Standardize HF_ACCESS_TOKEN -> HF_TOKEN (#391)

Same as huggingface/transformers.js#431.
Related to [slack
thread](https://huggingface.slack.com/archives/C021H1P1HKR/p1701775450614369?thread_ts=1701719404.424999&cid=C021H1P1HKR)
(internal).

This PR standardizes the name of the environment variable used to make
requests with a HF API token. The goal is to harmonize it across the HF
ecosystem. For what I've seen, `HF_ACCESS_TOKEN` was not an "official"
environment variable but simply the one used in docs and tests. Updating
the naming should not break anything in the libraries.

**TODO before merging** update `HF_ACCESS_TOKEN` to `HF_TOKEN` in Github
repo secrets.

cc @xenova @julien-c

Co-authored-by: Eliott C <coyotte508@gmail.com>

.github/workflows/lint-and-test.yml

@@ -50,12 +50,12 @@ jobs:
       - name: Test
         run: VCR_MODE=playback pnpm --filter ...[${{ steps.since.outputs.SINCE }}] test
         env:
-          HF_ACCESS_TOKEN: ${{ secrets.HF_ACCESS_TOKEN }}
+          HF_TOKEN: ${{ secrets.HF_TOKEN }}
 
       - name: Test in browser
         run: VCR_MODE=playback pnpm --filter ...[${{ steps.since.outputs.SINCE }}] test:browser
         env:
-          HF_ACCESS_TOKEN: ${{ secrets.HF_ACCESS_TOKEN }}
+          HF_TOKEN: ${{ secrets.HF_TOKEN }}
 
       - name: E2E - start mock npm registry
         run: |
@@ -86,7 +86,7 @@ jobs:
           pnpm i --ignore-workspace --registry http://localhost:4874/
           pnpm start
         env:
-          token: ${{ secrets.HF_ACCESS_TOKEN }}
+          token: ${{ secrets.HF_TOKEN }}
 
       - name: E2E test - svelte app build
         working-directory: e2e/svelte

README.md

@@ -105,9 +105,9 @@ Get your HF access token in your [account settings](https://huggingface.co/setti
 ```ts
 import { HfInference } from "@huggingface/inference";
 
-const HF_ACCESS_TOKEN = "hf_...";
+const HF_TOKEN = "hf_...";
 
-const inference = new HfInference(HF_ACCESS_TOKEN);
+const inference = new HfInference(HF_TOKEN);
 
 // You can also omit "model" to use the recommended model for the task
 await inference.translation({
@@ -137,11 +137,11 @@ const { generated_text } = await gpt2.textGeneration({inputs: 'The answer to the
 ```ts
 import {HfAgent, LLMFromHub, defaultTools} from '@huggingface/agents';
 
-const HF_ACCESS_TOKEN = "hf_...";
+const HF_TOKEN = "hf_...";
 
 const agent = new HfAgent(
-  HF_ACCESS_TOKEN,
-  LLMFromHub(HF_ACCESS_TOKEN),
+  HF_TOKEN,
+  LLMFromHub(HF_TOKEN),
   [...defaultTools]
 );
 
@@ -162,16 +162,16 @@ console.log(messages);
 ```ts
 import { createRepo, uploadFile, deleteFiles } from "@huggingface/hub";
 
-const HF_ACCESS_TOKEN = "hf_...";
+const HF_TOKEN = "hf_...";
 
 await createRepo({
   repo: "my-user/nlp-model", // or {type: "model", name: "my-user/nlp-test"},
-  credentials: {accessToken: HF_ACCESS_TOKEN}
+  credentials: {accessToken: HF_TOKEN}
 });
 
 await uploadFile({
   repo: "my-user/nlp-model",
-  credentials: {accessToken: HF_ACCESS_TOKEN},
+  credentials: {accessToken: HF_TOKEN},
   // Can work with native File in browsers
   file: {
     path: "pytorch_model.bin",
@@ -181,7 +181,7 @@ await uploadFile({
 
 await deleteFiles({
   repo: {type: "space", name: "my-user/my-space"}, // or "spaces/my-user/my-space"
-  credentials: {accessToken: HF_ACCESS_TOKEN},
+  credentials: {accessToken: HF_TOKEN},
   paths: ["README.md", ".gitattributes"]
 });
 ```

packages/agents/README.md

@@ -131,7 +131,7 @@ const uppercaseTool: Tool = {
 };
 
 // pass it in the agent
-const agent = new HfAgent(process.env.HF_ACCESS_TOKEN, 
+const agent = new HfAgent(process.env.HF_TOKEN,
                 LLMFromHub("hf_...", "OpenAssistant/oasst-sft-4-pythia-12b-epoch-3.5"),
                 [uppercaseTool, ...defaultTools]);
 ```

packages/agents/test/HfAgent.spec.ts

@@ -4,20 +4,20 @@ import type { Data } from "../src/types";
 import type { HfInference } from "@huggingface/inference";
 
 const env = import.meta.env;
-if (!env.HF_ACCESS_TOKEN) {
-	console.warn("Set HF_ACCESS_TOKEN in the env to run the tests for better rate limits");
+if (!env.HF_TOKEN) {
+	console.warn("Set HF_TOKEN in the env to run the tests for better rate limits");
 }
 
 describe("HfAgent", () => {
 	it("You can create an agent from the hub", async () => {
-		const llm = LLMFromHub(env.HF_ACCESS_TOKEN, "OpenAssistant/oasst-sft-4-pythia-12b-epoch-3.5");
-		const agent = new HfAgent(env.HF_ACCESS_TOKEN, llm);
+		const llm = LLMFromHub(env.HF_TOKEN, "OpenAssistant/oasst-sft-4-pythia-12b-epoch-3.5");
+		const agent = new HfAgent(env.HF_TOKEN, llm);
 		expect(agent).toBeDefined();
 	});
 
 	it("You can create an agent from an endpoint", async () => {
-		const llm = LLMFromEndpoint(env.HF_ACCESS_TOKEN ?? "", "endpoint");
-		const agent = new HfAgent(env.HF_ACCESS_TOKEN, llm);
+		const llm = LLMFromEndpoint(env.HF_TOKEN ?? "", "endpoint");
+		const agent = new HfAgent(env.HF_TOKEN, llm);
 		expect(agent).toBeDefined();
 	});
 
@@ -42,7 +42,7 @@ describe("HfAgent", () => {
 			},
 		};
 
-		const agent = new HfAgent(env.HF_ACCESS_TOKEN, undefined, [uppercaseTool, ...defaultTools]);
+		const agent = new HfAgent(env.HF_TOKEN, undefined, [uppercaseTool, ...defaultTools]);
 		const code = `
 async function generate() {
 	const output = uppercase("hello friends");
@@ -61,7 +61,7 @@ async function generate() {
 	message(output);
 }`;
 
-		const agent = new HfAgent(env.HF_ACCESS_TOKEN);
+		const agent = new HfAgent(env.HF_TOKEN);
 
 		await agent.evaluateCode(code).then((output) => {
 			expect(output.length).toBeGreaterThan(0);
@@ -75,7 +75,7 @@ async function generate() {
 	toolThatDoesntExist(aaa);
 }`;
 
-		const hf = new HfAgent(env.HF_ACCESS_TOKEN);
+		const hf = new HfAgent(env.HF_TOKEN);
 
 		await hf.evaluateCode(code).then((output) => {
 			expect(output.length).toBeGreaterThan(0);

packages/inference/README.md

@@ -504,7 +504,7 @@ const { generated_text } = await gpt2.textGeneration({inputs: 'The answer to the
 ## Running tests
 
 ```console
-HF_ACCESS_TOKEN="your access token" pnpm run test
+HF_TOKEN="your access token" pnpm run test
 ```
 
 ## Finding appropriate models

packages/inference/test/HfInference.spec.ts

@@ -8,15 +8,15 @@ import { readTestFile } from "./test-files";
 const TIMEOUT = 60000 * 3;
 const env = import.meta.env;
 
-if (!env.HF_ACCESS_TOKEN) {
-	console.warn("Set HF_ACCESS_TOKEN in the env to run the tests for better rate limits");
+if (!env.HF_TOKEN) {
+	console.warn("Set HF_TOKEN in the env to run the tests for better rate limits");
 }
 
 describe.concurrent(
 	"HfInference",
 	() => {
 		// Individual tests can be ran without providing an api key, however running all tests without an api key will result in rate limiting error.
-		const hf = new HfInference(env.HF_ACCESS_TOKEN);
+		const hf = new HfInference(env.HF_TOKEN);
 
 		it("throws error if model does not exist", () => {
 			expect(

packages/inference/test/vcr.ts

@@ -27,7 +27,7 @@ if (env.VCR_MODE) {
 
 	VCR_MODE = env.VCR_MODE as MODE;
 } else {
-	VCR_MODE = env.HF_ACCESS_TOKEN ? MODE.DISABLED : MODE.PLAYBACK;
+	VCR_MODE = env.HF_TOKEN ? MODE.DISABLED : MODE.PLAYBACK;
 }
 
 const originalFetch = globalThis.fetch;

packages/tasks/src/tasks/audio-classification/about.md

@@ -53,7 +53,7 @@ You can use [huggingface.js](https://github.com/huggingface/huggingface.js) to i
 ```javascript
 import { HfInference } from "@huggingface/inference";
 
-const inference = new HfInference(HF_ACCESS_TOKEN);
+const inference = new HfInference(HF_TOKEN);
 await inference.audioClassification({
 	data: await (await fetch("sample.flac")).blob(),
 	model: "facebook/mms-lid-126",

packages/tasks/src/tasks/audio-to-audio/about.md

@@ -35,7 +35,7 @@ You can use [huggingface.js](https://github.com/huggingface/huggingface.js) to i
 ```javascript
 import { HfInference } from "@huggingface/inference";
 
-const inference = new HfInference(HF_ACCESS_TOKEN);
+const inference = new HfInference(HF_TOKEN);
 await inference.audioToAudio({
 	data: await (await fetch("sample.flac")).blob(),
 	model: "speechbrain/sepformer-wham",

packages/tasks/src/tasks/automatic-speech-recognition/about.md

@@ -54,7 +54,7 @@ You can use [huggingface.js](https://github.com/huggingface/huggingface.js) to t
 ```javascript
 import { HfInference } from "@huggingface/inference";
 
-const inference = new HfInference(HF_ACCESS_TOKEN);
+const inference = new HfInference(HF_TOKEN);
 await inference.automaticSpeechRecognition({
 	data: await (await fetch("sample.flac")).blob(),
 	model: "openai/whisper-large-v2",

packages/tasks/src/tasks/conversational/about.md

@@ -34,7 +34,7 @@ You can use [huggingface.js](https://github.com/huggingface/huggingface.js) to i
 ```javascript
 import { HfInference } from "@huggingface/inference";
 
-const inference = new HfInference(HF_ACCESS_TOKEN);
+const inference = new HfInference(HF_TOKEN);
 await inference.conversational({
 	model: "facebook/blenderbot-400M-distill",
 	inputs: "Going to the movies tonight - any suggestions?",

packages/tasks/src/tasks/image-classification/about.md

@@ -29,7 +29,7 @@ You can use [huggingface.js](https://github.com/huggingface/huggingface.js) to c
 ```javascript
 import { HfInference } from "@huggingface/inference";
 
-const inference = new HfInference(HF_ACCESS_TOKEN);
+const inference = new HfInference(HF_TOKEN);
 await inference.imageClassification({
 	data: await (await fetch("https://picsum.photos/300/300")).blob(),
 	model: "microsoft/resnet-50",

packages/tasks/src/tasks/image-segmentation/about.md

@@ -45,7 +45,7 @@ You can use [huggingface.js](https://github.com/huggingface/huggingface.js) to i
 ```javascript
 import { HfInference } from "@huggingface/inference";
 
-const inference = new HfInference(HF_ACCESS_TOKEN);
+const inference = new HfInference(HF_TOKEN);
 await inference.imageSegmentation({
 	data: await (await fetch("https://picsum.photos/300/300")).blob(),
 	model: "facebook/detr-resnet-50-panoptic",

packages/tasks/src/tasks/image-to-image/about.md

@@ -43,7 +43,7 @@ You can use [huggingface.js](https://github.com/huggingface/huggingface.js) to i
 ```javascript
 import { HfInference } from "@huggingface/inference";
 
-const inference = new HfInference(HF_ACCESS_TOKEN);
+const inference = new HfInference(HF_TOKEN);
 await inference.imageToImage({
 	data: await (await fetch("image")).blob(),
 	model: "timbrooks/instruct-pix2pix",

packages/tasks/src/tasks/image-to-text/about.md

@@ -48,7 +48,7 @@ You can use [huggingface.js](https://github.com/huggingface/huggingface.js) to i
 ```javascript
 import { HfInference } from "@huggingface/inference";
 
-const inference = new HfInference(HF_ACCESS_TOKEN);
+const inference = new HfInference(HF_TOKEN);
 await inference.imageToText({
 	data: await (await fetch("https://picsum.photos/300/300")).blob(),
 	model: "Salesforce/blip-image-captioning-base",

packages/tasks/src/tasks/summarization/about.md

@@ -25,7 +25,7 @@ You can use [huggingface.js](https://github.com/huggingface/huggingface.js) to i
 ```javascript
 import { HfInference } from "@huggingface/inference";
 
-const inference = new HfInference(HF_ACCESS_TOKEN);
+const inference = new HfInference(HF_TOKEN);
 const inputs =
 	"Paris is the capital and most populous city of France, with an estimated population of 2,175,601 residents as of 2018, in an area of more than 105 square kilometres (41 square miles). The City of Paris is the centre and seat of government of the region and province of Île-de-France, or Paris Region, which has an estimated population of 12,174,880, or about 18 percent of the population of France as of 2017.";
 

packages/tasks/src/tasks/text-classification/about.md

@@ -117,7 +117,7 @@ You can use [huggingface.js](https://github.com/huggingface/huggingface.js) to i
 ```javascript
 import { HfInference } from "@huggingface/inference";
 
-const inference = new HfInference(HF_ACCESS_TOKEN);
+const inference = new HfInference(HF_TOKEN);
 await inference.conversational({
 	model: "distilbert-base-uncased-finetuned-sst-2-english",
 	inputs: "I love this movie!",

packages/tasks/src/tasks/text-generation/about.md

@@ -72,7 +72,7 @@ You can use [huggingface.js](https://github.com/huggingface/huggingface.js) to i
 ```javascript
 import { HfInference } from "@huggingface/inference";
 
-const inference = new HfInference(HF_ACCESS_TOKEN);
+const inference = new HfInference(HF_TOKEN);
 await inference.conversational({
 	model: "distilbert-base-uncased-finetuned-sst-2-english",
 	inputs: "I love this movie!",

packages/tasks/src/tasks/text-to-image/about.md

@@ -41,7 +41,7 @@ You can use [huggingface.js](https://github.com/huggingface/huggingface.js) to i
 ```javascript
 import { HfInference } from "@huggingface/inference";
 
-const inference = new HfInference(HF_ACCESS_TOKEN);
+const inference = new HfInference(HF_TOKEN);
 await inference.textToImage({
 	model: "stabilityai/stable-diffusion-2",
 	inputs: "award winning high resolution photo of a giant tortoise/((ladybird)) hybrid, [trending on artstation]",

packages/tasks/src/tasks/text-to-speech/about.md

@@ -47,7 +47,7 @@ You can use [huggingface.js](https://github.com/huggingface/huggingface.js) to i
 ```javascript
 import { HfInference } from "@huggingface/inference";
 
-const inference = new HfInference(HF_ACCESS_TOKEN);
+const inference = new HfInference(HF_TOKEN);
 await inference.textToSpeech({
 	model: "facebook/mms-tts",
 	inputs: "text to generate speech from",

packages/tasks/src/tasks/translation/about.md

@@ -37,7 +37,7 @@ You can use [huggingface.js](https://github.com/huggingface/huggingface.js) to i
 ```javascript
 import { HfInference } from "@huggingface/inference";
 
-const inference = new HfInference(HF_ACCESS_TOKEN);
+const inference = new HfInference(HF_TOKEN);
 await inference.translation({
 	model: "t5-base",
 	inputs: "My name is Wolfgang and I live in Berlin",