@huggingface/inference 3.5.0 → 3.5.2

package/README.md

@@ -1,7 +1,7 @@
 # 🤗 Hugging Face Inference
 
-A Typescript powered wrapper for the HF Inference API (serverless), Inference Endpoints (dedicated), and third-party Inference Providers.
-It works with [Inference API (serverless)](https://huggingface.co/docs/api-inference/index) and [Inference Endpoints (dedicated)](https://huggingface.co/docs/inference-endpoints/index), and even with supported third-party Inference Providers.
+A Typescript powered wrapper for the HF Inference API (serverless), Inference Endpoints (dedicated), and all supported Inference Providers.
+It works with [Inference API (serverless)](https://huggingface.co/docs/api-inference/index) and [Inference Endpoints (dedicated)](https://huggingface.co/docs/inference-endpoints/index), and even with all supported third-party Inference Providers.
 
 Check out the [full documentation](https://huggingface.co/docs/huggingface.js/inference/README).
 
@@ -25,24 +25,24 @@ yarn add @huggingface/inference
 
 ```ts
 // esm.sh
-import { HfInference } from "https://esm.sh/@huggingface/inference"
+import { InferenceClient } from "https://esm.sh/@huggingface/inference"
 // or npm:
-import { HfInference } from "npm:@huggingface/inference"
+import { InferenceClient } from "npm:@huggingface/inference"
 ```
 
 ### Initialize
 
 ```typescript
-import { HfInference } from '@huggingface/inference'
+import { InferenceClient } from '@huggingface/inference'
 
-const hf = new HfInference('your access token')
+const hf = new InferenceClient('your access token')
 ```
 
 ❗**Important note:** Using an access token is optional to get started, however you will be rate limited eventually. Join [Hugging Face](https://huggingface.co/join) and then visit [access tokens](https://huggingface.co/settings/tokens) to generate your access token for **free**.
 
 Your access token should be kept private. If you need to protect it in front-end applications, we suggest setting up a proxy server that stores the access token.
 
-### Third-party inference providers
+### All supported inference providers
 
 You can send inference requests to third-party providers with the inference client.
 
@@ -63,7 +63,7 @@ To send requests to a third-party provider, you have to pass the `provider` para
 ```ts
 const accessToken = "hf_..."; // Either a HF access token, or an API key from the third-party provider (Replicate in this example)
 
-const client = new HfInference(accessToken);
+const client = new InferenceClient(accessToken);
 await client.textToImage({
   provider: "replicate",
   model:"black-forest-labs/Flux.1-dev",
@@ -93,7 +93,7 @@ This is not an issue for LLMs as everyone converged on the OpenAI API anyways, b
 
 ### Tree-shaking
 
-You can import the functions you need directly from the module instead of using the `HfInference` class.
+You can import the functions you need directly from the module instead of using the `InferenceClient` class.
 
 ```ts
 import { textGeneration } from "@huggingface/inference";
@@ -165,7 +165,7 @@ for await (const chunk of hf.chatCompletionStream({
 It's also possible to call Mistral or OpenAI endpoints directly:
 
 ```typescript
-const openai = new HfInference(OPENAI_TOKEN).endpoint("https://api.openai.com");
+const openai = new InferenceClient(OPENAI_TOKEN).endpoint("https://api.openai.com");
 
 let out = "";
 for await (const chunk of openai.chatCompletionStream({
@@ -602,7 +602,7 @@ You can use any Chat Completion API-compatible provider with the `chatCompletion
 ```typescript
 // Chat Completion Example
 const MISTRAL_KEY = process.env.MISTRAL_KEY;
-const hf = new HfInference(MISTRAL_KEY);
+const hf = new InferenceClient(MISTRAL_KEY);
 const ep = hf.endpoint("https://api.mistral.ai");
 const stream = ep.chatCompletionStream({
   model: "mistral-tiny",

package/dist/index.cjs

@@ -21,8 +21,9 @@ var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: tru
 var src_exports = {};
 __export(src_exports, {
   HfInference: () => HfInference,
-  HfInferenceEndpoint: () => HfInferenceEndpoint,
   INFERENCE_PROVIDERS: () => INFERENCE_PROVIDERS,
+  InferenceClient: () => InferenceClient,
+  InferenceClientEndpoint: () => InferenceClientEndpoint,
   InferenceOutputError: () => InferenceOutputError,
   audioClassification: () => audioClassification,
   audioToAudio: () => audioToAudio,
@@ -102,7 +103,7 @@ var HF_HUB_URL = "https://huggingface.co";
 var HF_ROUTER_URL = "https://router.huggingface.co";
 
 // src/providers/black-forest-labs.ts
-var BLACK_FOREST_LABS_AI_API_BASE_URL = "https://api.us1.bfl.ai/v1";
+var BLACK_FOREST_LABS_AI_API_BASE_URL = "https://api.us1.bfl.ai";
 var makeBody = (params) => {
   return params.args;
 };
@@ -114,7 +115,7 @@ var makeHeaders = (params) => {
   }
 };
 var makeUrl = (params) => {
-  return `${params.baseUrl}/${params.model}`;
+  return `${params.baseUrl}/v1/${params.model}`;
 };
 var BLACK_FOREST_LABS_CONFIG = {
   baseUrl: BLACK_FOREST_LABS_AI_API_BASE_URL,
@@ -186,7 +187,7 @@ var FAL_AI_CONFIG = {
 };
 
 // src/providers/fireworks-ai.ts
-var FIREWORKS_AI_API_BASE_URL = "https://api.fireworks.ai/inference";
+var FIREWORKS_AI_API_BASE_URL = "https://api.fireworks.ai";
 var makeBody5 = (params) => {
   return {
     ...params.args,
@@ -198,9 +199,9 @@ var makeHeaders5 = (params) => {
 };
 var makeUrl5 = (params) => {
   if (params.task === "text-generation" && params.chatCompletion) {
-    return `${params.baseUrl}/v1/chat/completions`;
+    return `${params.baseUrl}/inference/v1/chat/completions`;
   }
-  return params.baseUrl;
+  return `${params.baseUrl}/inference`;
 };
 var FIREWORKS_AI_CONFIG = {
   baseUrl: FIREWORKS_AI_API_BASE_URL,
@@ -325,7 +326,7 @@ var makeBody10 = (params) => {
   };
 };
 var makeHeaders10 = (params) => {
-  return { Authorization: `Bearer ${params.accessToken}` };
+  return { Authorization: `Bearer ${params.accessToken}`, Prefer: "wait" };
 };
 var makeUrl10 = (params) => {
   if (params.model.includes(":")) {
@@ -429,7 +430,7 @@ function isUrl(modelOrUrl) {
 
 // package.json
 var name = "@huggingface/inference";
-var version = "3.5.0";
+var version = "3.5.2";
 
 // src/providers/consts.ts
 var HARDCODED_MODEL_ID_MAPPING = {
@@ -1582,8 +1583,8 @@ async function tabularClassification(args, options) {
   return res;
 }
 
-// src/HfInference.ts
-var HfInference = class {
+// src/InferenceClient.ts
+var InferenceClient = class {
   accessToken;
   defaultOptions;
   constructor(accessToken = "", defaultOptions = {}) {
@@ -1600,13 +1601,13 @@ var HfInference = class {
     }
   }
   /**
-   * Returns copy of HfInference tied to a specified endpoint.
+   * Returns copy of InferenceClient tied to a specified endpoint.
    */
   endpoint(endpointUrl) {
-    return new HfInferenceEndpoint(endpointUrl, this.accessToken, this.defaultOptions);
+    return new InferenceClientEndpoint(endpointUrl, this.accessToken, this.defaultOptions);
   }
 };
-var HfInferenceEndpoint = class {
+var InferenceClientEndpoint = class {
   constructor(endpointUrl, accessToken = "", defaultOptions = {}) {
     accessToken;
     defaultOptions;
@@ -1621,6 +1622,8 @@ var HfInferenceEndpoint = class {
     }
   }
 };
+var HfInference = class extends InferenceClient {
+};
 
 // src/types.ts
 var INFERENCE_PROVIDERS = [
@@ -1774,18 +1777,7 @@ function getCurlInferenceSnippet(model, accessToken, provider, providerModelId,
 // src/snippets/python.ts
 var python_exports = {};
 __export(python_exports, {
-  getPythonInferenceSnippet: () => getPythonInferenceSnippet,
-  pythonSnippets: () => pythonSnippets,
-  snippetBasic: () => snippetBasic2,
-  snippetConversational: () => snippetConversational,
-  snippetDocumentQuestionAnswering: () => snippetDocumentQuestionAnswering,
-  snippetFile: () => snippetFile2,
-  snippetTabular: () => snippetTabular,
-  snippetTextToAudio: () => snippetTextToAudio,
-  snippetTextToImage: () => snippetTextToImage,
-  snippetTextToVideo: () => snippetTextToVideo,
-  snippetZeroShotClassification: () => snippetZeroShotClassification2,
-  snippetZeroShotImageClassification: () => snippetZeroShotImageClassification
+  getPythonInferenceSnippet: () => getPythonInferenceSnippet
 });
 var import_tasks3 = require("@huggingface/tasks");
 var import_tasks4 = require("@huggingface/tasks");
@@ -1821,8 +1813,8 @@ var HFH_INFERENCE_CLIENT_METHODS = {
 var snippetImportInferenceClient = (accessToken, provider) => `from huggingface_hub import InferenceClient
 
 client = InferenceClient(
-	provider="${provider}",
-	api_key="${accessToken || "{API_TOKEN}"}"
+    provider="${provider}",
+    api_key="${accessToken || "{API_TOKEN}"}",
 )`;
 var snippetConversational = (model, accessToken, provider, providerModelId, opts) => {
   const streaming = opts?.streaming ?? true;
@@ -1850,7 +1842,7 @@ stream = client.chat.completions.create(
 	model="${model.id}", 
 	messages=messages, 
 	${configStr}
-	stream=True
+	stream=True,
 )
 
 for chunk in stream:
@@ -1960,9 +1952,8 @@ var snippetBasic2 = (model, accessToken, provider) => {
         content: `${snippetImportInferenceClient(accessToken, provider)}
 
 result = client.${HFH_INFERENCE_CLIENT_METHODS[model.pipeline_tag]}(
-	model="${model.id}",
 	inputs=${(0, import_tasks4.getModelInputSnippet)(model)},
-	provider="${provider}",
+	model="${model.id}",
 )
 
 print(result)
@@ -2004,7 +1995,7 @@ var snippetTextToImage = (model, accessToken, provider, providerModelId) => {
 # output is a PIL.Image object
 image = client.text_to_image(
 	${(0, import_tasks4.getModelInputSnippet)(model)},
-	model="${model.id}"
+	model="${model.id}",
 )`
     },
     ...provider === "fal-ai" ? [
@@ -2049,7 +2040,7 @@ var snippetTextToVideo = (model, accessToken, provider) => {
 
 video = client.text_to_video(
 	${(0, import_tasks4.getModelInputSnippet)(model)},
-	model="${model.id}"
+	model="${model.id}",
 )`
     }
   ] : [];
@@ -2103,23 +2094,79 @@ Audio(audio, rate=sampling_rate)`
     ];
   }
 };
-var snippetDocumentQuestionAnswering = (model) => {
+var snippetAutomaticSpeechRecognition = (model, accessToken, provider) => {
   return [
+    {
+      client: "huggingface_hub",
+      content: `${snippetImportInferenceClient(accessToken, provider)}
+output = client.automatic_speech_recognition(${(0, import_tasks4.getModelInputSnippet)(model)}, model="${model.id}")`
+    },
+    snippetFile2(model)[0]
+  ];
+};
+var snippetDocumentQuestionAnswering = (model, accessToken, provider) => {
+  const inputsAsStr = (0, import_tasks4.getModelInputSnippet)(model);
+  const inputsAsObj = JSON.parse(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")  
+		payload["image"] = base64.b64encode(img).decode("utf-8")
 	response = requests.post(API_URL, headers=headers, json=payload)
 	return response.json()
 
 output = query({
-    "inputs": ${(0, import_tasks4.getModelInputSnippet)(model)},
+    "inputs": ${inputsAsStr},
 })`
     }
   ];
 };
+var snippetImageToImage = (model, accessToken, provider) => {
+  const inputsAsStr = (0, import_tasks4.getModelInputSnippet)(model);
+  const inputsAsObj = JSON.parse(inputsAsStr);
+  return [
+    {
+      client: "huggingface_hub",
+      content: `${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": "${inputsAsObj.prompt}"},
+})
+
+# You can access the image with PIL.Image for example
+import io
+from PIL import Image
+image = Image.open(io.BytesIO(image_bytes))`
+    }
+  ];
+};
 var pythonSnippets = {
   // Same order as in tasks/src/pipelines.ts
   "text-classification": snippetBasic2,
@@ -2135,7 +2182,7 @@ var pythonSnippets = {
   "image-text-to-text": snippetConversational,
   "fill-mask": snippetBasic2,
   "sentence-similarity": snippetBasic2,
-  "automatic-speech-recognition": snippetFile2,
+  "automatic-speech-recognition": snippetAutomaticSpeechRecognition,
   "text-to-image": snippetTextToImage,
   "text-to-video": snippetTextToVideo,
   "text-to-speech": snippetTextToAudio,
@@ -2149,6 +2196,7 @@ var pythonSnippets = {
   "image-segmentation": snippetFile2,
   "document-question-answering": snippetDocumentQuestionAnswering,
   "image-to-text": snippetFile2,
+  "image-to-image": snippetImageToImage,
   "zero-shot-image-classification": snippetZeroShotImageClassification
 };
 function getPythonInferenceSnippet(model, accessToken, provider, providerModelId, opts) {
@@ -2159,23 +2207,33 @@ function getPythonInferenceSnippet(model, accessToken, provider, providerModelId
     return snippets.map((snippet) => {
       return {
         ...snippet,
-        content: snippet.client === "requests" ? `import requests
-
-API_URL = "${(0, import_tasks3.openAIbaseUrl)(provider)}"
-headers = {"Authorization": ${accessToken ? `"Bearer ${accessToken}"` : `f"Bearer {API_TOKEN}"`}}
-
-${snippet.content}` : snippet.content
+        content: addImportsToSnippet(snippet.content, model, accessToken)
       };
     });
   }
 }
+var addImportsToSnippet = (snippet, model, accessToken) => {
+  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}`;
+  }
+  return snippet;
+};
 
 // src/snippets/js.ts
 var js_exports = {};
 __export(js_exports, {
   getJsInferenceSnippet: () => getJsInferenceSnippet,
   jsSnippets: () => jsSnippets,
-  snippetAutomaticSpeechRecognition: () => snippetAutomaticSpeechRecognition,
+  snippetAutomaticSpeechRecognition: () => snippetAutomaticSpeechRecognition2,
   snippetBasic: () => snippetBasic3,
   snippetFile: () => snippetFile3,
   snippetTextGeneration: () => snippetTextGeneration2,
@@ -2204,9 +2262,9 @@ var snippetBasic3 = (model, accessToken, provider) => {
     ...model.pipeline_tag && model.pipeline_tag in HFJS_METHODS ? [
       {
         client: "huggingface.js",
-        content: `import { HfInference } from "@huggingface/inference";
+        content: `import { InferenceClient } from "@huggingface/inference";
 
-const client = new HfInference("${accessToken || `{API_TOKEN}`}");
+const client = new InferenceClient("${accessToken || `{API_TOKEN}`}");
 
 const output = await client.${HFJS_METHODS[model.pipeline_tag]}({
 	model: "${model.id}",
@@ -2261,9 +2319,9 @@ var snippetTextGeneration2 = (model, accessToken, provider, providerModelId, opt
       return [
         {
           client: "huggingface.js",
-          content: `import { HfInference } from "@huggingface/inference";
+          content: `import { InferenceClient } from "@huggingface/inference";
 
-const client = new HfInference("${accessToken || `{API_TOKEN}`}");
+const client = new InferenceClient("${accessToken || `{API_TOKEN}`}");
 
 let out = "";
 
@@ -2313,9 +2371,9 @@ for await (const chunk of stream) {
       return [
         {
           client: "huggingface.js",
-          content: `import { HfInference } from "@huggingface/inference";
+          content: `import { InferenceClient } from "@huggingface/inference";
 
-const client = new HfInference("${accessToken || `{API_TOKEN}`}");
+const client = new InferenceClient("${accessToken || `{API_TOKEN}`}");
 
 const chatCompletion = await client.chatCompletion({
 	model: "${model.id}",
@@ -2383,9 +2441,9 @@ var snippetTextToImage2 = (model, accessToken, provider) => {
   return [
     {
       client: "huggingface.js",
-      content: `import { HfInference } from "@huggingface/inference";
+      content: `import { InferenceClient } from "@huggingface/inference";
 
-const client = new HfInference("${accessToken || `{API_TOKEN}`}");
+const client = new InferenceClient("${accessToken || `{API_TOKEN}`}");
 
 const image = await client.textToImage({
 	model: "${model.id}",
@@ -2425,9 +2483,9 @@ var snippetTextToVideo2 = (model, accessToken, provider) => {
   return ["fal-ai", "replicate"].includes(provider) ? [
     {
       client: "huggingface.js",
-      content: `import { HfInference } from "@huggingface/inference";
+      content: `import { InferenceClient } from "@huggingface/inference";
 
-const client = new HfInference("${accessToken || `{API_TOKEN}`}");
+const client = new InferenceClient("${accessToken || `{API_TOKEN}`}");
 
 const video = await client.textToVideo({
 	model: "${model.id}",
@@ -2485,13 +2543,13 @@ var snippetTextToAudio2 = (model, accessToken, provider) => {
     ];
   }
 };
-var snippetAutomaticSpeechRecognition = (model, accessToken, provider) => {
+var snippetAutomaticSpeechRecognition2 = (model, accessToken, provider) => {
   return [
     {
       client: "huggingface.js",
-      content: `import { HfInference } from "@huggingface/inference";
+      content: `import { InferenceClient } from "@huggingface/inference";
 
-const client = new HfInference("${accessToken || `{API_TOKEN}`}");
+const client = new InferenceClient("${accessToken || `{API_TOKEN}`}");
 
 const data = fs.readFileSync(${(0, import_tasks6.getModelInputSnippet)(model)});
 
@@ -2552,7 +2610,7 @@ var jsSnippets = {
   "text2text-generation": snippetBasic3,
   "fill-mask": snippetBasic3,
   "sentence-similarity": snippetBasic3,
-  "automatic-speech-recognition": snippetAutomaticSpeechRecognition,
+  "automatic-speech-recognition": snippetAutomaticSpeechRecognition2,
   "text-to-image": snippetTextToImage2,
   "text-to-video": snippetTextToVideo2,
   "text-to-speech": snippetTextToAudio2,
@@ -2570,8 +2628,9 @@ function getJsInferenceSnippet(model, accessToken, provider, providerModelId, op
 // Annotate the CommonJS export names for ESM import in node:
 0 && (module.exports = {
   HfInference,
-  HfInferenceEndpoint,
   INFERENCE_PROVIDERS,
+  InferenceClient,
+  InferenceClientEndpoint,
   InferenceOutputError,
   audioClassification,
   audioToAudio,