@huggingface/inference 3.1.5 → 3.2.0

package/dist/src/providers/consts.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"consts.d.ts","sourceRoot":"","sources":["../../../src/providers/consts.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,OAAO,EAAE,MAAM,UAAU,CAAC;AAExC,KAAK,UAAU,GAAG,MAAM,CAAC;AAEzB;;;;GAIG;AACH,eAAO,MAAM,0BAA0B,EAAE,MAAM,CAAC,OAAO,EAAE,UAAU,CAKlE,CAAC"}

package/dist/src/providers/fal-ai.d.ts

@@ -1,6 +1,18 @@
-import type { ProviderMapping } from "./types";
 export declare const FAL_AI_API_BASE_URL = "https://fal.run";
-type FalAiId = string;
-export declare const FAL_AI_SUPPORTED_MODEL_IDS: ProviderMapping<FalAiId>;
-export {};
+/**
+ * See the registered mapping of HF model ID => Fal model ID here:
+ *
+ * https://huggingface.co/api/partners/fal-ai/models
+ *
+ * This is a publicly available mapping.
+ *
+ * If you want to try to run inference for a new model locally before it's registered on huggingface.co,
+ * you can add it to the dictionary "HARDCODED_MODEL_ID_MAPPING" in consts.ts, for dev purposes.
+ *
+ * - If you work at Fal and want to update this mapping, please use the model mapping API we provide on huggingface.co
+ * - If you're a community member and want to add a new supported HF model to Fal, please open an issue on the present repo
+ * and we will tag Fal team members.
+ *
+ * Thanks!
+ */
 //# sourceMappingURL=fal-ai.d.ts.map

package/dist/src/providers/fal-ai.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"fal-ai.d.ts","sourceRoot":"","sources":["../../../src/providers/fal-ai.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,SAAS,CAAC;AAE/C,eAAO,MAAM,mBAAmB,oBAAoB,CAAC;AAErD,KAAK,OAAO,GAAG,MAAM,CAAC;AAEtB,eAAO,MAAM,0BAA0B,EAAE,eAAe,CAAC,OAAO,CAwB/D,CAAC"}
+{"version":3,"file":"fal-ai.d.ts","sourceRoot":"","sources":["../../../src/providers/fal-ai.ts"],"names":[],"mappings":"AAAA,eAAO,MAAM,mBAAmB,oBAAoB,CAAC;AAErD;;;;;;;;;;;;;;;GAeG"}

package/dist/src/providers/replicate.d.ts

@@ -1,6 +1,18 @@
-import type { ProviderMapping } from "./types";
 export declare const REPLICATE_API_BASE_URL = "https://api.replicate.com";
-type ReplicateId = string;
-export declare const REPLICATE_SUPPORTED_MODEL_IDS: ProviderMapping<ReplicateId>;
-export {};
+/**
+ * See the registered mapping of HF model ID => Replicate model ID here:
+ *
+ * https://huggingface.co/api/partners/replicate/models
+ *
+ * This is a publicly available mapping.
+ *
+ * If you want to try to run inference for a new model locally before it's registered on huggingface.co,
+ * you can add it to the dictionary "HARDCODED_MODEL_ID_MAPPING" in consts.ts, for dev purposes.
+ *
+ * - If you work at Replicate and want to update this mapping, please use the model mapping API we provide on huggingface.co
+ * - If you're a community member and want to add a new supported HF model to Replicate, please open an issue on the present repo
+ * and we will tag Replicate team members.
+ *
+ * Thanks!
+ */
 //# sourceMappingURL=replicate.d.ts.map

package/dist/src/providers/replicate.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"replicate.d.ts","sourceRoot":"","sources":["../../../src/providers/replicate.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,SAAS,CAAC;AAE/C,eAAO,MAAM,sBAAsB,8BAA8B,CAAC;AAElE,KAAK,WAAW,GAAG,MAAM,CAAC;AAE1B,eAAO,MAAM,6BAA6B,EAAE,eAAe,CAAC,WAAW,CAsBtE,CAAC"}
+{"version":3,"file":"replicate.d.ts","sourceRoot":"","sources":["../../../src/providers/replicate.ts"],"names":[],"mappings":"AAAA,eAAO,MAAM,sBAAsB,8BAA8B,CAAC;AAElE;;;;;;;;;;;;;;;GAeG"}

package/dist/src/providers/sambanova.d.ts

@@ -1,6 +1,18 @@
-import type { ProviderMapping } from "./types";
 export declare const SAMBANOVA_API_BASE_URL = "https://api.sambanova.ai";
-type SambanovaId = string;
-export declare const SAMBANOVA_SUPPORTED_MODEL_IDS: ProviderMapping<SambanovaId>;
-export {};
+/**
+ * See the registered mapping of HF model ID => Sambanova model ID here:
+ *
+ * https://huggingface.co/api/partners/sambanova/models
+ *
+ * This is a publicly available mapping.
+ *
+ * If you want to try to run inference for a new model locally before it's registered on huggingface.co,
+ * you can add it to the dictionary "HARDCODED_MODEL_ID_MAPPING" in consts.ts, for dev purposes.
+ *
+ * - If you work at Sambanova and want to update this mapping, please use the model mapping API we provide on huggingface.co
+ * - If you're a community member and want to add a new supported HF model to Sambanova, please open an issue on the present repo
+ * and we will tag Sambanova team members.
+ *
+ * Thanks!
+ */
 //# sourceMappingURL=sambanova.d.ts.map

package/dist/src/providers/sambanova.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"sambanova.d.ts","sourceRoot":"","sources":["../../../src/providers/sambanova.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,SAAS,CAAC;AAE/C,eAAO,MAAM,sBAAsB,6BAA6B,CAAC;AAEjE,KAAK,WAAW,GAAG,MAAM,CAAC;AAE1B,eAAO,MAAM,6BAA6B,EAAE,eAAe,CAAC,WAAW,CAgBtE,CAAC"}
+{"version":3,"file":"sambanova.d.ts","sourceRoot":"","sources":["../../../src/providers/sambanova.ts"],"names":[],"mappings":"AAAA,eAAO,MAAM,sBAAsB,6BAA6B,CAAC;AAEjE;;;;;;;;;;;;;;;GAeG"}

package/dist/src/providers/together.d.ts

@@ -1,12 +1,18 @@
-import type { ProviderMapping } from "./types";
 export declare const TOGETHER_API_BASE_URL = "https://api.together.xyz";
 /**
- * Same comment as in sambanova.ts
+ * See the registered mapping of HF model ID => Together model ID here:
+ *
+ * https://huggingface.co/api/partners/together/models
+ *
+ * This is a publicly available mapping.
+ *
+ * If you want to try to run inference for a new model locally before it's registered on huggingface.co,
+ * you can add it to the dictionary "HARDCODED_MODEL_ID_MAPPING" in consts.ts, for dev purposes.
+ *
+ * - If you work at Together and want to update this mapping, please use the model mapping API we provide on huggingface.co
+ * - If you're a community member and want to add a new supported HF model to Together, please open an issue on the present repo
+ * and we will tag Together team members.
+ *
+ * Thanks!
  */
-type TogetherId = string;
-/**
- * https://docs.together.ai/reference/models-1
- */
-export declare const TOGETHER_SUPPORTED_MODEL_IDS: ProviderMapping<TogetherId>;
-export {};
 //# sourceMappingURL=together.d.ts.map

package/dist/src/providers/together.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"together.d.ts","sourceRoot":"","sources":["../../../src/providers/together.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,SAAS,CAAC;AAE/C,eAAO,MAAM,qBAAqB,6BAA6B,CAAC;AAEhE;;GAEG;AACH,KAAK,UAAU,GAAG,MAAM,CAAC;AAEzB;;GAEG;AACH,eAAO,MAAM,4BAA4B,EAAE,eAAe,CAAC,UAAU,CA8CpE,CAAC"}
+{"version":3,"file":"together.d.ts","sourceRoot":"","sources":["../../../src/providers/together.ts"],"names":[],"mappings":"AAAA,eAAO,MAAM,qBAAqB,6BAA6B,CAAC;AAEhE;;;;;;;;;;;;;;;GAeG"}

package/dist/src/tasks/audio/textToSpeech.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"textToSpeech.d.ts","sourceRoot":"","sources":["../../../../src/tasks/audio/textToSpeech.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,iBAAiB,EAAE,MAAM,oBAAoB,CAAC;AAE5D,OAAO,KAAK,EAAE,QAAQ,EAAE,OAAO,EAAE,MAAM,aAAa,CAAC;AAGrD,KAAK,gBAAgB,GAAG,QAAQ,GAAG,iBAAiB,CAAC;AAKrD;;;GAGG;AACH,wBAAsB,YAAY,CAAC,IAAI,EAAE,gBAAgB,EAAE,OAAO,CAAC,EAAE,OAAO,GAAG,OAAO,CAAC,IAAI,CAAC,CAsB3F"}
+{"version":3,"file":"textToSpeech.d.ts","sourceRoot":"","sources":["../../../../src/tasks/audio/textToSpeech.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,iBAAiB,EAAE,MAAM,oBAAoB,CAAC;AAE5D,OAAO,KAAK,EAAE,QAAQ,EAAE,OAAO,EAAE,MAAM,aAAa,CAAC;AAGrD,KAAK,gBAAgB,GAAG,QAAQ,GAAG,iBAAiB,CAAC;AAKrD;;;GAGG;AACH,wBAAsB,YAAY,CAAC,IAAI,EAAE,gBAAgB,EAAE,OAAO,CAAC,EAAE,OAAO,GAAG,OAAO,CAAC,IAAI,CAAC,CA+B3F"}

package/dist/src/types.d.ts

@@ -1,5 +1,4 @@
-import type { PipelineType } from "@huggingface/tasks";
-import type { ChatCompletionInput } from "@huggingface/tasks";
+import type { ChatCompletionInput, PipelineType } from "@huggingface/tasks";
 /**
  * HF model id, like "meta-llama/Llama-3.3-70B-Instruct"
  */
@@ -78,6 +77,8 @@ export type RequestArgs = BaseArgs & ({
     inputs: unknown;
 } | {
     prompt: string;
+} | {
+    text: string;
 } | {
     audio_url: string;
 } | ChatCompletionInput) & {

package/dist/src/types.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"types.d.ts","sourceRoot":"","sources":["../../src/types.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,oBAAoB,CAAC;AACvD,OAAO,KAAK,EAAE,mBAAmB,EAAE,MAAM,oBAAoB,CAAC;AAE9D;;GAEG;AACH,MAAM,MAAM,OAAO,GAAG,MAAM,CAAC;AAE7B,MAAM,WAAW,OAAO;IACvB;;OAEG;IACH,cAAc,CAAC,EAAE,OAAO,CAAC;IACzB;;OAEG;IACH,SAAS,CAAC,EAAE,OAAO,CAAC;IACpB;;OAEG;IACH,eAAe,CAAC,EAAE,OAAO,CAAC;IAC1B;;OAEG;IACH,OAAO,CAAC,EAAE,OAAO,CAAC;IAElB;;OAEG;IACH,cAAc,CAAC,EAAE,OAAO,CAAC;IACzB;;OAEG;IACH,KAAK,CAAC,EAAE,OAAO,KAAK,CAAC;IACrB;;OAEG;IACH,MAAM,CAAC,EAAE,WAAW,CAAC;IAErB;;OAEG;IACH,kBAAkB,CAAC,EAAE,MAAM,GAAG,OAAO,CAAC;CACtC;AAED,MAAM,MAAM,aAAa,GAAG,OAAO,CAAC,YAAY,EAAE,OAAO,CAAC,CAAC;AAE3D,eAAO,MAAM,mBAAmB,2EAA4E,CAAC;AAC7G,MAAM,MAAM,iBAAiB,GAAG,CAAC,OAAO,mBAAmB,CAAC,CAAC,MAAM,CAAC,CAAC;AAErE,MAAM,WAAW,QAAQ;IACxB;;;;;;OAMG;IACH,WAAW,CAAC,EAAE,MAAM,CAAC;IAErB;;;;;;;OAOG;IACH,KAAK,CAAC,EAAE,OAAO,CAAC;IAEhB;;;;OAIG;IACH,WAAW,CAAC,EAAE,MAAM,CAAC;IAErB;;;;OAIG;IACH,QAAQ,CAAC,EAAE,iBAAiB,CAAC;CAC7B;AAED,MAAM,MAAM,WAAW,GAAG,QAAQ,GACjC,CACG;IAAE,IAAI,EAAE,IAAI,GAAG,WAAW,CAAA;CAAE,GAC5B;IAAE,MAAM,EAAE,OAAO,CAAA;CAAE,GACnB;IAAE,MAAM,EAAE,MAAM,CAAA;CAAE,GAClB;IAAE,SAAS,EAAE,MAAM,CAAA;CAAE,GACrB,mBAAmB,CACrB,GAAG;IACH,UAAU,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;IACrC,WAAW,CAAC,EAAE,MAAM,CAAC;CACrB,CAAC"}
+{"version":3,"file":"types.d.ts","sourceRoot":"","sources":["../../src/types.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,mBAAmB,EAAE,YAAY,EAAE,MAAM,oBAAoB,CAAC;AAE5E;;GAEG;AACH,MAAM,MAAM,OAAO,GAAG,MAAM,CAAC;AAE7B,MAAM,WAAW,OAAO;IACvB;;OAEG;IACH,cAAc,CAAC,EAAE,OAAO,CAAC;IACzB;;OAEG;IACH,SAAS,CAAC,EAAE,OAAO,CAAC;IACpB;;OAEG;IACH,eAAe,CAAC,EAAE,OAAO,CAAC;IAC1B;;OAEG;IACH,OAAO,CAAC,EAAE,OAAO,CAAC;IAElB;;OAEG;IACH,cAAc,CAAC,EAAE,OAAO,CAAC;IACzB;;OAEG;IACH,KAAK,CAAC,EAAE,OAAO,KAAK,CAAC;IACrB;;OAEG;IACH,MAAM,CAAC,EAAE,WAAW,CAAC;IAErB;;OAEG;IACH,kBAAkB,CAAC,EAAE,MAAM,GAAG,OAAO,CAAC;CACtC;AAED,MAAM,MAAM,aAAa,GAAG,OAAO,CAAC,YAAY,EAAE,OAAO,CAAC,CAAC;AAE3D,eAAO,MAAM,mBAAmB,2EAA4E,CAAC;AAC7G,MAAM,MAAM,iBAAiB,GAAG,CAAC,OAAO,mBAAmB,CAAC,CAAC,MAAM,CAAC,CAAC;AAErE,MAAM,WAAW,QAAQ;IACxB;;;;;;OAMG;IACH,WAAW,CAAC,EAAE,MAAM,CAAC;IAErB;;;;;;;OAOG;IACH,KAAK,CAAC,EAAE,OAAO,CAAC;IAEhB;;;;OAIG;IACH,WAAW,CAAC,EAAE,MAAM,CAAC;IAErB;;;;OAIG;IACH,QAAQ,CAAC,EAAE,iBAAiB,CAAC;CAC7B;AAED,MAAM,MAAM,WAAW,GAAG,QAAQ,GACjC,CACG;IAAE,IAAI,EAAE,IAAI,GAAG,WAAW,CAAA;CAAE,GAC5B;IAAE,MAAM,EAAE,OAAO,CAAA;CAAE,GACnB;IAAE,MAAM,EAAE,MAAM,CAAA;CAAE,GAClB;IAAE,IAAI,EAAE,MAAM,CAAA;CAAE,GAChB;IAAE,SAAS,EAAE,MAAM,CAAA;CAAE,GACrB,mBAAmB,CACrB,GAAG;IACH,UAAU,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;IACrC,WAAW,CAAC,EAAE,MAAM,CAAC;CACrB,CAAC"}

package/dist/test/HfInference.spec.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"HfInference.spec.d.ts","sourceRoot":"","sources":["../../test/HfInference.spec.ts"],"names":[],"mappings":"AAKA,OAAO,OAAO,CAAC"}
+{"version":3,"file":"HfInference.spec.d.ts","sourceRoot":"","sources":["../../test/HfInference.spec.ts"],"names":[],"mappings":"AAOA,OAAO,OAAO,CAAC"}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@huggingface/inference",
-  "version": "3.1.5",
+  "version": "3.2.0",
   "packageManager": "pnpm@8.10.5",
   "license": "MIT",
   "author": "Tim Mikeladze <tim.mikeladze@gmail.com>",
@@ -39,7 +39,7 @@
   },
   "type": "module",
   "dependencies": {
-    "@huggingface/tasks": "^0.15.5"
+    "@huggingface/tasks": "^0.15.9"
   },
   "devDependencies": {
     "@types/node": "18.13.0"

package/src/config.ts

@@ -1 +1,2 @@
 export const HF_HUB_URL = "https://huggingface.co";
+export const HF_ROUTER_URL = "https://router.huggingface.co";

package/src/index.ts

@@ -1,9 +1,4 @@
-export type { ProviderMapping } from "./providers/types";
 export { HfInference, HfInferenceEndpoint } from "./HfInference";
 export { InferenceOutputError } from "./lib/InferenceOutputError";
-export { FAL_AI_SUPPORTED_MODEL_IDS } from "./providers/fal-ai";
-export { REPLICATE_SUPPORTED_MODEL_IDS } from "./providers/replicate";
-export { SAMBANOVA_SUPPORTED_MODEL_IDS } from "./providers/sambanova";
-export { TOGETHER_SUPPORTED_MODEL_IDS } from "./providers/together";
 export * from "./types";
 export * from "./tasks";

package/src/lib/getProviderModelId.ts

@@ -0,0 +1,74 @@
+import type { WidgetType } from "@huggingface/tasks";
+import type { InferenceProvider, InferenceTask, ModelId, Options, RequestArgs } from "../types";
+import { HF_HUB_URL } from "../config";
+import { HARDCODED_MODEL_ID_MAPPING } from "../providers/consts";
+
+type InferenceProviderMapping = Partial<
+	Record<InferenceProvider, { providerId: string; status: "live" | "staging"; task: WidgetType }>
+>;
+const inferenceProviderMappingCache = new Map<ModelId, InferenceProviderMapping>();
+
+export async function getProviderModelId(
+	params: {
+		model: string;
+		provider: InferenceProvider;
+	},
+	args: RequestArgs,
+	options: {
+		taskHint?: InferenceTask;
+		chatCompletion?: boolean;
+		fetch?: Options["fetch"];
+	} = {}
+): Promise<string> {
+	if (params.provider === "hf-inference") {
+		return params.model;
+	}
+	if (!options.taskHint) {
+		throw new Error("taskHint must be specified when using a third-party provider");
+	}
+	const task: WidgetType =
+		options.taskHint === "text-generation" && options.chatCompletion ? "conversational" : options.taskHint;
+
+	// A dict called HARDCODED_MODEL_ID_MAPPING takes precedence in all cases (useful for dev purposes)
+	if (HARDCODED_MODEL_ID_MAPPING[params.model]) {
+		return HARDCODED_MODEL_ID_MAPPING[params.model];
+	}
+
+	let inferenceProviderMapping: InferenceProviderMapping | null;
+	if (inferenceProviderMappingCache.has(params.model)) {
+		// eslint-disable-next-line @typescript-eslint/no-non-null-assertion
+		inferenceProviderMapping = inferenceProviderMappingCache.get(params.model)!;
+	} else {
+		inferenceProviderMapping = await (options?.fetch ?? fetch)(
+			`${HF_HUB_URL}/api/models/${params.model}?expand[]=inferenceProviderMapping`,
+			{
+				headers: args.accessToken?.startsWith("hf_") ? { Authorization: `Bearer ${args.accessToken}` } : {},
+			}
+		)
+			.then((resp) => resp.json())
+			.then((json) => json.inferenceProviderMapping)
+			.catch(() => null);
+	}
+
+	if (!inferenceProviderMapping) {
+		throw new Error(`We have not been able to find inference provider information for model ${params.model}.`);
+	}
+
+	const providerMapping = inferenceProviderMapping[params.provider];
+	if (providerMapping) {
+		if (providerMapping.task !== task) {
+			throw new Error(
+				`Model ${params.model} is not supported for task ${task} and provider ${params.provider}. Supported task: ${providerMapping.task}.`
+			);
+		}
+		if (providerMapping.status === "staging") {
+			console.warn(
+				`Model ${params.model} is in staging mode for provider ${params.provider}. Meant for test purposes only.`
+			);
+		}
+		// TODO: how is it handled server-side if model has multiple tasks (e.g. `text-generation` + `conversational`)?
+		return providerMapping.providerId;
+	}
+
+	throw new Error(`Model ${params.model} is not supported provider ${params.provider}.`);
+}

package/src/lib/makeRequestOptions.ts

@@ -1,15 +1,15 @@
-import type { WidgetType } from "@huggingface/tasks";
-import { HF_HUB_URL } from "../config";
-import { FAL_AI_API_BASE_URL, FAL_AI_SUPPORTED_MODEL_IDS } from "../providers/fal-ai";
-import { REPLICATE_API_BASE_URL, REPLICATE_SUPPORTED_MODEL_IDS } from "../providers/replicate";
-import { SAMBANOVA_API_BASE_URL, SAMBANOVA_SUPPORTED_MODEL_IDS } from "../providers/sambanova";
-import { TOGETHER_API_BASE_URL, TOGETHER_SUPPORTED_MODEL_IDS } from "../providers/together";
+import { HF_HUB_URL, HF_ROUTER_URL } from "../config";
+import { FAL_AI_API_BASE_URL } from "../providers/fal-ai";
+import { REPLICATE_API_BASE_URL } from "../providers/replicate";
+import { SAMBANOVA_API_BASE_URL } from "../providers/sambanova";
+import { TOGETHER_API_BASE_URL } from "../providers/together";
 import type { InferenceProvider } from "../types";
 import type { InferenceTask, Options, RequestArgs } from "../types";
 import { isUrl } from "./isUrl";
 import { version as packageVersion, name as packageName } from "../../package.json";
+import { getProviderModelId } from "./getProviderModelId";
 
-const HF_HUB_INFERENCE_PROXY_TEMPLATE = `${HF_HUB_URL}/api/inference-proxy/{{PROVIDER}}`;
+const HF_HUB_INFERENCE_PROXY_TEMPLATE = `${HF_ROUTER_URL}/{{PROVIDER}}`;
 
 /**
  * Lazy-loaded from huggingface.co/api/tasks when needed
@@ -49,18 +49,16 @@ export async function makeRequestOptions(
 	if (maybeModel && isUrl(maybeModel)) {
 		throw new Error(`Model URLs are no longer supported. Use endpointUrl instead.`);
 	}
-
-	let model: string;
-	if (!maybeModel) {
-		if (taskHint) {
-			model = mapModel({ model: await loadDefaultModel(taskHint), provider, taskHint, chatCompletion });
-		} else {
-			throw new Error("No model provided, and no default model found for this task");
-			/// TODO : change error message ^
-		}
-	} else {
-		model = mapModel({ model: maybeModel, provider, taskHint, chatCompletion });
+	if (!maybeModel && !taskHint) {
+		throw new Error("No model provided, and no task has been specified.");
 	}
+	// eslint-disable-next-line @typescript-eslint/no-non-null-assertion
+	const hfModel = maybeModel ?? (await loadDefaultModel(taskHint!));
+	const model = await getProviderModelId({ model: hfModel, provider }, args, {
+		taskHint,
+		chatCompletion,
+		fetch: options?.fetch,
+	});
 
 	/// If accessToken is passed, it should take precedence over includeCredentials
 	const authMethod = accessToken
@@ -153,39 +151,6 @@ export async function makeRequestOptions(
 	return { url, info };
 }
 
-function mapModel(params: {
-	model: string;
-	provider: InferenceProvider;
-	taskHint: InferenceTask | undefined;
-	chatCompletion: boolean | undefined;
-}): string {
-	if (params.provider === "hf-inference") {
-		return params.model;
-	}
-	if (!params.taskHint) {
-		throw new Error("taskHint must be specified when using a third-party provider");
-	}
-	const task: WidgetType =
-		params.taskHint === "text-generation" && params.chatCompletion ? "conversational" : params.taskHint;
-	const model = (() => {
-		switch (params.provider) {
-			case "fal-ai":
-				return FAL_AI_SUPPORTED_MODEL_IDS[task]?.[params.model];
-			case "replicate":
-				return REPLICATE_SUPPORTED_MODEL_IDS[task]?.[params.model];
-			case "sambanova":
-				return SAMBANOVA_SUPPORTED_MODEL_IDS[task]?.[params.model];
-			case "together":
-				return TOGETHER_SUPPORTED_MODEL_IDS[task]?.[params.model];
-		}
-	})();
-
-	if (!model) {
-		throw new Error(`Model ${params.model} is not supported for task ${task} and provider ${params.provider}`);
-	}
-	return model;
-}
-
 function makeUrl(params: {
 	authMethod: "none" | "hf-token" | "credentials-include" | "provider-key";
 	chatCompletion: boolean;

package/src/providers/consts.ts

@@ -0,0 +1,15 @@
+import type { ModelId } from "../types";
+
+type ProviderId = string;
+
+/**
+ * If you want to try to run inference for a new model locally before it's registered on huggingface.co
+ * for a given Inference Provider,
+ * you can add it to the following dictionary, for dev purposes.
+ */
+export const HARDCODED_MODEL_ID_MAPPING: Record<ModelId, ProviderId> = {
+	/**
+	 * "HF model ID" => "Model ID on Inference Provider's side"
+	 */
+	// "Qwen/Qwen2.5-Coder-32B-Instruct": "Qwen2.5-Coder-32B-Instruct",
+};

package/src/providers/fal-ai.ts

@@ -1,31 +1,18 @@
-import type { ProviderMapping } from "./types";
-
 export const FAL_AI_API_BASE_URL = "https://fal.run";
 
-type FalAiId = string;
-
-export const FAL_AI_SUPPORTED_MODEL_IDS: ProviderMapping<FalAiId> = {
-	"text-to-image": {
-		"black-forest-labs/FLUX.1-schnell": "fal-ai/flux/schnell",
-		"black-forest-labs/FLUX.1-dev": "fal-ai/flux/dev",
-		"playgroundai/playground-v2.5-1024px-aesthetic": "fal-ai/playground-v25",
-		"ByteDance/SDXL-Lightning": "fal-ai/lightning-models",
-		"PixArt-alpha/PixArt-Sigma-XL-2-1024-MS": "fal-ai/pixart-sigma",
-		"stabilityai/stable-diffusion-3-medium": "fal-ai/stable-diffusion-v3-medium",
-		"Warlord-K/Sana-1024": "fal-ai/sana",
-		"fal/AuraFlow-v0.2": "fal-ai/aura-flow",
-		"stabilityai/stable-diffusion-3.5-large": "fal-ai/stable-diffusion-v35-large",
-		"stabilityai/stable-diffusion-3.5-large-turbo": "fal-ai/stable-diffusion-v35-large/turbo",
-		"stabilityai/stable-diffusion-3.5-medium": "fal-ai/stable-diffusion-v35-medium",
-		"Kwai-Kolors/Kolors": "fal-ai/kolors",
-	},
-	"automatic-speech-recognition": {
-		"openai/whisper-large-v3": "fal-ai/whisper",
-	},
-	"text-to-video": {
-		"genmo/mochi-1-preview": "fal-ai/mochi-v1",
-		"tencent/HunyuanVideo": "fal-ai/hunyuan-video",
-		"THUDM/CogVideoX-5b": "fal-ai/cogvideox-5b",
-		"Lightricks/LTX-Video": "fal-ai/ltx-video",
-	},
-};
+/**
+ * See the registered mapping of HF model ID => Fal model ID here:
+ *
+ * https://huggingface.co/api/partners/fal-ai/models
+ *
+ * This is a publicly available mapping.
+ *
+ * If you want to try to run inference for a new model locally before it's registered on huggingface.co,
+ * you can add it to the dictionary "HARDCODED_MODEL_ID_MAPPING" in consts.ts, for dev purposes.
+ *
+ * - If you work at Fal and want to update this mapping, please use the model mapping API we provide on huggingface.co
+ * - If you're a community member and want to add a new supported HF model to Fal, please open an issue on the present repo
+ * and we will tag Fal team members.
+ *
+ * Thanks!
+ */

package/src/providers/replicate.ts

@@ -1,29 +1,18 @@
-import type { ProviderMapping } from "./types";
-
 export const REPLICATE_API_BASE_URL = "https://api.replicate.com";
 
-type ReplicateId = string;
-
-export const REPLICATE_SUPPORTED_MODEL_IDS: ProviderMapping<ReplicateId> = {
-	"text-to-image": {
-		"black-forest-labs/FLUX.1-dev": "black-forest-labs/flux-dev",
-		"black-forest-labs/FLUX.1-schnell": "black-forest-labs/flux-schnell",
-		"ByteDance/Hyper-SD":
-			"bytedance/hyper-flux-16step:382cf8959fb0f0d665b26e7e80b8d6dc3faaef1510f14ce017e8c732bb3d1eb7",
-		"ByteDance/SDXL-Lightning":
-			"bytedance/sdxl-lightning-4step:5599ed30703defd1d160a25a63321b4dec97101d98b4674bcc56e41f62f35637",
-		"playgroundai/playground-v2.5-1024px-aesthetic":
-			"playgroundai/playground-v2.5-1024px-aesthetic:a45f82a1382bed5c7aeb861dac7c7d191b0fdf74d8d57c4a0e6ed7d4d0bf7d24",
-		"stabilityai/stable-diffusion-3.5-large-turbo": "stability-ai/stable-diffusion-3.5-large-turbo",
-		"stabilityai/stable-diffusion-3.5-large": "stability-ai/stable-diffusion-3.5-large",
-		"stabilityai/stable-diffusion-3.5-medium": "stability-ai/stable-diffusion-3.5-medium",
-		"stabilityai/stable-diffusion-xl-base-1.0":
-			"stability-ai/sdxl:7762fd07cf82c948538e41f63f77d685e02b063e37e496e96eefd46c929f9bdc",
-	},
-	"text-to-speech": {
-		"OuteAI/OuteTTS-0.3-500M": "jbilcke/oute-tts:39a59319327b27327fa3095149c5a746e7f2aee18c75055c3368237a6503cd26",
-	},
-	"text-to-video": {
-		"genmo/mochi-1-preview": "genmoai/mochi-1:1944af04d098ef69bed7f9d335d102e652203f268ec4aaa2d836f6217217e460",
-	},
-};
+/**
+ * See the registered mapping of HF model ID => Replicate model ID here:
+ *
+ * https://huggingface.co/api/partners/replicate/models
+ *
+ * This is a publicly available mapping.
+ *
+ * If you want to try to run inference for a new model locally before it's registered on huggingface.co,
+ * you can add it to the dictionary "HARDCODED_MODEL_ID_MAPPING" in consts.ts, for dev purposes.
+ *
+ * - If you work at Replicate and want to update this mapping, please use the model mapping API we provide on huggingface.co
+ * - If you're a community member and want to add a new supported HF model to Replicate, please open an issue on the present repo
+ * and we will tag Replicate team members.
+ *
+ * Thanks!
+ */

package/src/providers/sambanova.ts

@@ -1,23 +1,18 @@
-import type { ProviderMapping } from "./types";
-
 export const SAMBANOVA_API_BASE_URL = "https://api.sambanova.ai";
 
-type SambanovaId = string;
-
-export const SAMBANOVA_SUPPORTED_MODEL_IDS: ProviderMapping<SambanovaId> = {
-	/** Chat completion / conversational */
-	conversational: {
-		"Qwen/Qwen2.5-Coder-32B-Instruct": "Qwen2.5-Coder-32B-Instruct",
-		"Qwen/Qwen2.5-72B-Instruct": "Qwen2.5-72B-Instruct",
-		"Qwen/QwQ-32B-Preview": "QwQ-32B-Preview",
-		"meta-llama/Llama-3.3-70B-Instruct": "Meta-Llama-3.3-70B-Instruct",
-		"meta-llama/Llama-3.2-1B-Instruct": "Meta-Llama-3.2-1B-Instruct",
-		"meta-llama/Llama-3.2-3B-Instruct": "Meta-Llama-3.2-3B-Instruct",
-		"meta-llama/Llama-3.2-11B-Vision-Instruct": "Llama-3.2-11B-Vision-Instruct",
-		"meta-llama/Llama-3.2-90B-Vision-Instruct": "Llama-3.2-90B-Vision-Instruct",
-		"meta-llama/Llama-3.1-8B-Instruct": "Meta-Llama-3.1-8B-Instruct",
-		"meta-llama/Llama-3.1-70B-Instruct": "Meta-Llama-3.1-70B-Instruct",
-		"meta-llama/Llama-3.1-405B-Instruct": "Meta-Llama-3.1-405B-Instruct",
-		"meta-llama/Llama-Guard-3-8B": "Meta-Llama-Guard-3-8B",
-	},
-};
+/**
+ * See the registered mapping of HF model ID => Sambanova model ID here:
+ *
+ * https://huggingface.co/api/partners/sambanova/models
+ *
+ * This is a publicly available mapping.
+ *
+ * If you want to try to run inference for a new model locally before it's registered on huggingface.co,
+ * you can add it to the dictionary "HARDCODED_MODEL_ID_MAPPING" in consts.ts, for dev purposes.
+ *
+ * - If you work at Sambanova and want to update this mapping, please use the model mapping API we provide on huggingface.co
+ * - If you're a community member and want to add a new supported HF model to Sambanova, please open an issue on the present repo
+ * and we will tag Sambanova team members.
+ *
+ * Thanks!
+ */

package/src/providers/together.ts

@@ -1,59 +1,18 @@
-import type { ProviderMapping } from "./types";
-
 export const TOGETHER_API_BASE_URL = "https://api.together.xyz";
 
 /**
- * Same comment as in sambanova.ts
- */
-type TogetherId = string;
-
-/**
- * https://docs.together.ai/reference/models-1
+ * See the registered mapping of HF model ID => Together model ID here:
+ *
+ * https://huggingface.co/api/partners/together/models
+ *
+ * This is a publicly available mapping.
+ *
+ * If you want to try to run inference for a new model locally before it's registered on huggingface.co,
+ * you can add it to the dictionary "HARDCODED_MODEL_ID_MAPPING" in consts.ts, for dev purposes.
+ *
+ * - If you work at Together and want to update this mapping, please use the model mapping API we provide on huggingface.co
+ * - If you're a community member and want to add a new supported HF model to Together, please open an issue on the present repo
+ * and we will tag Together team members.
+ *
+ * Thanks!
  */
-export const TOGETHER_SUPPORTED_MODEL_IDS: ProviderMapping<TogetherId> = {
-	"text-to-image": {
-		"black-forest-labs/FLUX.1-Canny-dev": "black-forest-labs/FLUX.1-canny",
-		"black-forest-labs/FLUX.1-Depth-dev": "black-forest-labs/FLUX.1-depth",
-		"black-forest-labs/FLUX.1-dev": "black-forest-labs/FLUX.1-dev",
-		"black-forest-labs/FLUX.1-Redux-dev": "black-forest-labs/FLUX.1-redux",
-		"black-forest-labs/FLUX.1-schnell": "black-forest-labs/FLUX.1-pro",
-		"stabilityai/stable-diffusion-xl-base-1.0": "stabilityai/stable-diffusion-xl-base-1.0",
-	},
-	conversational: {
-		"databricks/dbrx-instruct": "databricks/dbrx-instruct",
-		"deepseek-ai/DeepSeek-R1": "deepseek-ai/DeepSeek-R1",
-		"deepseek-ai/DeepSeek-V3": "deepseek-ai/DeepSeek-V3",
-		"deepseek-ai/deepseek-llm-67b-chat": "deepseek-ai/deepseek-llm-67b-chat",
-		"google/gemma-2-9b-it": "google/gemma-2-9b-it",
-		"google/gemma-2b-it": "google/gemma-2-27b-it",
-		"meta-llama/Llama-2-13b-chat-hf": "meta-llama/Llama-2-13b-chat-hf",
-		"meta-llama/Llama-2-7b-chat-hf": "meta-llama/Llama-2-7b-chat-hf",
-		"meta-llama/Llama-3.2-11B-Vision-Instruct": "meta-llama/Llama-Vision-Free",
-		"meta-llama/Llama-3.2-3B-Instruct": "meta-llama/Llama-3.2-3B-Instruct-Turbo",
-		"meta-llama/Llama-3.2-90B-Vision-Instruct": "meta-llama/Llama-3.2-90B-Vision-Instruct-Turbo",
-		"meta-llama/Llama-3.3-70B-Instruct": "meta-llama/Llama-3.3-70B-Instruct-Turbo",
-		"meta-llama/Meta-Llama-3-70B-Instruct": "meta-llama/Llama-3-70b-chat-hf",
-		"meta-llama/Meta-Llama-3-8B-Instruct": "meta-llama/Meta-Llama-3-8B-Instruct-Turbo",
-		"meta-llama/Meta-Llama-3.1-405B-Instruct": "meta-llama/Llama-3.2-11B-Vision-Instruct-Turbo",
-		"meta-llama/Meta-Llama-3.1-70B-Instruct": "meta-llama/Meta-Llama-3.1-70B-Instruct-Turbo",
-		"meta-llama/Meta-Llama-3.1-8B-Instruct": "meta-llama/Meta-Llama-3.1-8B-Instruct-Turbo-128K",
-		"microsoft/WizardLM-2-8x22B": "microsoft/WizardLM-2-8x22B",
-		"mistralai/Mistral-7B-Instruct-v0.3": "mistralai/Mistral-7B-Instruct-v0.3",
-		"mistralai/Mistral-Small-24B-Instruct-2501": "mistralai/Mistral-Small-24B-Instruct-2501",
-		"mistralai/Mixtral-8x22B-Instruct-v0.1": "mistralai/Mixtral-8x22B-Instruct-v0.1",
-		"mistralai/Mixtral-8x7B-Instruct-v0.1": "mistralai/Mixtral-8x7B-Instruct-v0.1",
-		"NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO": "NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO",
-		"nvidia/Llama-3.1-Nemotron-70B-Instruct-HF": "nvidia/Llama-3.1-Nemotron-70B-Instruct-HF",
-		"Qwen/Qwen2-72B-Instruct": "Qwen/Qwen2-72B-Instruct",
-		"Qwen/Qwen2.5-72B-Instruct": "Qwen/Qwen2.5-72B-Instruct-Turbo",
-		"Qwen/Qwen2.5-7B-Instruct": "Qwen/Qwen2.5-7B-Instruct-Turbo",
-		"Qwen/Qwen2.5-Coder-32B-Instruct": "Qwen/Qwen2.5-Coder-32B-Instruct",
-		"Qwen/QwQ-32B-Preview": "Qwen/QwQ-32B-Preview",
-		"scb10x/llama-3-typhoon-v1.5-8b-instruct": "scb10x/scb10x-llama3-typhoon-v1-5-8b-instruct",
-		"scb10x/llama-3-typhoon-v1.5x-70b-instruct-awq": "scb10x/scb10x-llama3-typhoon-v1-5x-4f316",
-	},
-	"text-generation": {
-		"meta-llama/Llama-2-70b-hf": "meta-llama/Llama-2-70b-hf",
-		"mistralai/Mixtral-8x7B-v0.1": "mistralai/Mixtral-8x7B-v0.1",
-	},
-};

package/src/tasks/audio/textToSpeech.ts

@@ -1,8 +1,8 @@
 import type { TextToSpeechInput } from "@huggingface/tasks";
 import { InferenceOutputError } from "../../lib/InferenceOutputError";
 import type { BaseArgs, Options } from "../../types";
+import { omit } from "../../utils/omit";
 import { request } from "../custom/request";
-
 type TextToSpeechArgs = BaseArgs & TextToSpeechInput;
 
 interface OutputUrlTextToSpeechGeneration {
@@ -13,7 +13,16 @@ interface OutputUrlTextToSpeechGeneration {
  * Recommended model: espnet/kan-bayashi_ljspeech_vits
  */
 export async function textToSpeech(args: TextToSpeechArgs, options?: Options): Promise<Blob> {
-	const res = await request<Blob | OutputUrlTextToSpeechGeneration>(args, {
+	// Replicate models expects "text" instead of "inputs"
+	const payload =
+		args.provider === "replicate"
+			? {
+					...omit(args, ["inputs", "parameters"]),
+					...args.parameters,
+					text: args.inputs,
+			  }
+			: args;
+	const res = await request<Blob | OutputUrlTextToSpeechGeneration>(payload, {
 		...options,
 		taskHint: "text-to-speech",
 	});

package/src/types.ts

@@ -1,5 +1,4 @@
-import type { PipelineType } from "@huggingface/tasks";
-import type { ChatCompletionInput } from "@huggingface/tasks";
+import type { ChatCompletionInput, PipelineType } from "@huggingface/tasks";
 
 /**
  * HF model id, like "meta-llama/Llama-3.3-70B-Instruct"
@@ -88,6 +87,7 @@ export type RequestArgs = BaseArgs &
 		| { data: Blob | ArrayBuffer }
 		| { inputs: unknown }
 		| { prompt: string }
+		| { text: string }
 		| { audio_url: string }
 		| ChatCompletionInput
 	) & {

package/dist/src/providers/types.d.ts

@@ -1,4 +0,0 @@
-import type { WidgetType } from "@huggingface/tasks";
-import type { ModelId } from "../types";
-export type ProviderMapping<ProviderId extends string> = Partial<Record<WidgetType, Partial<Record<ModelId, ProviderId>>>>;
-//# sourceMappingURL=types.d.ts.map