api-inference.huggingface.co/models/${model.id} \\
 	-X POST \\
 	-d '{"inputs": ${getModelInputSnippet(model, true)}, "parameters": {"candidate_labels": ["refund", "legal", "faq"]}}' \\
 	-H 'Content-Type: application/json' \\
-	-H "Authorization: Bearer ${accessToken || `{API_TOKEN}`}"`;
+	-H "Authorization: Bearer ${accessToken || `{API_TOKEN}`}"`,
+});
 
-export const snippetFile = (model: ModelDataMinimal, accessToken: string): string =>
-	`curl https://api-inference.huggingface.co/models/${model.id} \\
+export const snippetFile = (model: ModelDataMinimal, accessToken: string): InferenceSnippet => ({
+	content: `curl https://api-inference.huggingface.co/models/${model.id} \\
 	-X POST \\
 	--data-binary '@${getModelInputSnippet(model, true, true)}' \\
-	-H "Authorization: Bearer ${accessToken || `{API_TOKEN}`}"`;
+	-H "Authorization: Bearer ${accessToken || `{API_TOKEN}`}"`,
+});
 
-export const curlSnippets: Partial<Record<PipelineType, (model: ModelDataMinimal, accessToken: string) => string>> = {
+export const curlSnippets: Partial<
+	Record<
+		PipelineType,
+		(model: ModelDataMinimal, accessToken: string, opts?: Record<string, unknown>) => InferenceSnippet
+	>
+> = {
 	// Same order as in js/src/lib/interfaces/Types.ts
 	"text-classification": snippetBasic,
 	"token-classification": snippetBasic,
@@ -93,10 +138,10 @@ export const curlSnippets: Partial<Record<PipelineType, (model: ModelDataMinimal
 	"image-segmentation": snippetFile,
 };
 
-export function getCurlInferenceSnippet(model: ModelDataMinimal, accessToken: string): string {
+export function getCurlInferenceSnippet(model: ModelDataMinimal, accessToken: string): InferenceSnippet {
 	return model.pipeline_tag && model.pipeline_tag in curlSnippets
-		? curlSnippets[model.pipeline_tag]?.(model, accessToken) ?? ""
-		: "";
+		? curlSnippets[model.pipeline_tag]?.(model, accessToken) ?? { content: "" }
+		: { content: "" };
 }
 
 export function hasCurlInferenceSnippet(model: Pick<ModelDataMinimal, "pipeline_tag">): boolean {

package/src/snippets/js.ts

@@ -1,9 +1,11 @@
 import type { PipelineType } from "../pipelines.js";
+import type { ChatCompletionInputMessage, GenerationParameters } from "../tasks/index.js";
+import { stringifyGenerationConfig, stringifyMessages } from "./common.js";
 import { getModelInputSnippet } from "./inputs.js";
-import type { ModelDataMinimal } from "./types.js";
+import type { InferenceSnippet, ModelDataMinimal } from "./types.js";
 
-export const snippetBasic = (model: ModelDataMinimal, accessToken: string): string =>
-	`async function query(data) {
+export const snippetBasic = (model: ModelDataMinimal, accessToken: string): InferenceSnippet => ({
+	content: `async function query(data) {
 	const response = await fetch(
 		"https://api-inference.huggingface.co/models/${model.id}",
 		{
@@ -21,31 +23,136 @@ export const snippetBasic = (model: ModelDataMinimal, accessToken: string): stri
 
 query({"inputs": ${getModelInputSnippet(model)}}).then((response) => {
 	console.log(JSON.stringify(response));
-});`;
+});`,
+});
 
-export const snippetTextGeneration = (model: ModelDataMinimal, accessToken: string): string => {
+export const snippetTextGeneration = (
+	model: ModelDataMinimal,
+	accessToken: string,
+	opts?: {
+		streaming?: boolean;
+		messages?: ChatCompletionInputMessage[];
+		temperature?: GenerationParameters["temperature"];
+		max_tokens?: GenerationParameters["max_tokens"];
+		top_p?: GenerationParameters["top_p"];
+	}
+): InferenceSnippet | InferenceSnippet[] => {
 	if (model.tags.includes("conversational")) {
 		// Conversational model detected, so we display a code snippet that features the Messages API
-		return `import { HfInference } from "@huggingface/inference";
+		const streaming = opts?.streaming ?? true;
+		const messages: ChatCompletionInputMessage[] = opts?.messages ?? [
+			{ role: "user", content: "What is the capital of France?" },
+		];
+		const messagesStr = stringifyMessages(messages, { sep: ",\n\t\t", start: "[\n\t\t", end: "\n\t]" });
 
-const inference = new HfInference("${accessToken || `{API_TOKEN}`}");
+		const config = {
+			...(opts?.temperature ? { temperature: opts.temperature } : undefined),
+			max_tokens: opts?.max_tokens ?? 500,
+			...(opts?.top_p ? { top_p: opts.top_p } : undefined),
+		};
+		const configStr = stringifyGenerationConfig(config, {
+			sep: ",\n\t",
+			start: "",
+			end: "",
+			attributeValueConnector: ": ",
+		});
 
-for await (const chunk of inference.chatCompletionStream({
+		if (streaming) {
+			return [
+				{
+					client: "huggingface_hub",
+					content: `import { HfInference } from "@huggingface/inference"
+
+const client = new HfInference("${accessToken || `{API_TOKEN}`}")
+
+let out = "";
+
+const stream = client.chatCompletionStream({
 	model: "${model.id}",
-	messages: [{ role: "user", content: "What is the capital of France?" }],
-	max_tokens: 500,
-})) {
-	process.stdout.write(chunk.choices[0]?.delta?.content || "");
-}`;
+	messages: ${messagesStr},
+	${configStr}
+});
+
+for await (const chunk of stream) {
+	if (chunk.choices && chunk.choices.length > 0) {
+		const newContent = chunk.choices[0].delta.content;
+		out += newContent;
+		console.log(newContent);
+	}  
+}`,
+				},
+				{
+					client: "openai",
+					content: `import { OpenAI } from "openai"
+
+const client = new OpenAI({
+	baseURL: "https://api-inference.huggingface.co/v1/",
+    apiKey: "${accessToken || `{API_TOKEN}`}"
+})
+
+let out = "";
+
+const stream = await client.chat.completions.create({
+	model: "${model.id}",
+	messages: ${messagesStr},
+	${configStr},
+	stream: true,
+});
+
+for await (const chunk of stream) {
+	if (chunk.choices && chunk.choices.length > 0) {
+		const newContent = chunk.choices[0].delta.content;
+		out += newContent;
+		console.log(newContent);
+	}  
+}`,
+				},
+			];
+		} else {
+			return [
+				{
+					client: "huggingface_hub",
+					content: `import { HfInference } from '@huggingface/inference'
+
+const client = new HfInference("${accessToken || `{API_TOKEN}`}")
+
+const chatCompletion = await client.chatCompletion({
+	model: "${model.id}",
+	messages: ${messagesStr},
+	${configStr}
+});
+
+console.log(chatCompletion.choices[0].message);`,
+				},
+				{
+					client: "openai",
+					content: `import { OpenAI } from "openai"
+
+const client = new OpenAI({
+    baseURL: "https://api-inference.huggingface.co/v1/",
+    apiKey: "${accessToken || `{API_TOKEN}`}"
+})
+
+const chatCompletion = await client.chat.completions.create({
+	model: "${model.id}",
+	messages: ${messagesStr},
+	${configStr}
+});
+
+console.log(chatCompletion.choices[0].message);`,
+				},
+			];
+		}
 	} else {
 		return snippetBasic(model, accessToken);
 	}
 };
 
-export const snippetImageTextToTextGeneration = (model: ModelDataMinimal, accessToken: string): string => {
+export const snippetImageTextToTextGeneration = (model: ModelDataMinimal, accessToken: string): InferenceSnippet => {
 	if (model.tags.includes("conversational")) {
 		// Conversational model detected, so we display a code snippet that features the Messages API
-		return `import { HfInference } from "@huggingface/inference";
+		return {
+			content: `import { HfInference } from "@huggingface/inference";
 
 const inference = new HfInference("${accessToken || `{API_TOKEN}`}");
 const imageUrl = "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg";
@@ -64,14 +171,15 @@ for await (const chunk of inference.chatCompletionStream({
 	max_tokens: 500,
 })) {
 	process.stdout.write(chunk.choices[0]?.delta?.content || "");
-}`;
+}`,
+		};
 	} else {
 		return snippetBasic(model, accessToken);
 	}
 };
 
-export const snippetZeroShotClassification = (model: ModelDataMinimal, accessToken: string): string =>
-	`async function query(data) {
+export const snippetZeroShotClassification = (model: ModelDataMinimal, accessToken: string): InferenceSnippet => ({
+	content: `async function query(data) {
 	const response = await fetch(
 		"https://api-inference.huggingface.co/models/${model.id}",
 		{
@@ -91,10 +199,11 @@ query({"inputs": ${getModelInputSnippet(
 		model
 	)}, "parameters": {"candidate_labels": ["refund", "legal", "faq"]}}).then((response) => {
 	console.log(JSON.stringify(response));
-});`;
+});`,
+});
 
-export const snippetTextToImage = (model: ModelDataMinimal, accessToken: string): string =>
-	`async function query(data) {
+export const snippetTextToImage = (model: ModelDataMinimal, accessToken: string): InferenceSnippet => ({
+	content: `async function query(data) {
 	const response = await fetch(
 		"https://api-inference.huggingface.co/models/${model.id}",
 		{
@@ -111,9 +220,10 @@ export const snippetTextToImage = (model: ModelDataMinimal, accessToken: string)
 }
 query({"inputs": ${getModelInputSnippet(model)}}).then((response) => {
 	// Use image
-});`;
+});`,
+});
 
-export const snippetTextToAudio = (model: ModelDataMinimal, accessToken: string): string => {
+export const snippetTextToAudio = (model: ModelDataMinimal, accessToken: string): InferenceSnippet => {
 	const commonSnippet = `async function query(data) {
 		const response = await fetch(
 			"https://api-inference.huggingface.co/models/${model.id}",
@@ -127,33 +237,35 @@ export const snippetTextToAudio = (model: ModelDataMinimal, accessToken: string)
 			}
 		);`;
 	if (model.library_name === "transformers") {
-		return (
-			commonSnippet +
-			`
+		return {
+			content:
+				commonSnippet +
+				`
 			const result = await response.blob();
 			return result;
 		}
 		query({"inputs": ${getModelInputSnippet(model)}}).then((response) => {
 			// Returns a byte object of the Audio wavform. Use it directly!
-		});`
-		);
+		});`,
+		};
 	} else {
-		return (
-			commonSnippet +
-			`
+		return {
+			content:
+				commonSnippet +
+				`
 			const result = await response.json();
 			return result;
 		}
 		
 		query({"inputs": ${getModelInputSnippet(model)}}).then((response) => {
 			console.log(JSON.stringify(response));
-		});`
-		);
+		});`,
+		};
 	}
 };
 
-export const snippetFile = (model: ModelDataMinimal, accessToken: string): string =>
-	`async function query(filename) {
+export const snippetFile = (model: ModelDataMinimal, accessToken: string): InferenceSnippet => ({
+	content: `async function query(filename) {
 	const data = fs.readFileSync(filename);
 	const response = await fetch(
 		"https://api-inference.huggingface.co/models/${model.id}",
@@ -172,9 +284,19 @@ export const snippetFile = (model: ModelDataMinimal, accessToken: string): strin
 
 query(${getModelInputSnippet(model)}).then((response) => {
 	console.log(JSON.stringify(response));
-});`;
+});`,
+});
 
-export const jsSnippets: Partial<Record<PipelineType, (model: ModelDataMinimal, accessToken: string) => string>> = {
+export const jsSnippets: Partial<
+	Record<
+		PipelineType,
+		(
+			model: ModelDataMinimal,
+			accessToken: string,
+			opts?: Record<string, unknown>
+		) => InferenceSnippet | InferenceSnippet[]
+	>
+> = {
 	// Same order as in js/src/lib/interfaces/Types.ts
 	"text-classification": snippetBasic,
 	"token-classification": snippetBasic,
@@ -201,10 +323,13 @@ export const jsSnippets: Partial<Record<PipelineType, (model: ModelDataMinimal,
 	"image-segmentation": snippetFile,
 };
 
-export function getJsInferenceSnippet(model: ModelDataMinimal, accessToken: string): string {
+export function getJsInferenceSnippet(
+	model: ModelDataMinimal,
+	accessToken: string
+): InferenceSnippet | InferenceSnippet[] {
 	return model.pipeline_tag && model.pipeline_tag in jsSnippets
-		? jsSnippets[model.pipeline_tag]?.(model, accessToken) ?? ""
-		: "";
+		? jsSnippets[model.pipeline_tag]?.(model, accessToken) ?? { content: "" }
+		: { content: "" };
 }
 
 export function hasJsInferenceSnippet(model: ModelDataMinimal): boolean {