@huggingface/inference - npm Package Compare versions

Comparing version 2.6.7 to 2.7.0

dist/index.js

@@ -1,2 +0,1 @@
-/// <reference path="./index.d.ts" />
 var __defProp = Object.defineProperty;
@@ -14,2 +13,4 @@ var __export = (target, all) => {
   automaticSpeechRecognition: () => automaticSpeechRecognition,
+  chatCompletion: () => chatCompletion,
+  chatCompletionStream: () => chatCompletionStream,
   documentQuestionAnswering: () => documentQuestionAnswering,
@@ -43,2 +44,26 @@ featureExtraction: () => featureExtraction,
 
+// src/utils/pick.ts
+function pick(o, props) {
+  return Object.assign(
+    {},
+    ...props.map((prop) => {
+      if (o[prop] !== void 0) {
+        return { [prop]: o[prop] };
+      }
+    })
+  );
+}
+
+// src/utils/typedInclude.ts
+function typedInclude(arr, v) {
+  return arr.includes(v);
+}
+
+// src/utils/omit.ts
+function omit(o, props) {
+  const propsArr = Array.isArray(props) ? props : [props];
+  const letsKeep = Object.keys(o).filter((prop) => !typedInclude(propsArr, prop));
+  return pick(o, letsKeep);
+}
+
 // src/lib/isUrl.ts
@@ -84,3 +109,3 @@ function isUrl(modelOrUrl) {
 async function makeRequestOptions(args, options) {
-  const { accessToken, model: _model, ...otherArgs } = args;
+  const { accessToken, endpointUrl, ...otherArgs } = args;
   let { model } = args;
@@ -94,3 +119,3 @@ const {
     dont_load_model,
-    ...otherOptions
+    chatCompletion: chatCompletion2
   } = options ?? {};
@@ -129,6 +154,13 @@ const headers = {};
   }
-  const url = (() => {
+  let url = (() => {
+    if (endpointUrl && isUrl(model)) {
+      throw new TypeError("Both model and endpointUrl cannot be URLs");
+    }
     if (isUrl(model)) {
+      console.warn("Using a model URL is deprecated, please use the `endpointUrl` parameter instead");
       return model;
     }
+    if (endpointUrl) {
+      return endpointUrl;
+    }
     if (task) {
@@ -139,2 +171,5 @@ return `${HF_INFERENCE_API_BASE_URL}/pipeline/${task}/${model}`;
   })();
+  if (chatCompletion2 && !url.endsWith("/chat/completions")) {
+    url += "/v1/chat/completions";
+  }
   let credentials;
@@ -150,4 +185,3 @@ if (typeof includeCredentials === "string") {
     body: binary ? args.data : JSON.stringify({
-      ...otherArgs,
-      options: options && otherOptions
+      ...otherArgs.model && isUrl(otherArgs.model) ? omit(otherArgs, "model") : otherArgs
     }),
@@ -173,2 +207,5 @@ ...credentials && { credentials },
       const output = await response.json();
+      if ([400, 422, 404, 500].includes(response.status) && options?.chatCompletion) {
+        throw new Error(`Server ${args.model} does not seem to support chat completion. Error: ${output.error}`);
+      }
       if (output.error) {
@@ -298,2 +335,5 @@ throw new Error(output.error);
       const output = await response.json();
+      if ([400, 422, 404, 500].includes(response.status) && options?.chatCompletion) {
+        throw new Error(`Server ${args.model} does not seem to support chat completion. Error: ${output.error}`);
+      }
       if (output.error) {
@@ -335,2 +375,5 @@ throw new Error(output.error);
         if (event.data.length > 0) {
+          if (event.data === "[DONE]") {
+            return;
+          }
           const data = JSON.parse(event.data);
@@ -482,3 +525,3 @@ if (typeof data === "object" && data !== null && "error" in data) {
 
-// ../shared/src/base64FromBytes.ts
+// src/utils/base64FromBytes.ts
 function base64FromBytes(arr) {
@@ -496,6 +539,2 @@ if (globalThis.Buffer) {
 
-// ../shared/src/isBackend.ts
-var isBrowser = typeof window !== "undefined" && typeof window.document !== "undefined";
-var isWebWorker = typeof self === "object" && self.constructor && self.constructor.name === "DedicatedWorkerGlobalScope";
-
 // src/tasks/cv/imageToImage.ts
@@ -741,2 +780,25 @@ async function imageToImage(args, options) {
 
+// src/tasks/nlp/chatCompletion.ts
+async function chatCompletion(args, options) {
+  const res = await request(args, {
+    ...options,
+    taskHint: "text-generation",
+    chatCompletion: true
+  });
+  const isValidOutput = typeof res === "object" && Array.isArray(res?.choices) && typeof res?.created === "number" && typeof res?.id === "string" && typeof res?.model === "string" && typeof res?.system_fingerprint === "string" && typeof res?.usage === "object";
+  if (!isValidOutput) {
+    throw new InferenceOutputError("Expected ChatCompletionOutput");
+  }
+  return res;
+}
+
+// src/tasks/nlp/chatCompletionStream.ts
+async function* chatCompletionStream(args, options) {
+  yield* streamingRequest(args, {
+    ...options,
+    taskHint: "text-generation",
+    chatCompletion: true
+  });
+}
+
 // src/tasks/multimodal/documentQuestionAnswering.ts
@@ -853,3 +915,3 @@ async function documentQuestionAnswering(args, options) {
           // eslint-disable-next-line @typescript-eslint/no-explicit-any
-          fn({ ...params, accessToken, model: endpointUrl }, { ...defaultOptions, ...options })
+          fn({ ...params, accessToken, endpointUrl }, { ...defaultOptions, ...options })
         )
@@ -867,2 +929,4 @@ });
   automaticSpeechRecognition,
+  chatCompletion,
+  chatCompletionStream,
   documentQuestionAnswering,
@@ -869,0 +933,0 @@ featureExtraction,

package.json

 {
   "name": "@huggingface/inference",
-  "version": "2.6.7",
+  "version": "2.7.0",
   "packageManager": "pnpm@8.10.5",
@@ -32,7 +32,7 @@ "license": "MIT",
   "source": "src/index.ts",
-  "types": "./dist/index.d.ts",
+  "types": "./dist/src/index.d.ts",
   "main": "./dist/index.cjs",
   "module": "./dist/index.js",
   "exports": {
-    "types": "./dist/index.d.ts",
+    "types": "./dist/src/index.d.ts",
     "require": "./dist/index.cjs",
@@ -42,10 +42,12 @@ "import": "./dist/index.js"
   "type": "module",
+  "dependencies": {
+    "@huggingface/tasks": "^0.10.0"
+  },
   "devDependencies": {
-    "@types/node": "18.13.0",
-    "@huggingface/tasks": "^0.8.0"
+    "@types/node": "18.13.0"
   },
   "resolutions": {},
   "scripts": {
-    "build": "tsup src/index.ts --format cjs,esm --clean && pnpm run dts",
-    "dts": "tsx scripts/generate-dts.ts",
+    "build": "tsup src/index.ts --format cjs,esm --clean && tsc --emitDeclarationOnly --declaration",
+    "dts": "tsx scripts/generate-dts.ts && tsc --noEmit dist/index.d.ts",
     "lint": "eslint --quiet --fix --ext .cjs,.ts .",
@@ -52,0 +54,0 @@ "lint:check": "eslint --ext .cjs,.ts .",

README.md

@@ -8,3 +8,3 @@ # 🤗 Hugging Face Inference Endpoints
 
-You can also try out a live [interactive notebook](https://observablehq.com/@huggingface/hello-huggingface-js-inference), see some demos on [hf.co/huggingfacejs](https://huggingface.co/huggingfacejs), or watch a [Scrimba tutorial that explains how Inference Endpoints works](https://scrimba.com/scrim/cod8248f5adfd6e129582c523). 
+You can also try out a live [interactive notebook](https://observablehq.com/@huggingface/hello-huggingface-js-inference), see some demos on [hf.co/huggingfacejs](https://huggingface.co/huggingfacejs), or watch a [Scrimba tutorial that explains how Inference Endpoints works](https://scrimba.com/scrim/cod8248f5adfd6e129582c523).
 
@@ -34,3 +34,2 @@ ## Getting Started
 
-
 ### Initialize
@@ -48,3 +47,2 @@
 
-
 #### Tree-shaking
@@ -69,2 +67,81 @@
 
+### Text Generation
+
+Generates text from an input prompt.
+
+[Demo](https://huggingface.co/spaces/huggingfacejs/streaming-text-generation)
+
+```typescript
+await hf.textGeneration({
+  model: 'gpt2',
+  inputs: 'The answer to the universe is'
+})
+
+for await (const output of hf.textGenerationStream({
+  model: "google/flan-t5-xxl",
+  inputs: 'repeat "one two three four"',
+  parameters: { max_new_tokens: 250 }
+})) {
+  console.log(output.token.text, output.generated_text);
+}
+```
+
+### Text Generation (Chat Completion API Compatible)
+
+Using the `chatCompletion` method, you can generate text with models compatible with the OpenAI Chat Completion API. All models served by [TGI](https://api-inference.huggingface.co/framework/text-generation-inference) on Hugging Face support Messages API.
+
+[Demo](https://huggingface.co/spaces/huggingfacejs/streaming-chat-completion)
+
+```typescript
+// Non-streaming API
+const out = await hf.chatCompletion({
+  model: "mistralai/Mistral-7B-Instruct-v0.2",
+  messages: [{ role: "user", content: "Complete the this sentence with words one plus one is equal " }],
+  max_tokens: 500,
+  temperature: 0.1,
+  seed: 0,
+});
+
+// Streaming API
+let out = "";
+for await (const chunk of hf.chatCompletionStream({
+  model: "mistralai/Mistral-7B-Instruct-v0.2",
+  messages: [
+    { role: "user", content: "Complete the equation 1+1= ,just the answer" },
+  ],
+  max_tokens: 500,
+  temperature: 0.1,
+  seed: 0,
+})) {
+  if (chunk.choices && chunk.choices.length > 0) {
+    out += chunk.choices[0].delta.content;
+  }
+}
+```
+
+It's also possible to call Mistral or OpenAI endpoints directly:
+
+```typescript
+const openai = new HfInference(OPENAI_TOKEN).endpoint("https://api.openai.com");
+
+let out = "";
+for await (const chunk of openai.chatCompletionStream({
+  model: "gpt-3.5-turbo",
+  messages: [
+    { role: "user", content: "Complete the equation 1+1= ,just the answer" },
+  ],
+  max_tokens: 500,
+  temperature: 0.1,
+  seed: 0,
+})) {
+  if (chunk.choices && chunk.choices.length > 0) {
+    out += chunk.choices[0].delta.content;
+  }
+}
+
+// For mistral AI:
+// endpointUrl: "https://api.mistral.ai"
+// model: "mistral-tiny"
+```
+
 ### Fill Mask
@@ -138,23 +215,2 @@
 
-### Text Generation
-
-Generates text from an input prompt.
-
-[Demo](https://huggingface.co/spaces/huggingfacejs/streaming-text-generation)
-
-```typescript
-await hf.textGeneration({
-  model: 'gpt2',
-  inputs: 'The answer to the universe is'
-})
-
-for await (const output of hf.textGenerationStream({
-  model: "google/flan-t5-xxl",
-  inputs: 'repeat "one two three four"',
-  parameters: { max_new_tokens: 250 }
-})) {
-  console.log(output.token.text, output.generated_text);
-}
-```
-
 ### Token Classification
@@ -185,5 +241,5 @@
   parameters: {
-		"src_lang": "en_XX",
-		"tgt_lang": "fr_XX"
-	}
+  "src_lang": "en_XX",
+  "tgt_lang": "fr_XX"
+ }
 })
@@ -506,2 +562,22 @@ ```
 
+You can use any Chat Completion API-compatible provider with the `chatCompletion` method.
+
+```typescript
+// Chat Completion Example
+const MISTRAL_KEY = process.env.MISTRAL_KEY;
+const hf = new HfInference(MISTRAL_KEY);
+const ep = hf.endpoint("https://api.mistral.ai");
+const stream = ep.chatCompletionStream({
+  model: "mistral-tiny",
+  messages: [{ role: "user", content: "Complete the equation one + one = , just the answer" }],
+});
+let out = "";
+for await (const chunk of stream) {
+  if (chunk.choices && chunk.choices.length > 0) {
+    out += chunk.choices[0].delta.content;
+    console.log(out);
+  }
+}
+```
+
 ## Custom Inference Endpoints
@@ -514,2 +590,21 @@
 const { generated_text } = await gpt2.textGeneration({inputs: 'The answer to the universe is'});
+
+// Chat Completion Example
+const ep = hf.endpoint(
+  "https://api-inference.huggingface.co/models/mistralai/Mistral-7B-Instruct-v0.2"
+);
+const stream = ep.chatCompletionStream({
+  model: "tgi",
+  messages: [{ role: "user", content: "Complete the equation 1+1= ,just the answer" }],
+  max_tokens: 500,
+  temperature: 0.1,
+  seed: 0,
+});
+let out = "";
+for await (const chunk of stream) {
+  if (chunk.choices && chunk.choices.length > 0) {
+    out += chunk.choices[0].delta.content;
+    console.log(out);
+  }
+}
 ```
@@ -543,1 +638,5 @@
 It also contains demos, example outputs, and other resources should you want to dig deeper into the ML side of things.
+
+## Dependencies
+
+- `@huggingface/tasks` : Typings only

src/HfInference.ts

@@ -17,5 +17,5 @@ import * as tasks from "./tasks";
 
-type TaskWithNoAccessTokenNoModel = {
+type TaskWithNoAccessTokenNoEndpointUrl = {
 	[key in keyof Task]: (
-		args: DistributiveOmit<Parameters<Task[key]>[0], "accessToken" | "model">,
+		args: DistributiveOmit<Parameters<Task[key]>[0], "accessToken" | "endpointUrl">,
 		options?: Parameters<Task[key]>[1]
@@ -61,3 +61,3 @@ ) => ReturnType<Task[key]>;
 					// eslint-disable-next-line @typescript-eslint/no-explicit-any
-					fn({ ...params, accessToken, model: endpointUrl } as any, { ...defaultOptions, ...options }),
+					fn({ ...params, accessToken, endpointUrl } as any, { ...defaultOptions, ...options }),
 			});
@@ -70,2 +70,2 @@ }
 
-export interface HfInferenceEndpoint extends TaskWithNoAccessTokenNoModel {}
+export interface HfInferenceEndpoint extends TaskWithNoAccessTokenNoEndpointUrl {}

src/lib/makeRequestOptions.ts

 import type { InferenceTask, Options, RequestArgs } from "../types";
+import { omit } from "../utils/omit";
 import { HF_HUB_URL } from "./getDefaultTask";
@@ -25,6 +26,6 @@ import { isUrl } from "./isUrl";
 		taskHint?: InferenceTask;
+		chatCompletion?: boolean;
 	}
 ): Promise<{ url: string; info: RequestInit }> {
-	// eslint-disable-next-line @typescript-eslint/no-unused-vars
-	const { accessToken, model: _model, ...otherArgs } = args;
+	const { accessToken, endpointUrl, ...otherArgs } = args;
 	let { model } = args;
@@ -38,3 +39,3 @@ const {
 		dont_load_model,
-		...otherOptions
+		chatCompletion,
 	} = options ?? {};
@@ -82,7 +83,13 @@
 
-	const url = (() => {
+	let url = (() => {
+		if (endpointUrl && isUrl(model)) {
+			throw new TypeError("Both model and endpointUrl cannot be URLs");
+		}
 		if (isUrl(model)) {
+			console.warn("Using a model URL is deprecated, please use the `endpointUrl` parameter instead");
 			return model;
 		}
-
+		if (endpointUrl) {
+			return endpointUrl;
+		}
 		if (task) {
@@ -95,2 +102,6 @@ return `${HF_INFERENCE_API_BASE_URL}/pipeline/${task}/${model}`;
 
+	if (chatCompletion && !url.endsWith("/chat/completions")) {
+		url += "/v1/chat/completions";
+	}
+
 	/**
@@ -112,4 +123,3 @@ * For edge runtimes, leave 'credentials' undefined, otherwise cloudflare workers will error
 			: JSON.stringify({
-					...otherArgs,
-					options: options && otherOptions,
+					...(otherArgs.model && isUrl(otherArgs.model) ? omit(otherArgs, "model") : otherArgs),
 			  }),
@@ -116,0 +126,0 @@ ...(credentials && { credentials }),

src/tasks/custom/request.ts

@@ -14,2 +14,4 @@ import type { InferenceTask, Options, RequestArgs } from "../../types";
 		taskHint?: InferenceTask;
+		/** Is chat completion compatible */
+		chatCompletion?: boolean;
 	}
@@ -30,2 +32,5 @@ ): Promise<T> {
 			const output = await response.json();
+			if ([400, 422, 404, 500].includes(response.status) && options?.chatCompletion) {
+				throw new Error(`Server ${args.model} does not seem to support chat completion. Error: ${output.error}`);
+			}
 			if (output.error) {
@@ -32,0 +37,0 @@ throw new Error(output.error);

src/tasks/custom/streamingRequest.ts

@@ -16,2 +16,4 @@ import type { InferenceTask, Options, RequestArgs } from "../../types";
 		taskHint?: InferenceTask;
+		/** Is chat completion compatible */
+		chatCompletion?: boolean;
 	}
@@ -31,2 +33,5 @@ ): AsyncGenerator<T> {
 			const output = await response.json();
+			if ([400, 422, 404, 500].includes(response.status) && options?.chatCompletion) {
+				throw new Error(`Server ${args.model} does not seem to support chat completion. Error: ${output.error}`);
+			}
 			if (output.error) {
@@ -72,2 +77,5 @@ throw new Error(output.error);
 				if (event.data.length > 0) {
+					if (event.data === "[DONE]") {
+						return;
+					}
 					const data = JSON.parse(event.data);
@@ -74,0 +82,0 @@ if (typeof data === "object" && data !== null && "error" in data) {

src/tasks/cv/imageToImage.ts

 import { InferenceOutputError } from "../../lib/InferenceOutputError";
 import type { BaseArgs, Options, RequestArgs } from "../../types";
+import { base64FromBytes } from "../../utils/base64FromBytes";
 import { request } from "../custom/request";
-import { base64FromBytes } from "../../../../shared";
 
@@ -6,0 +6,0 @@ export type ImageToImageArgs = BaseArgs & {

src/tasks/cv/zeroShotImageClassification.ts

@@ -5,3 +5,3 @@ import { InferenceOutputError } from "../../lib/InferenceOutputError";
 import type { RequestArgs } from "../../types";
-import { base64FromBytes } from "../../../../shared";
+import { base64FromBytes } from "../../utils/base64FromBytes";
 
@@ -8,0 +8,0 @@ export type ZeroShotImageClassificationArgs = BaseArgs & {

src/tasks/index.ts

@@ -33,2 +33,4 @@ // Custom tasks with arbitrary inputs and outputs
 export * from "./nlp/zeroShotClassification";
+export * from "./nlp/chatCompletion";
+export * from "./nlp/chatCompletionStream";
 
@@ -35,0 +37,0 @@ // Multimodal tasks

src/tasks/multimodal/documentQuestionAnswering.ts

@@ -5,4 +5,4 @@ import { InferenceOutputError } from "../../lib/InferenceOutputError";
 import type { RequestArgs } from "../../types";
-import { base64FromBytes } from "../../../../shared";
 import { toArray } from "../../utils/toArray";
+import { base64FromBytes } from "../../utils/base64FromBytes";
 
@@ -9,0 +9,0 @@ export type DocumentQuestionAnsweringArgs = BaseArgs & {

src/tasks/multimodal/visualQuestionAnswering.ts

 import { InferenceOutputError } from "../../lib/InferenceOutputError";
 import type { BaseArgs, Options, RequestArgs } from "../../types";
+import { base64FromBytes } from "../../utils/base64FromBytes";
 import { request } from "../custom/request";
-import { base64FromBytes } from "../../../../shared";
 
@@ -6,0 +6,0 @@ export type VisualQuestionAnsweringArgs = BaseArgs & {

src/tasks/nlp/textGeneration.ts

@@ -0,1 +1,2 @@
+import type { TextGenerationInput, TextGenerationOutput } from "@huggingface/tasks";
 import { InferenceOutputError } from "../../lib/InferenceOutputError";
@@ -5,206 +6,5 @@ import type { BaseArgs, Options } from "../../types";
 
-/**
- * Inputs for Text Generation inference
- */
-export interface TextGenerationInput {
-	/**
-	 * The text to initialize generation with
-	 */
-	inputs: string;
-	/**
-	 * Additional inference parameters
-	 */
-	parameters?: TextGenerationParameters;
-	/**
-	 * Whether to stream output tokens
-	 */
-	stream?: boolean;
-	[property: string]: unknown;
-}
+export type { TextGenerationInput, TextGenerationOutput };
 
 /**
- * Additional inference parameters
- *
- * Additional inference parameters for Text Generation
- */
-export interface TextGenerationParameters {
-	/**
-	 * The number of sampling queries to run. Only the best one (in terms of total logprob) will
-	 * be returned.
-	 */
-	best_of?: number;
-	/**
-	 * Whether or not to output decoder input details
-	 */
-	decoder_input_details?: boolean;
-	/**
-	 * Whether or not to output details
-	 */
-	details?: boolean;
-	/**
-	 * Whether to use logits sampling instead of greedy decoding when generating new tokens.
-	 */
-	do_sample?: boolean;
-	/**
-	 * The maximum number of tokens to generate.
-	 */
-	max_new_tokens?: number;
-	/**
-	 * The parameter for repetition penalty. A value of 1.0 means no penalty. See [this
-	 * paper](https://hf.co/papers/1909.05858) for more details.
-	 */
-	repetition_penalty?: number;
-	/**
-	 * Whether to prepend the prompt to the generated text.
-	 */
-	return_full_text?: boolean;
-	/**
-	 * The random sampling seed.
-	 */
-	seed?: number;
-	/**
-	 * Stop generating tokens if a member of `stop_sequences` is generated.
-	 */
-	stop_sequences?: string[];
-	/**
-	 * The value used to modulate the logits distribution.
-	 */
-	temperature?: number;
-	/**
-	 * The number of highest probability vocabulary tokens to keep for top-k-filtering.
-	 */
-	top_k?: number;
-	/**
-	 * If set to < 1, only the smallest set of most probable tokens with probabilities that add
-	 * up to `top_p` or higher are kept for generation.
-	 */
-	top_p?: number;
-	/**
-	 * Truncate input tokens to the given size.
-	 */
-	truncate?: number;
-	/**
-	 * Typical Decoding mass. See [Typical Decoding for Natural Language
-	 * Generation](https://hf.co/papers/2202.00666) for more information
-	 */
-	typical_p?: number;
-	/**
-	 * Watermarking with [A Watermark for Large Language Models](https://hf.co/papers/2301.10226)
-	 */
-	watermark?: boolean;
-	[property: string]: unknown;
-}
-
-/**
- * Outputs for Text Generation inference
- */
-export interface TextGenerationOutput {
-	/**
-	 * When enabled, details about the generation
-	 */
-	details?: TextGenerationOutputDetails;
-	/**
-	 * The generated text
-	 */
-	generated_text: string;
-	[property: string]: unknown;
-}
-
-/**
- * When enabled, details about the generation
- */
-export interface TextGenerationOutputDetails {
-	/**
-	 * Details about additional sequences when best_of is provided
-	 */
-	best_of_sequences?: TextGenerationOutputSequenceDetails[];
-	/**
-	 * The reason why the generation was stopped.
-	 */
-	finish_reason: TextGenerationFinishReason;
-	/**
-	 * The number of generated tokens
-	 */
-	generated_tokens: number;
-	prefill: TextGenerationPrefillToken[];
-	/**
-	 * The random seed used for generation
-	 */
-	seed?: number;
-	/**
-	 * The generated tokens and associated details
-	 */
-	tokens: TextGenerationOutputToken[];
-	/**
-	 * Most likely tokens
-	 */
-	top_tokens?: Array<TextGenerationOutputToken[]>;
-	[property: string]: unknown;
-}
-
-export interface TextGenerationOutputSequenceDetails {
-	finish_reason: TextGenerationFinishReason;
-	/**
-	 * The generated text
-	 */
-	generated_text: string;
-	/**
-	 * The number of generated tokens
-	 */
-	generated_tokens: number;
-	prefill: TextGenerationPrefillToken[];
-	/**
-	 * The random seed used for generation
-	 */
-	seed?: number;
-	/**
-	 * The generated tokens and associated details
-	 */
-	tokens: TextGenerationOutputToken[];
-	/**
-	 * Most likely tokens
-	 */
-	top_tokens?: Array<TextGenerationOutputToken[]>;
-	[property: string]: unknown;
-}
-
-export interface TextGenerationPrefillToken {
-	id: number;
-	logprob: number;
-	/**
-	 * The text associated with that token
-	 */
-	text: string;
-	[property: string]: unknown;
-}
-
-/**
- * Generated token.
- */
-export interface TextGenerationOutputToken {
-	id: number;
-	logprob?: number;
-	/**
-	 * Whether or not that token is a special one
-	 */
-	special: boolean;
-	/**
-	 * The text associated with that token
-	 */
-	text: string;
-	[property: string]: unknown;
-}
-
-/**
- * The reason why the generation was stopped.
- *
- * length: The generated sequence reached the maximum allowed length
- *
- * eos_token: The model generated an end-of-sentence (EOS) token
- *
- * stop_sequence: One of the sequence in stop_sequences was generated
- */
-export type TextGenerationFinishReason = "length" | "eos_token" | "stop_sequence";
-
-/**
  * Use to continue text from a prompt. This is a very generic task. Recommended model: gpt2 (it’s a simple model, but fun to play with).
@@ -211,0 +11,0 @@ */

src/tasks/nlp/textGenerationStream.ts

@@ -0,4 +1,4 @@
+import type { TextGenerationInput } from "@huggingface/tasks";
 import type { BaseArgs, Options } from "../../types";
 import { streamingRequest } from "../custom/streamingRequest";
-import type { TextGenerationInput } from "./textGeneration";
 
@@ -70,2 +70,3 @@ export interface TextGenerationStreamToken {
 export interface TextGenerationStreamOutput {
+	index?: number;
 	/** Generated token, one at a time */
@@ -72,0 +73,0 @@ token: TextGenerationStreamToken;

src/types.ts

 import type { PipelineType } from "@huggingface/tasks";
+import type { ChatCompletionInput } from "@huggingface/tasks";
 
@@ -35,3 +36,3 @@ export interface Options {
 	/**
-	 * Credentials to use for the request. If this is a string, it will be passed straight on. If it's a boolean, true will be "include" and false will not send credentials at all (which defaults to "same-origin" inside browsers).
+	 * (Default: "same-origin"). String | Boolean. Credentials to use for the request. If this is a string, it will be passed straight on. If it's a boolean, true will be "include" and false will not send credentials at all.
 	 */
@@ -51,13 +52,23 @@ includeCredentials?: string | boolean;
 	/**
-	 * The model to use. Can be a full URL for a dedicated inference endpoint.
+	 * The model to use.
 	 *
 	 * If not specified, will call huggingface.co/api/tasks to get the default model for the task.
+	 *
+	 * /!\ Legacy behavior allows this to be an URL, but this is deprecated and will be removed in the future.
+	 * Use the `endpointUrl` parameter instead.
 	 */
 	model?: string;
+
+	/**
+	 * The URL of the endpoint to use. If not specified, will call huggingface.co/api/tasks to get the default endpoint for the task.
+	 *
+	 * If specified, will use this URL instead of the default one.
+	 */
+	endpointUrl?: string;
 }
 
 export type RequestArgs = BaseArgs &
-	({ data: Blob | ArrayBuffer } | { inputs: unknown }) & {
+	({ data: Blob | ArrayBuffer } | { inputs: unknown } | ChatCompletionInput) & {
 		parameters?: Record<string, unknown>;
 		accessToken?: string;
 	};

No alert changes

Improved metrics

Total package byte prevSize

234974

increased by12.45%

Dev dependency count

1

decreased by-50%

Number of package files

160

increased by220%

Number of lines in readme file

634

increased by18.5%

Worsened metrics

Dependency count

1

increased byInfinity%

Lines of code

5283

decreased by-7.78%

Dependency changes

• + Added@huggingface/tasks@^0.10.0

• + Added@huggingface/tasks@0.10.22(transitive)