@huggingface/inference 3.3.2 → 3.3.4

package/README.md

@@ -49,10 +49,13 @@ You can send inference requests to third-party providers with the inference clie
 Currently, we support the following providers:
 - [Fal.ai](https://fal.ai)
 - [Fireworks AI](https://fireworks.ai)
+- [Hyperbolic](https://hyperbolic.xyz)
 - [Nebius](https://studio.nebius.ai)
+- [Novita](https://novita.ai/?utm_source=github_huggingface&utm_medium=github_readme&utm_campaign=link)
 - [Replicate](https://replicate.com)
 - [Sambanova](https://sambanova.ai)
 - [Together](https://together.xyz)
+- [Blackforestlabs](https://blackforestlabs.ai)
 
 To send requests to a third-party provider, you have to pass the `provider` parameter to the inference function. Make sure your request is authenticated with an access token.
 ```ts
@@ -72,6 +75,7 @@ When authenticated with a third-party provider key, the request is made directly
 Only a subset of models are supported when requesting third-party providers. You can check the list of supported models per pipeline tasks here:
 - [Fal.ai supported models](https://huggingface.co/api/partners/fal-ai/models)
 - [Fireworks AI supported models](https://huggingface.co/api/partners/fireworks-ai/models)
+- [Hyperbolic supported models](https://huggingface.co/api/partners/hyperbolic/models)
 - [Nebius supported models](https://huggingface.co/api/partners/nebius/models)
 - [Replicate supported models](https://huggingface.co/api/partners/replicate/models)
 - [Sambanova supported models](https://huggingface.co/api/partners/sambanova/models)
@@ -125,7 +129,7 @@ for await (const output of hf.textGenerationStream({
 
 ### 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.
+Using the `chatCompletion` method, you can generate text with models compatible with the OpenAI Chat Completion API. All models served by [TGI](https://huggingface.co/docs/text-generation-inference/) on Hugging Face support Messages API.
 
 [Demo](https://huggingface.co/spaces/huggingfacejs/streaming-chat-completion)
 

package/dist/index.cjs

@@ -98,6 +98,7 @@ __export(tasks_exports, {
 
 // src/config.ts
 var HF_HUB_URL = "https://huggingface.co";
+var HF_ROUTER_URL = "https://router.huggingface.co";
 
 // src/providers/fal-ai.ts
 var FAL_AI_API_BASE_URL = "https://fal.run";
@@ -114,9 +115,18 @@ var SAMBANOVA_API_BASE_URL = "https://api.sambanova.ai";
 // src/providers/together.ts
 var TOGETHER_API_BASE_URL = "https://api.together.xyz";
 
+// src/providers/novita.ts
+var NOVITA_API_BASE_URL = "https://api.novita.ai/v3/openai";
+
 // src/providers/fireworks-ai.ts
 var FIREWORKS_AI_API_BASE_URL = "https://api.fireworks.ai/inference";
 
+// src/providers/hyperbolic.ts
+var HYPERBOLIC_API_BASE_URL = "https://api.hyperbolic.xyz";
+
+// src/providers/black-forest-labs.ts
+var BLACKFORESTLABS_AI_API_BASE_URL = "https://api.us1.bfl.ai/v1";
+
 // src/lib/isUrl.ts
 function isUrl(modelOrUrl) {
   return /^http(s?):/.test(modelOrUrl) || modelOrUrl.startsWith("/");
@@ -124,7 +134,7 @@ function isUrl(modelOrUrl) {
 
 // package.json
 var name = "@huggingface/inference";
-var version = "3.3.2";
+var version = "3.3.4";
 
 // src/providers/consts.ts
 var HARDCODED_MODEL_ID_MAPPING = {
@@ -134,13 +144,16 @@ var HARDCODED_MODEL_ID_MAPPING = {
    * Example:
    * "Qwen/Qwen2.5-Coder-32B-Instruct": "Qwen2.5-Coder-32B-Instruct",
    */
+  "black-forest-labs": {},
   "fal-ai": {},
   "fireworks-ai": {},
   "hf-inference": {},
+  hyperbolic: {},
   nebius: {},
   replicate: {},
   sambanova: {},
-  together: {}
+  together: {},
+  novita: {}
 };
 
 // src/lib/getProviderModelId.ts
@@ -188,19 +201,16 @@ async function getProviderModelId(params, args, options = {}) {
 }
 
 // src/lib/makeRequestOptions.ts
-var HF_HUB_INFERENCE_PROXY_TEMPLATE = `${HF_HUB_URL}/api/inference-proxy/{{PROVIDER}}`;
+var HF_HUB_INFERENCE_PROXY_TEMPLATE = `${HF_ROUTER_URL}/{{PROVIDER}}`;
 var tasks = null;
 async function makeRequestOptions(args, options) {
   const { accessToken, endpointUrl, provider: maybeProvider, model: maybeModel, ...remainingArgs } = args;
   let otherArgs = remainingArgs;
   const provider = maybeProvider ?? "hf-inference";
-  const { forceTask, includeCredentials, taskHint, chatCompletion: chatCompletion2 } = options ?? {};
+  const { includeCredentials, taskHint, chatCompletion: chatCompletion2 } = options ?? {};
   if (endpointUrl && provider !== "hf-inference") {
     throw new Error(`Cannot use endpointUrl with a third-party provider.`);
   }
-  if (forceTask && provider !== "hf-inference") {
-    throw new Error(`Cannot use forceTask with a third-party provider.`);
-  }
   if (maybeModel && isUrl(maybeModel)) {
     throw new Error(`Model URLs are no longer supported. Use endpointUrl instead.`);
   }
@@ -217,14 +227,19 @@ async function makeRequestOptions(args, options) {
   const url = endpointUrl ? chatCompletion2 ? endpointUrl + `/v1/chat/completions` : endpointUrl : makeUrl({
     authMethod,
     chatCompletion: chatCompletion2 ?? false,
-    forceTask,
     model,
     provider: provider ?? "hf-inference",
     taskHint
   });
   const headers = {};
   if (accessToken) {
-    headers["Authorization"] = provider === "fal-ai" && authMethod === "provider-key" ? `Key ${accessToken}` : `Bearer ${accessToken}`;
+    if (provider === "fal-ai" && authMethod === "provider-key") {
+      headers["Authorization"] = `Key ${accessToken}`;
+    } else if (provider === "black-forest-labs" && authMethod === "provider-key") {
+      headers["X-Key"] = accessToken;
+    } else {
+      headers["Authorization"] = `Bearer ${accessToken}`;
+    }
   }
   const ownUserAgent = `${name}/${version}`;
   headers["User-Agent"] = [ownUserAgent, typeof navigator !== "undefined" ? navigator.userAgent : void 0].filter((x) => x !== void 0).join(" ");
@@ -250,7 +265,7 @@ async function makeRequestOptions(args, options) {
     method: "POST",
     body: binary ? args.data : JSON.stringify({
       ...otherArgs,
-      ...chatCompletion2 || provider === "together" || provider === "nebius" ? { model } : void 0
+      ...taskHint === "text-to-image" && provider === "hyperbolic" ? { model_name: model } : chatCompletion2 || provider === "together" || provider === "nebius" || provider === "hyperbolic" ? { model } : void 0
     }),
     ...credentials ? { credentials } : void 0,
     signal: options?.signal
@@ -263,6 +278,10 @@ function makeUrl(params) {
   }
   const shouldProxy = params.provider !== "hf-inference" && params.authMethod !== "provider-key";
   switch (params.provider) {
+    case "black-forest-labs": {
+      const baseUrl = shouldProxy ? HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", params.provider) : BLACKFORESTLABS_AI_API_BASE_URL;
+      return `${baseUrl}/${params.model}`;
+    }
     case "fal-ai": {
       const baseUrl = shouldProxy ? HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", params.provider) : FAL_AI_API_BASE_URL;
       return `${baseUrl}/${params.model}`;
@@ -314,13 +333,32 @@ function makeUrl(params) {
       }
       return baseUrl;
     }
+    case "hyperbolic": {
+      const baseUrl = shouldProxy ? HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", params.provider) : HYPERBOLIC_API_BASE_URL;
+      if (params.taskHint === "text-to-image") {
+        return `${baseUrl}/v1/images/generations`;
+      }
+      return `${baseUrl}/v1/chat/completions`;
+    }
+    case "novita": {
+      const baseUrl = shouldProxy ? HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", params.provider) : NOVITA_API_BASE_URL;
+      if (params.taskHint === "text-generation") {
+        if (params.chatCompletion) {
+          return `${baseUrl}/chat/completions`;
+        }
+        return `${baseUrl}/completions`;
+      }
+      return baseUrl;
+    }
     default: {
       const baseUrl = HF_HUB_INFERENCE_PROXY_TEMPLATE.replaceAll("{{PROVIDER}}", "hf-inference");
-      const url = params.forceTask ? `${baseUrl}/pipeline/${params.forceTask}/${params.model}` : `${baseUrl}/models/${params.model}`;
+      if (params.taskHint && ["feature-extraction", "sentence-similarity"].includes(params.taskHint)) {
+        return `${baseUrl}/pipeline/${params.taskHint}/${params.model}`;
+      }
       if (params.taskHint === "text-generation" && params.chatCompletion) {
-        return url + `/v1/chat/completions`;
+        return `${baseUrl}/models/${params.model}/v1/chat/completions`;
       }
-      return url;
+      return `${baseUrl}/models/${params.model}`;
     }
   }
 }
@@ -767,6 +805,13 @@ async function objectDetection(args, options) {
   return res;
 }
 
+// src/utils/delay.ts
+function delay(ms) {
+  return new Promise((resolve) => {
+    setTimeout(() => resolve(), ms);
+  });
+}
+
 // src/tasks/cv/textToImage.ts
 function getResponseFormatArg(provider) {
   switch (provider) {
@@ -794,10 +839,18 @@ async function textToImage(args, options) {
     taskHint: "text-to-image"
   });
   if (res && typeof res === "object") {
+    if (args.provider === "black-forest-labs" && "polling_url" in res && typeof res.polling_url === "string") {
+      return await pollBflResponse(res.polling_url);
+    }
     if (args.provider === "fal-ai" && "images" in res && Array.isArray(res.images) && res.images[0].url) {
       const image = await fetch(res.images[0].url);
       return await image.blob();
     }
+    if (args.provider === "hyperbolic" && "images" in res && Array.isArray(res.images) && res.images[0] && typeof res.images[0].image === "string") {
+      const base64Response = await fetch(`data:image/jpeg;base64,${res.images[0].image}`);
+      const blob = await base64Response.blob();
+      return blob;
+    }
     if ("data" in res && Array.isArray(res.data) && res.data[0].b64_json) {
       const base64Data = res.data[0].b64_json;
       const base64Response = await fetch(`data:image/jpeg;base64,${base64Data}`);
@@ -816,6 +869,24 @@ async function textToImage(args, options) {
   }
   return res;
 }
+async function pollBflResponse(url) {
+  const urlObj = new URL(url);
+  for (let step = 0; step < 5; step++) {
+    await delay(1e3);
+    console.debug(`Polling Black Forest Labs API for the result... ${step + 1}/5`);
+    urlObj.searchParams.set("attempt", step.toString(10));
+    const resp = await fetch(urlObj, { headers: { "Content-Type": "application/json" } });
+    if (!resp.ok) {
+      throw new InferenceOutputError("Failed to fetch result from black forest labs API");
+    }
+    const payload = await resp.json();
+    if (typeof payload === "object" && payload && "status" in payload && typeof payload.status === "string" && payload.status === "Ready" && "result" in payload && typeof payload.result === "object" && payload.result && "sample" in payload.result && typeof payload.result.sample === "string") {
+      const image = await fetch(payload.result.sample);
+      return await image.blob();
+    }
+  }
+  throw new InferenceOutputError("Failed to fetch result from black forest labs API");
+}
 
 // src/tasks/cv/imageToImage.ts
 async function imageToImage(args, options) {
@@ -910,43 +981,11 @@ async function textToVideo(args, options) {
   }
 }
 
-// src/lib/getDefaultTask.ts
-var taskCache = /* @__PURE__ */ new Map();
-var CACHE_DURATION = 10 * 60 * 1e3;
-var MAX_CACHE_ITEMS = 1e3;
-async function getDefaultTask(model, accessToken, options) {
-  if (isUrl(model)) {
-    return null;
-  }
-  const key = `${model}:${accessToken}`;
-  let cachedTask = taskCache.get(key);
-  if (cachedTask && cachedTask.date < new Date(Date.now() - CACHE_DURATION)) {
-    taskCache.delete(key);
-    cachedTask = void 0;
-  }
-  if (cachedTask === void 0) {
-    const modelTask = await (options?.fetch ?? fetch)(`${HF_HUB_URL}/api/models/${model}?expand[]=pipeline_tag`, {
-      headers: accessToken ? { Authorization: `Bearer ${accessToken}` } : {}
-    }).then((resp) => resp.json()).then((json) => json.pipeline_tag).catch(() => null);
-    if (!modelTask) {
-      return null;
-    }
-    cachedTask = { task: modelTask, date: /* @__PURE__ */ new Date() };
-    taskCache.set(key, { task: modelTask, date: /* @__PURE__ */ new Date() });
-    if (taskCache.size > MAX_CACHE_ITEMS) {
-      taskCache.delete(taskCache.keys().next().value);
-    }
-  }
-  return cachedTask.task;
-}
-
 // src/tasks/nlp/featureExtraction.ts
 async function featureExtraction(args, options) {
-  const defaultTask = args.model ? await getDefaultTask(args.model, args.accessToken, options) : void 0;
   const res = await request(args, {
     ...options,
-    taskHint: "feature-extraction",
-    ...defaultTask === "sentence-similarity" && { forceTask: "feature-extraction" }
+    taskHint: "feature-extraction"
   });
   let isValidOutput = true;
   const isNumArrayRec = (arr, maxDepth, curDepth = 0) => {
@@ -999,11 +1038,9 @@ async function questionAnswering(args, options) {
 
 // src/tasks/nlp/sentenceSimilarity.ts
 async function sentenceSimilarity(args, options) {
-  const defaultTask = args.model ? await getDefaultTask(args.model, args.accessToken, options) : void 0;
   const res = await request(prepareInput(args), {
     ...options,
-    taskHint: "sentence-similarity",
-    ...defaultTask === "feature-extraction" && { forceTask: "sentence-similarity" }
+    taskHint: "sentence-similarity"
   });
   const isValidOutput = Array.isArray(res) && res.every((x) => typeof x === "number");
   if (!isValidOutput) {
@@ -1089,6 +1126,27 @@ async function textGeneration(args, options) {
     return {
       generated_text: completion.text
     };
+  } else if (args.provider === "hyperbolic") {
+    const payload = {
+      messages: [{ content: args.inputs, role: "user" }],
+      ...args.parameters ? {
+        max_tokens: args.parameters.max_new_tokens,
+        ...omit(args.parameters, "max_new_tokens")
+      } : void 0,
+      ...omit(args, ["inputs", "parameters"])
+    };
+    const raw = await request(payload, {
+      ...options,
+      taskHint: "text-generation"
+    });
+    const isValidOutput = typeof raw === "object" && "choices" in raw && Array.isArray(raw?.choices) && typeof raw?.model === "string";
+    if (!isValidOutput) {
+      throw new InferenceOutputError("Expected ChatCompletionOutput");
+    }
+    const completion = raw.choices[0];
+    return {
+      generated_text: completion.message.content
+    };
   } else {
     const res = toArray(
       await request(args, {
@@ -1301,10 +1359,13 @@ var HfInferenceEndpoint = class {
 
 // src/types.ts
 var INFERENCE_PROVIDERS = [
+  "black-forest-labs",
   "fal-ai",
   "fireworks-ai",
-  "nebius",
   "hf-inference",
+  "hyperbolic",
+  "nebius",
+  "novita",
   "replicate",
   "sambanova",
   "together"

package/dist/index.js

@@ -43,6 +43,7 @@ __export(tasks_exports, {
 
 // src/config.ts
 var HF_HUB_URL = "https://huggingface.co";
+var HF_ROUTER_URL = "https://router.huggingface.co";
 
 // src/providers/fal-ai.ts
 var FAL_AI_API_BASE_URL = "https://fal.run";
@@ -59,9 +60,18 @@ var SAMBANOVA_API_BASE_URL = "https://api.sambanova.ai";
 // src/providers/together.ts
 var TOGETHER_API_BASE_URL = "https://api.together.xyz";
 
+// src/providers/novita.ts
+var NOVITA_API_BASE_URL = "https://api.novita.ai/v3/openai";
+
 // src/providers/fireworks-ai.ts
 var FIREWORKS_AI_API_BASE_URL = "https://api.fireworks.ai/inference";
 
+// src/providers/hyperbolic.ts
+var HYPERBOLIC_API_BASE_URL = "https://api.hyperbolic.xyz";
+
+// src/providers/black-forest-labs.ts
+var BLACKFORESTLABS_AI_API_BASE_URL = "https://api.us1.bfl.ai/v1";
+
 // src/lib/isUrl.ts
 function isUrl(modelOrUrl) {
   return /^http(s?):/.test(modelOrUrl) || modelOrUrl.startsWith("/");
@@ -69,7 +79,7 @@ function isUrl(modelOrUrl) {
 
 // package.json
 var name = "@huggingface/inference";
-var version = "3.3.2";
+var version = "3.3.4";
 
 // src/providers/consts.ts
 var HARDCODED_MODEL_ID_MAPPING = {
@@ -79,13 +89,16 @@ var HARDCODED_MODEL_ID_MAPPING = {
    * Example:
    * "Qwen/Qwen2.5-Coder-32B-Instruct": "Qwen2.5-Coder-32B-Instruct",
    */
+  "black-forest-labs": {},
   "fal-ai": {},
   "fireworks-ai": {},
   "hf-inference": {},
+  hyperbolic: {},
   nebius: {},
   replicate: {},
   sambanova: {},
-  together: {}
+  together: {},
+  novita: {}
 };
 
 // src/lib/getProviderModelId.ts
@@ -133,19 +146,16 @@ async function getProviderModelId(params, args, options = {}) {
 }
 
 // src/lib/makeRequestOptions.ts
-var HF_HUB_INFERENCE_PROXY_TEMPLATE = `${HF_HUB_URL}/api/inference-proxy/{{PROVIDER}}`;
+var HF_HUB_INFERENCE_PROXY_TEMPLATE = `${HF_ROUTER_URL}/{{PROVIDER}}`;
 var tasks = null;
 async function makeRequestOptions(args, options) {
   const { accessToken, endpointUrl, provider: maybeProvider, model: maybeModel, ...remainingArgs } = args;
   let otherArgs = remainingArgs;
   const provider = maybeProvider ?? "hf-inference";
-  const { forceTask, includeCredentials, taskHint, chatCompletion: chatCompletion2 } = options ?? {};
+  const { includeCredentials, taskHint, chatCompletion: chatCompletion2 } = options ?? {};
   if (endpointUrl && provider !== "hf-inference") {
     throw new Error(`Cannot use endpointUrl with a third-party provider.`);
   }
-  if (forceTask && provider !== "hf-inference") {
-    throw new Error(`Cannot use forceTask with a third-party provider.`);
-  }
   if (maybeModel && isUrl(maybeModel)) {
     throw new Error(`Model URLs are no longer supported. Use endpointUrl instead.`);
   }
@@ -162,14 +172,19 @@ async function makeRequestOptions(args, options) {
   const url = endpointUrl ? chatCompletion2 ? endpointUrl + `/v1/chat/completions` : endpointUrl : makeUrl({
     authMethod,
     chatCompletion: chatCompletion2 ?? false,
-    forceTask,
     model,
     provider: provider ?? "hf-inference",
     taskHint
   });
   const headers = {};
   if (accessToken) {
-    headers["Authorization"] = provider === "fal-ai" && authMethod === "provider-key" ? `Key ${accessToken}` : `Bearer ${accessToken}`;
+    if (provider === "fal-ai" && authMethod === "provider-key") {
+      headers["Authorization"] = `Key ${accessToken}`;
+    } else if (provider === "black-forest-labs" && authMethod === "provider-key") {
+      headers["X-Key"] = accessToken;
+    } else {
+      headers["Authorization"] = `Bearer ${accessToken}`;
+    }
   }
   const ownUserAgent = `${name}/${version}`;
   headers["User-Agent"] = [ownUserAgent, typeof navigator !== "undefined" ? navigator.userAgent : void 0].filter((x) => x !== void 0).join(" ");
@@ -195,7 +210,7 @@ async function makeRequestOptions(args, options) {
     method: "POST",
     body: binary ? args.data : JSON.stringify({
       ...otherArgs,
-      ...chatCompletion2 || provider === "together" || provider === "nebius" ? { model } : void 0
+      ...taskHint === "text-to-image" && provider === "hyperbolic" ? { model_name: model } : chatCompletion2 || provider === "together" || provider === "nebius" || provider === "hyperbolic" ? { model } : void 0
     }),
     ...credentials ? { credentials } : void 0,
     signal: options?.signal
@@ -208,6 +223,10 @@ function makeUrl(params) {
   }
   const shouldProxy = params.provider !== "hf-inference" && params.authMethod !== "provider-key";
   switch (params.provider) {
+    case "black-forest-labs": {
+      const baseUrl = shouldProxy ? HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", params.provider) : BLACKFORESTLABS_AI_API_BASE_URL;
+      return `${baseUrl}/${params.model}`;
+    }
     case "fal-ai": {
       const baseUrl = shouldProxy ? HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", params.provider) : FAL_AI_API_BASE_URL;
       return `${baseUrl}/${params.model}`;
@@ -259,13 +278,32 @@ function makeUrl(params) {
       }
       return baseUrl;
     }
+    case "hyperbolic": {
+      const baseUrl = shouldProxy ? HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", params.provider) : HYPERBOLIC_API_BASE_URL;
+      if (params.taskHint === "text-to-image") {
+        return `${baseUrl}/v1/images/generations`;
+      }
+      return `${baseUrl}/v1/chat/completions`;
+    }
+    case "novita": {
+      const baseUrl = shouldProxy ? HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", params.provider) : NOVITA_API_BASE_URL;
+      if (params.taskHint === "text-generation") {
+        if (params.chatCompletion) {
+          return `${baseUrl}/chat/completions`;
+        }
+        return `${baseUrl}/completions`;
+      }
+      return baseUrl;
+    }
     default: {
       const baseUrl = HF_HUB_INFERENCE_PROXY_TEMPLATE.replaceAll("{{PROVIDER}}", "hf-inference");
-      const url = params.forceTask ? `${baseUrl}/pipeline/${params.forceTask}/${params.model}` : `${baseUrl}/models/${params.model}`;
+      if (params.taskHint && ["feature-extraction", "sentence-similarity"].includes(params.taskHint)) {
+        return `${baseUrl}/pipeline/${params.taskHint}/${params.model}`;
+      }
       if (params.taskHint === "text-generation" && params.chatCompletion) {
-        return url + `/v1/chat/completions`;
+        return `${baseUrl}/models/${params.model}/v1/chat/completions`;
       }
-      return url;
+      return `${baseUrl}/models/${params.model}`;
     }
   }
 }
@@ -712,6 +750,13 @@ async function objectDetection(args, options) {
   return res;
 }
 
+// src/utils/delay.ts
+function delay(ms) {
+  return new Promise((resolve) => {
+    setTimeout(() => resolve(), ms);
+  });
+}
+
 // src/tasks/cv/textToImage.ts
 function getResponseFormatArg(provider) {
   switch (provider) {
@@ -739,10 +784,18 @@ async function textToImage(args, options) {
     taskHint: "text-to-image"
   });
   if (res && typeof res === "object") {
+    if (args.provider === "black-forest-labs" && "polling_url" in res && typeof res.polling_url === "string") {
+      return await pollBflResponse(res.polling_url);
+    }
     if (args.provider === "fal-ai" && "images" in res && Array.isArray(res.images) && res.images[0].url) {
       const image = await fetch(res.images[0].url);
       return await image.blob();
     }
+    if (args.provider === "hyperbolic" && "images" in res && Array.isArray(res.images) && res.images[0] && typeof res.images[0].image === "string") {
+      const base64Response = await fetch(`data:image/jpeg;base64,${res.images[0].image}`);
+      const blob = await base64Response.blob();
+      return blob;
+    }
     if ("data" in res && Array.isArray(res.data) && res.data[0].b64_json) {
       const base64Data = res.data[0].b64_json;
       const base64Response = await fetch(`data:image/jpeg;base64,${base64Data}`);
@@ -761,6 +814,24 @@ async function textToImage(args, options) {
   }
   return res;
 }
+async function pollBflResponse(url) {
+  const urlObj = new URL(url);
+  for (let step = 0; step < 5; step++) {
+    await delay(1e3);
+    console.debug(`Polling Black Forest Labs API for the result... ${step + 1}/5`);
+    urlObj.searchParams.set("attempt", step.toString(10));
+    const resp = await fetch(urlObj, { headers: { "Content-Type": "application/json" } });
+    if (!resp.ok) {
+      throw new InferenceOutputError("Failed to fetch result from black forest labs API");
+    }
+    const payload = await resp.json();
+    if (typeof payload === "object" && payload && "status" in payload && typeof payload.status === "string" && payload.status === "Ready" && "result" in payload && typeof payload.result === "object" && payload.result && "sample" in payload.result && typeof payload.result.sample === "string") {
+      const image = await fetch(payload.result.sample);
+      return await image.blob();
+    }
+  }
+  throw new InferenceOutputError("Failed to fetch result from black forest labs API");
+}
 
 // src/tasks/cv/imageToImage.ts
 async function imageToImage(args, options) {
@@ -855,43 +926,11 @@ async function textToVideo(args, options) {
   }
 }
 
-// src/lib/getDefaultTask.ts
-var taskCache = /* @__PURE__ */ new Map();
-var CACHE_DURATION = 10 * 60 * 1e3;
-var MAX_CACHE_ITEMS = 1e3;
-async function getDefaultTask(model, accessToken, options) {
-  if (isUrl(model)) {
-    return null;
-  }
-  const key = `${model}:${accessToken}`;
-  let cachedTask = taskCache.get(key);
-  if (cachedTask && cachedTask.date < new Date(Date.now() - CACHE_DURATION)) {
-    taskCache.delete(key);
-    cachedTask = void 0;
-  }
-  if (cachedTask === void 0) {
-    const modelTask = await (options?.fetch ?? fetch)(`${HF_HUB_URL}/api/models/${model}?expand[]=pipeline_tag`, {
-      headers: accessToken ? { Authorization: `Bearer ${accessToken}` } : {}
-    }).then((resp) => resp.json()).then((json) => json.pipeline_tag).catch(() => null);
-    if (!modelTask) {
-      return null;
-    }
-    cachedTask = { task: modelTask, date: /* @__PURE__ */ new Date() };
-    taskCache.set(key, { task: modelTask, date: /* @__PURE__ */ new Date() });
-    if (taskCache.size > MAX_CACHE_ITEMS) {
-      taskCache.delete(taskCache.keys().next().value);
-    }
-  }
-  return cachedTask.task;
-}
-
 // src/tasks/nlp/featureExtraction.ts
 async function featureExtraction(args, options) {
-  const defaultTask = args.model ? await getDefaultTask(args.model, args.accessToken, options) : void 0;
   const res = await request(args, {
     ...options,
-    taskHint: "feature-extraction",
-    ...defaultTask === "sentence-similarity" && { forceTask: "feature-extraction" }
+    taskHint: "feature-extraction"
   });
   let isValidOutput = true;
   const isNumArrayRec = (arr, maxDepth, curDepth = 0) => {
@@ -944,11 +983,9 @@ async function questionAnswering(args, options) {
 
 // src/tasks/nlp/sentenceSimilarity.ts
 async function sentenceSimilarity(args, options) {
-  const defaultTask = args.model ? await getDefaultTask(args.model, args.accessToken, options) : void 0;
   const res = await request(prepareInput(args), {
     ...options,
-    taskHint: "sentence-similarity",
-    ...defaultTask === "feature-extraction" && { forceTask: "sentence-similarity" }
+    taskHint: "sentence-similarity"
   });
   const isValidOutput = Array.isArray(res) && res.every((x) => typeof x === "number");
   if (!isValidOutput) {
@@ -1034,6 +1071,27 @@ async function textGeneration(args, options) {
     return {
       generated_text: completion.text
     };
+  } else if (args.provider === "hyperbolic") {
+    const payload = {
+      messages: [{ content: args.inputs, role: "user" }],
+      ...args.parameters ? {
+        max_tokens: args.parameters.max_new_tokens,
+        ...omit(args.parameters, "max_new_tokens")
+      } : void 0,
+      ...omit(args, ["inputs", "parameters"])
+    };
+    const raw = await request(payload, {
+      ...options,
+      taskHint: "text-generation"
+    });
+    const isValidOutput = typeof raw === "object" && "choices" in raw && Array.isArray(raw?.choices) && typeof raw?.model === "string";
+    if (!isValidOutput) {
+      throw new InferenceOutputError("Expected ChatCompletionOutput");
+    }
+    const completion = raw.choices[0];
+    return {
+      generated_text: completion.message.content
+    };
   } else {
     const res = toArray(
       await request(args, {
@@ -1246,10 +1304,13 @@ var HfInferenceEndpoint = class {
 
 // src/types.ts
 var INFERENCE_PROVIDERS = [
+  "black-forest-labs",
   "fal-ai",
   "fireworks-ai",
-  "nebius",
   "hf-inference",
+  "hyperbolic",
+  "nebius",
+  "novita",
   "replicate",
   "sambanova",
   "together"

package/dist/src/config.d.ts

@@ -1,2 +1,3 @@
 export declare const HF_HUB_URL = "https://huggingface.co";
+export declare const HF_ROUTER_URL = "https://router.huggingface.co";
 //# sourceMappingURL=config.d.ts.map

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

@@ -1 +1 @@
-{"version":3,"file":"config.d.ts","sourceRoot":"","sources":["../../src/config.ts"],"names":[],"mappings":"AAAA,eAAO,MAAM,UAAU,2BAA2B,CAAC"}
+{"version":3,"file":"config.d.ts","sourceRoot":"","sources":["../../src/config.ts"],"names":[],"mappings":"AAAA,eAAO,MAAM,UAAU,2BAA2B,CAAC;AACnD,eAAO,MAAM,aAAa,kCAAkC,CAAC"}

package/dist/src/lib/makeRequestOptions.d.ts

@@ -6,8 +6,6 @@ export declare function makeRequestOptions(args: RequestArgs & {
     data?: Blob | ArrayBuffer;
     stream?: boolean;
 }, options?: Options & {
-    /** When a model can be used for multiple tasks, and we want to run a non-default task */
-    forceTask?: string | InferenceTask;
     /** To load default model if needed */
     taskHint?: InferenceTask;
     chatCompletion?: boolean;

package/dist/src/lib/makeRequestOptions.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"makeRequestOptions.d.ts","sourceRoot":"","sources":["../../../src/lib/makeRequestOptions.ts"],"names":[],"mappings":"AAQA,OAAO,KAAK,EAAE,aAAa,EAAE,OAAO,EAAE,WAAW,EAAE,MAAM,UAAU,CAAC;AAapE;;GAEG;AACH,wBAAsB,kBAAkB,CACvC,IAAI,EAAE,WAAW,GAAG;IACnB,IAAI,CAAC,EAAE,IAAI,GAAG,WAAW,CAAC;IAC1B,MAAM,CAAC,EAAE,OAAO,CAAC;CACjB,EACD,OAAO,CAAC,EAAE,OAAO,GAAG;IACnB,yFAAyF;IACzF,SAAS,CAAC,EAAE,MAAM,GAAG,aAAa,CAAC;IACnC,sCAAsC;IACtC,QAAQ,CAAC,EAAE,aAAa,CAAC;IACzB,cAAc,CAAC,EAAE,OAAO,CAAC;CACzB,GACC,OAAO,CAAC;IAAE,GAAG,EAAE,MAAM,CAAC;IAAC,IAAI,EAAE,WAAW,CAAA;CAAE,CAAC,CAwG7C"}
+{"version":3,"file":"makeRequestOptions.d.ts","sourceRoot":"","sources":["../../../src/lib/makeRequestOptions.ts"],"names":[],"mappings":"AAWA,OAAO,KAAK,EAAE,aAAa,EAAE,OAAO,EAAE,WAAW,EAAE,MAAM,UAAU,CAAC;AAapE;;GAEG;AACH,wBAAsB,kBAAkB,CACvC,IAAI,EAAE,WAAW,GAAG;IACnB,IAAI,CAAC,EAAE,IAAI,GAAG,WAAW,CAAC;IAC1B,MAAM,CAAC,EAAE,OAAO,CAAC;CACjB,EACD,OAAO,CAAC,EAAE,OAAO,GAAG;IACnB,sCAAsC;IACtC,QAAQ,CAAC,EAAE,aAAa,CAAC;IACzB,cAAc,CAAC,EAAE,OAAO,CAAC;CACzB,GACC,OAAO,CAAC;IAAE,GAAG,EAAE,MAAM,CAAC;IAAC,IAAI,EAAE,WAAW,CAAA;CAAE,CAAC,CA6G7C"}

package/dist/src/providers/black-forest-labs.d.ts

@@ -0,0 +1,18 @@
+export declare const BLACKFORESTLABS_AI_API_BASE_URL = "https://api.us1.bfl.ai/v1";
+/**
+ * See the registered mapping of HF model ID => Black Forest Labs model ID here:
+ *
+ * https://huggingface.co/api/partners/blackforestlabs/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 Black Forest Labs 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 Black Forest Labs, please open an issue on the present repo
+ * and we will tag Black Forest Labs team members.
+ *
+ * Thanks!
+ */
+//# sourceMappingURL=black-forest-labs.d.ts.map

package/dist/src/providers/black-forest-labs.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"black-forest-labs.d.ts","sourceRoot":"","sources":["../../../src/providers/black-forest-labs.ts"],"names":[],"mappings":"AAAA,eAAO,MAAM,+BAA+B,8BAA8B,CAAC;AAE3E;;;;;;;;;;;;;;;GAeG"}

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

@@ -1 +1 @@
-{"version":3,"file":"consts.d.ts","sourceRoot":"","sources":["../../../src/providers/consts.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,iBAAiB,EAAE,MAAM,UAAU,CAAC;AAClD,OAAO,EAAE,KAAK,OAAO,EAAE,MAAM,UAAU,CAAC;AAExC,KAAK,UAAU,GAAG,MAAM,CAAC;AACzB;;;;;;GAMG;AACH,eAAO,MAAM,0BAA0B,EAAE,MAAM,CAAC,iBAAiB,EAAE,MAAM,CAAC,OAAO,EAAE,UAAU,CAAC,CAc7F,CAAC"}
+{"version":3,"file":"consts.d.ts","sourceRoot":"","sources":["../../../src/providers/consts.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,iBAAiB,EAAE,MAAM,UAAU,CAAC;AAClD,OAAO,EAAE,KAAK,OAAO,EAAE,MAAM,UAAU,CAAC;AAExC,KAAK,UAAU,GAAG,MAAM,CAAC;AACzB;;;;;;GAMG;AACH,eAAO,MAAM,0BAA0B,EAAE,MAAM,CAAC,iBAAiB,EAAE,MAAM,CAAC,OAAO,EAAE,UAAU,CAAC,CAiB7F,CAAC"}

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

@@ -0,0 +1,18 @@
+export declare const HYPERBOLIC_API_BASE_URL = "https://api.hyperbolic.xyz";
+/**
+ * See the registered mapping of HF model ID => Hyperbolic model ID here:
+ *
+ * https://huggingface.co/api/partners/hyperbolic/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 Hyperbolic 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 Hyperbolic, please open an issue on the present repo
+ * and we will tag Hyperbolic team members.
+ *
+ * Thanks!
+ */
+//# sourceMappingURL=hyperbolic.d.ts.map

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

@@ -0,0 +1 @@
+{"version":3,"file":"hyperbolic.d.ts","sourceRoot":"","sources":["../../../src/providers/hyperbolic.ts"],"names":[],"mappings":"AAAA,eAAO,MAAM,uBAAuB,+BAA+B,CAAC;AAEpE;;;;;;;;;;;;;;;GAeG"}

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

@@ -0,0 +1,18 @@
+export declare const NOVITA_API_BASE_URL = "https://api.novita.ai/v3/openai";
+/**
+ * See the registered mapping of HF model ID => Novita model ID here:
+ *
+ * https://huggingface.co/api/partners/novita/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 Novita 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 Novita, please open an issue on the present repo
+ * and we will tag Novita team members.
+ *
+ * Thanks!
+ */
+//# sourceMappingURL=novita.d.ts.map

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

@@ -0,0 +1 @@
+{"version":3,"file":"novita.d.ts","sourceRoot":"","sources":["../../../src/providers/novita.ts"],"names":[],"mappings":"AAAA,eAAO,MAAM,mBAAmB,oCAAoC,CAAC;AAErE;;;;;;;;;;;;;;;GAeG"}

package/dist/src/tasks/cv/textToImage.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"textToImage.d.ts","sourceRoot":"","sources":["../../../../src/tasks/cv/textToImage.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,gBAAgB,EAAqB,MAAM,oBAAoB,CAAC;AAE9E,OAAO,KAAK,EAAE,QAAQ,EAAqB,OAAO,EAAE,MAAM,aAAa,CAAC;AAIxE,MAAM,MAAM,eAAe,GAAG,QAAQ,GAAG,gBAAgB,CAAC;AA0B1D;;;GAGG;AACH,wBAAsB,WAAW,CAAC,IAAI,EAAE,eAAe,EAAE,OAAO,CAAC,EAAE,OAAO,GAAG,OAAO,CAAC,IAAI,CAAC,CAqCzF"}
+{"version":3,"file":"textToImage.d.ts","sourceRoot":"","sources":["../../../../src/tasks/cv/textToImage.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,gBAAgB,EAAqB,MAAM,oBAAoB,CAAC;AAE9E,OAAO,KAAK,EAAE,QAAQ,EAAqB,OAAO,EAAE,MAAM,aAAa,CAAC;AAKxE,MAAM,MAAM,eAAe,GAAG,QAAQ,GAAG,gBAAgB,CAAC;AAkC1D;;;GAGG;AACH,wBAAsB,WAAW,CAAC,IAAI,EAAE,eAAe,EAAE,OAAO,CAAC,EAAE,OAAO,GAAG,OAAO,CAAC,IAAI,CAAC,CAyDzF"}

package/dist/src/tasks/nlp/featureExtraction.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"featureExtraction.d.ts","sourceRoot":"","sources":["../../../../src/tasks/nlp/featureExtraction.ts"],"names":[],"mappings":"AAEA,OAAO,KAAK,EAAE,QAAQ,EAAE,OAAO,EAAE,MAAM,aAAa,CAAC;AAGrD,MAAM,MAAM,qBAAqB,GAAG,QAAQ,GAAG;IAC9C;;;;;OAKG;IACH,MAAM,EAAE,MAAM,GAAG,MAAM,EAAE,CAAC;CAC1B,CAAC;AAEF;;GAEG;AACH,MAAM,MAAM,uBAAuB,GAAG,CAAC,MAAM,GAAG,MAAM,EAAE,GAAG,MAAM,EAAE,EAAE,CAAC,EAAE,CAAC;AAEzE;;GAEG;AACH,wBAAsB,iBAAiB,CACtC,IAAI,EAAE,qBAAqB,EAC3B,OAAO,CAAC,EAAE,OAAO,GACf,OAAO,CAAC,uBAAuB,CAAC,CAyBlC"}
+{"version":3,"file":"featureExtraction.d.ts","sourceRoot":"","sources":["../../../../src/tasks/nlp/featureExtraction.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,QAAQ,EAAE,OAAO,EAAE,MAAM,aAAa,CAAC;AAGrD,MAAM,MAAM,qBAAqB,GAAG,QAAQ,GAAG;IAC9C;;;;;OAKG;IACH,MAAM,EAAE,MAAM,GAAG,MAAM,EAAE,CAAC;CAC1B,CAAC;AAEF;;GAEG;AACH,MAAM,MAAM,uBAAuB,GAAG,CAAC,MAAM,GAAG,MAAM,EAAE,GAAG,MAAM,EAAE,EAAE,CAAC,EAAE,CAAC;AAEzE;;GAEG;AACH,wBAAsB,iBAAiB,CACtC,IAAI,EAAE,qBAAqB,EAC3B,OAAO,CAAC,EAAE,OAAO,GACf,OAAO,CAAC,uBAAuB,CAAC,CAsBlC"}

package/dist/src/tasks/nlp/sentenceSimilarity.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"sentenceSimilarity.d.ts","sourceRoot":"","sources":["../../../../src/tasks/nlp/sentenceSimilarity.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,uBAAuB,EAAE,wBAAwB,EAAE,MAAM,oBAAoB,CAAC;AAG5F,OAAO,KAAK,EAAE,QAAQ,EAAE,OAAO,EAAE,MAAM,aAAa,CAAC;AAIrD,MAAM,MAAM,sBAAsB,GAAG,QAAQ,GAAG,uBAAuB,CAAC;AAExE;;GAEG;AACH,wBAAsB,kBAAkB,CACvC,IAAI,EAAE,sBAAsB,EAC5B,OAAO,CAAC,EAAE,OAAO,GACf,OAAO,CAAC,wBAAwB,CAAC,CAanC"}
+{"version":3,"file":"sentenceSimilarity.d.ts","sourceRoot":"","sources":["../../../../src/tasks/nlp/sentenceSimilarity.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,uBAAuB,EAAE,wBAAwB,EAAE,MAAM,oBAAoB,CAAC;AAE5F,OAAO,KAAK,EAAE,QAAQ,EAAE,OAAO,EAAE,MAAM,aAAa,CAAC;AAIrD,MAAM,MAAM,sBAAsB,GAAG,QAAQ,GAAG,uBAAuB,CAAC;AAExE;;GAEG;AACH,wBAAsB,kBAAkB,CACvC,IAAI,EAAE,sBAAsB,EAC5B,OAAO,CAAC,EAAE,OAAO,GACf,OAAO,CAAC,wBAAwB,CAAC,CAWnC"}

package/dist/src/tasks/nlp/textGeneration.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"textGeneration.d.ts","sourceRoot":"","sources":["../../../../src/tasks/nlp/textGeneration.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAEX,mBAAmB,EACnB,oBAAoB,EAEpB,MAAM,oBAAoB,CAAC;AAE5B,OAAO,KAAK,EAAE,QAAQ,EAAE,OAAO,EAAE,MAAM,aAAa,CAAC;AAIrD,YAAY,EAAE,mBAAmB,EAAE,oBAAoB,EAAE,CAAC;AAY1D;;GAEG;AACH,wBAAsB,cAAc,CACnC,IAAI,EAAE,QAAQ,GAAG,mBAAmB,EACpC,OAAO,CAAC,EAAE,OAAO,GACf,OAAO,CAAC,oBAAoB,CAAC,CA+B/B"}
+{"version":3,"file":"textGeneration.d.ts","sourceRoot":"","sources":["../../../../src/tasks/nlp/textGeneration.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAEX,mBAAmB,EACnB,oBAAoB,EAEpB,MAAM,oBAAoB,CAAC;AAE5B,OAAO,KAAK,EAAE,QAAQ,EAAE,OAAO,EAAE,MAAM,aAAa,CAAC;AAKrD,YAAY,EAAE,mBAAmB,EAAE,oBAAoB,EAAE,CAAC;AAkB1D;;GAEG;AACH,wBAAsB,cAAc,CACnC,IAAI,EAAE,QAAQ,GAAG,mBAAmB,EACpC,OAAO,CAAC,EAAE,OAAO,GACf,OAAO,CAAC,oBAAoB,CAAC,CAuD/B"}

package/dist/src/types.d.ts

@@ -22,7 +22,7 @@ export interface Options {
     includeCredentials?: string | boolean;
 }
 export type InferenceTask = Exclude<PipelineType, "other">;
-export declare const INFERENCE_PROVIDERS: readonly ["fal-ai", "fireworks-ai", "nebius", "hf-inference", "replicate", "sambanova", "together"];
+export declare const INFERENCE_PROVIDERS: readonly ["black-forest-labs", "fal-ai", "fireworks-ai", "hf-inference", "hyperbolic", "nebius", "novita", "replicate", "sambanova", "together"];
 export type InferenceProvider = (typeof INFERENCE_PROVIDERS)[number];
 export interface BaseArgs {
     /**

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,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;IAEzB;;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,qGAQtB,CAAC;AACX,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"}
+{"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;IAEzB;;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,kJAWtB,CAAC;AAEX,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/src/utils/delay.d.ts

@@ -0,0 +1,2 @@
+export declare function delay(ms: number): Promise<void>;
+//# sourceMappingURL=delay.d.ts.map

package/dist/src/utils/delay.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"delay.d.ts","sourceRoot":"","sources":["../../../src/utils/delay.ts"],"names":[],"mappings":"AAAA,wBAAgB,KAAK,CAAC,EAAE,EAAE,MAAM,GAAG,OAAO,CAAC,IAAI,CAAC,CAI/C"}

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

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@huggingface/inference",
-  "version": "3.3.2",
+  "version": "3.3.4",
   "packageManager": "pnpm@8.10.5",
   "license": "MIT",
   "author": "Tim Mikeladze <tim.mikeladze@gmail.com>",

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/lib/makeRequestOptions.ts

@@ -1,17 +1,20 @@
-import { HF_HUB_URL } from "../config";
+import { HF_HUB_URL, HF_ROUTER_URL } from "../config";
 import { FAL_AI_API_BASE_URL } from "../providers/fal-ai";
 import { NEBIUS_API_BASE_URL } from "../providers/nebius";
 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 { NOVITA_API_BASE_URL } from "../providers/novita";
 import { FIREWORKS_AI_API_BASE_URL } from "../providers/fireworks-ai";
+import { HYPERBOLIC_API_BASE_URL } from "../providers/hyperbolic";
+import { BLACKFORESTLABS_AI_API_BASE_URL } from "../providers/black-forest-labs";
 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
@@ -28,8 +31,6 @@ export async function makeRequestOptions(
 		stream?: boolean;
 	},
 	options?: Options & {
-		/** When a model can be used for multiple tasks, and we want to run a non-default task */
-		forceTask?: string | InferenceTask;
 		/** To load default model if needed */
 		taskHint?: InferenceTask;
 		chatCompletion?: boolean;
@@ -39,14 +40,11 @@ export async function makeRequestOptions(
 	let otherArgs = remainingArgs;
 	const provider = maybeProvider ?? "hf-inference";
 
-	const { forceTask, includeCredentials, taskHint, chatCompletion } = options ?? {};
+	const { includeCredentials, taskHint, chatCompletion } = options ?? {};
 
 	if (endpointUrl && provider !== "hf-inference") {
 		throw new Error(`Cannot use endpointUrl with a third-party provider.`);
 	}
-	if (forceTask && provider !== "hf-inference") {
-		throw new Error(`Cannot use forceTask with a third-party provider.`);
-	}
 	if (maybeModel && isUrl(maybeModel)) {
 		throw new Error(`Model URLs are no longer supported. Use endpointUrl instead.`);
 	}
@@ -77,7 +75,6 @@ export async function makeRequestOptions(
 		: makeUrl({
 				authMethod,
 				chatCompletion: chatCompletion ?? false,
-				forceTask,
 				model,
 				provider: provider ?? "hf-inference",
 				taskHint,
@@ -85,8 +82,13 @@ export async function makeRequestOptions(
 
 	const headers: Record<string, string> = {};
 	if (accessToken) {
-		headers["Authorization"] =
-			provider === "fal-ai" && authMethod === "provider-key" ? `Key ${accessToken}` : `Bearer ${accessToken}`;
+		if (provider === "fal-ai" && authMethod === "provider-key") {
+			headers["Authorization"] = `Key ${accessToken}`;
+		} else if (provider === "black-forest-labs" && authMethod === "provider-key") {
+			headers["X-Key"] = accessToken;
+		} else {
+			headers["Authorization"] = `Bearer ${accessToken}`;
+		}
 	}
 
 	// e.g. @huggingface/inference/3.1.3
@@ -131,7 +133,11 @@ export async function makeRequestOptions(
 			? args.data
 			: JSON.stringify({
 					...otherArgs,
-					...(chatCompletion || provider === "together" || provider === "nebius" ? { model } : undefined),
+					...(taskHint === "text-to-image" && provider === "hyperbolic"
+						? { model_name: model }
+						: chatCompletion || provider === "together" || provider === "nebius" || provider === "hyperbolic"
+						  ? { model }
+						  : undefined),
 			  }),
 		...(credentials ? { credentials } : undefined),
 		signal: options?.signal,
@@ -146,7 +152,6 @@ function makeUrl(params: {
 	model: string;
 	provider: InferenceProvider;
 	taskHint: InferenceTask | undefined;
-	forceTask?: string | InferenceTask;
 }): string {
 	if (params.authMethod === "none" && params.provider !== "hf-inference") {
 		throw new Error("Authentication is required when requesting a third-party provider. Please provide accessToken");
@@ -154,6 +159,12 @@ function makeUrl(params: {
 
 	const shouldProxy = params.provider !== "hf-inference" && params.authMethod !== "provider-key";
 	switch (params.provider) {
+		case "black-forest-labs": {
+			const baseUrl = shouldProxy
+				? HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", params.provider)
+				: BLACKFORESTLABS_AI_API_BASE_URL;
+			return `${baseUrl}/${params.model}`;
+		}
 		case "fal-ai": {
 			const baseUrl = shouldProxy
 				? HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", params.provider)
@@ -213,6 +224,7 @@ function makeUrl(params: {
 			}
 			return baseUrl;
 		}
+
 		case "fireworks-ai": {
 			const baseUrl = shouldProxy
 				? HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", params.provider)
@@ -222,15 +234,38 @@ function makeUrl(params: {
 			}
 			return baseUrl;
 		}
+		case "hyperbolic": {
+			const baseUrl = shouldProxy
+				? HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", params.provider)
+				: HYPERBOLIC_API_BASE_URL;
+
+			if (params.taskHint === "text-to-image") {
+				return `${baseUrl}/v1/images/generations`;
+			}
+			return `${baseUrl}/v1/chat/completions`;
+		}
+		case "novita": {
+			const baseUrl = shouldProxy
+				? HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", params.provider)
+				: NOVITA_API_BASE_URL;
+			if (params.taskHint === "text-generation") {
+				if (params.chatCompletion) {
+					return `${baseUrl}/chat/completions`;
+				}
+				return `${baseUrl}/completions`;
+			}
+			return baseUrl;
+		}
 		default: {
 			const baseUrl = HF_HUB_INFERENCE_PROXY_TEMPLATE.replaceAll("{{PROVIDER}}", "hf-inference");
-			const url = params.forceTask
-				? `${baseUrl}/pipeline/${params.forceTask}/${params.model}`
-				: `${baseUrl}/models/${params.model}`;
+			if (params.taskHint && ["feature-extraction", "sentence-similarity"].includes(params.taskHint)) {
+				/// when deployed on hf-inference, those two tasks are automatically compatible with one another.
+				return `${baseUrl}/pipeline/${params.taskHint}/${params.model}`;
+			}
 			if (params.taskHint === "text-generation" && params.chatCompletion) {
-				return url + `/v1/chat/completions`;
+				return `${baseUrl}/models/${params.model}/v1/chat/completions`;
 			}
-			return url;
+			return `${baseUrl}/models/${params.model}`;
 		}
 	}
 }

package/src/providers/black-forest-labs.ts

@@ -0,0 +1,18 @@
+export const BLACKFORESTLABS_AI_API_BASE_URL = "https://api.us1.bfl.ai/v1";
+
+/**
+ * See the registered mapping of HF model ID => Black Forest Labs model ID here:
+ *
+ * https://huggingface.co/api/partners/blackforestlabs/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 Black Forest Labs 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 Black Forest Labs, please open an issue on the present repo
+ * and we will tag Black Forest Labs team members.
+ *
+ * Thanks!
+ */

package/src/providers/consts.ts

@@ -16,11 +16,14 @@ export const HARDCODED_MODEL_ID_MAPPING: Record<InferenceProvider, Record<ModelI
 	 * Example:
 	 * "Qwen/Qwen2.5-Coder-32B-Instruct": "Qwen2.5-Coder-32B-Instruct",
 	 */
+	"black-forest-labs": {},
 	"fal-ai": {},
 	"fireworks-ai": {},
 	"hf-inference": {},
+	hyperbolic: {},
 	nebius: {},
 	replicate: {},
 	sambanova: {},
 	together: {},
+	novita: {},
 };

package/src/providers/hyperbolic.ts

@@ -0,0 +1,18 @@
+export const HYPERBOLIC_API_BASE_URL = "https://api.hyperbolic.xyz";
+
+/**
+ * See the registered mapping of HF model ID => Hyperbolic model ID here:
+ *
+ * https://huggingface.co/api/partners/hyperbolic/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 Hyperbolic 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 Hyperbolic, please open an issue on the present repo
+ * and we will tag Hyperbolic team members.
+ *
+ * Thanks!
+ */

package/src/providers/novita.ts

@@ -0,0 +1,18 @@
+export const NOVITA_API_BASE_URL = "https://api.novita.ai/v3/openai";
+
+/**
+ * See the registered mapping of HF model ID => Novita model ID here:
+ *
+ * https://huggingface.co/api/partners/novita/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 Novita 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 Novita, please open an issue on the present repo
+ * and we will tag Novita team members.
+ *
+ * Thanks!
+ */

package/src/tasks/cv/textToImage.ts

@@ -3,6 +3,7 @@ import { InferenceOutputError } from "../../lib/InferenceOutputError";
 import type { BaseArgs, InferenceProvider, Options } from "../../types";
 import { omit } from "../../utils/omit";
 import { request } from "../custom/request";
+import { delay } from "../../utils/delay";
 
 export type TextToImageArgs = BaseArgs & TextToImageInput;
 
@@ -14,6 +15,14 @@ interface Base64ImageGeneration {
 interface OutputUrlImageGeneration {
 	output: string[];
 }
+interface HyperbolicTextToImageOutput {
+	images: Array<{ image: string }>;
+}
+
+interface BlackForestLabsResponse {
+	id: string;
+	polling_url: string;
+}
 
 function getResponseFormatArg(provider: InferenceProvider) {
 	switch (provider) {
@@ -44,16 +53,36 @@ export async function textToImage(args: TextToImageArgs, options?: Options): Pro
 					...getResponseFormatArg(args.provider),
 					prompt: args.inputs,
 			  };
-	const res = await request<TextToImageOutput | Base64ImageGeneration | OutputUrlImageGeneration>(payload, {
+	const res = await request<
+		| TextToImageOutput
+		| Base64ImageGeneration
+		| OutputUrlImageGeneration
+		| BlackForestLabsResponse
+		| HyperbolicTextToImageOutput
+	>(payload, {
 		...options,
 		taskHint: "text-to-image",
 	});
 
 	if (res && typeof res === "object") {
+		if (args.provider === "black-forest-labs" && "polling_url" in res && typeof res.polling_url === "string") {
+			return await pollBflResponse(res.polling_url);
+		}
 		if (args.provider === "fal-ai" && "images" in res && Array.isArray(res.images) && res.images[0].url) {
 			const image = await fetch(res.images[0].url);
 			return await image.blob();
 		}
+		if (
+			args.provider === "hyperbolic" &&
+			"images" in res &&
+			Array.isArray(res.images) &&
+			res.images[0] &&
+			typeof res.images[0].image === "string"
+		) {
+			const base64Response = await fetch(`data:image/jpeg;base64,${res.images[0].image}`);
+			const blob = await base64Response.blob();
+			return blob;
+		}
 		if ("data" in res && Array.isArray(res.data) && res.data[0].b64_json) {
 			const base64Data = res.data[0].b64_json;
 			const base64Response = await fetch(`data:image/jpeg;base64,${base64Data}`);
@@ -72,3 +101,33 @@ export async function textToImage(args: TextToImageArgs, options?: Options): Pro
 	}
 	return res;
 }
+
+async function pollBflResponse(url: string): Promise<Blob> {
+	const urlObj = new URL(url);
+	for (let step = 0; step < 5; step++) {
+		await delay(1000);
+		console.debug(`Polling Black Forest Labs API for the result... ${step + 1}/5`);
+		urlObj.searchParams.set("attempt", step.toString(10));
+		const resp = await fetch(urlObj, { headers: { "Content-Type": "application/json" } });
+		if (!resp.ok) {
+			throw new InferenceOutputError("Failed to fetch result from black forest labs API");
+		}
+		const payload = await resp.json();
+		if (
+			typeof payload === "object" &&
+			payload &&
+			"status" in payload &&
+			typeof payload.status === "string" &&
+			payload.status === "Ready" &&
+			"result" in payload &&
+			typeof payload.result === "object" &&
+			payload.result &&
+			"sample" in payload.result &&
+			typeof payload.result.sample === "string"
+		) {
+			const image = await fetch(payload.result.sample);
+			return await image.blob();
+		}
+	}
+	throw new InferenceOutputError("Failed to fetch result from black forest labs API");
+}

package/src/tasks/nlp/featureExtraction.ts

@@ -1,5 +1,4 @@
 import { InferenceOutputError } from "../../lib/InferenceOutputError";
-import { getDefaultTask } from "../../lib/getDefaultTask";
 import type { BaseArgs, Options } from "../../types";
 import { request } from "../custom/request";
 
@@ -25,12 +24,9 @@ export async function featureExtraction(
 	args: FeatureExtractionArgs,
 	options?: Options
 ): Promise<FeatureExtractionOutput> {
-	const defaultTask = args.model ? await getDefaultTask(args.model, args.accessToken, options) : undefined;
-
 	const res = await request<FeatureExtractionOutput>(args, {
 		...options,
 		taskHint: "feature-extraction",
-		...(defaultTask === "sentence-similarity" && { forceTask: "feature-extraction" }),
 	});
 	let isValidOutput = true;
 

package/src/tasks/nlp/sentenceSimilarity.ts

@@ -1,6 +1,5 @@
 import type { SentenceSimilarityInput, SentenceSimilarityOutput } from "@huggingface/tasks";
 import { InferenceOutputError } from "../../lib/InferenceOutputError";
-import { getDefaultTask } from "../../lib/getDefaultTask";
 import type { BaseArgs, Options } from "../../types";
 import { request } from "../custom/request";
 import { omit } from "../../utils/omit";
@@ -14,11 +13,9 @@ export async function sentenceSimilarity(
 	args: SentenceSimilarityArgs,
 	options?: Options
 ): Promise<SentenceSimilarityOutput> {
-	const defaultTask = args.model ? await getDefaultTask(args.model, args.accessToken, options) : undefined;
 	const res = await request<SentenceSimilarityOutput>(prepareInput(args), {
 		...options,
 		taskHint: "sentence-similarity",
-		...(defaultTask === "feature-extraction" && { forceTask: "sentence-similarity" }),
 	});
 
 	const isValidOutput = Array.isArray(res) && res.every((x) => typeof x === "number");

package/src/tasks/nlp/textGeneration.ts

@@ -8,6 +8,7 @@ import { InferenceOutputError } from "../../lib/InferenceOutputError";
 import type { BaseArgs, Options } from "../../types";
 import { toArray } from "../../utils/toArray";
 import { request } from "../custom/request";
+import { omit } from "../../utils/omit";
 
 export type { TextGenerationInput, TextGenerationOutput };
 
@@ -21,6 +22,12 @@ interface TogeteherTextCompletionOutput extends Omit<ChatCompletionOutput, "choi
 	}>;
 }
 
+interface HyperbolicTextCompletionOutput extends Omit<ChatCompletionOutput, "choices"> {
+	choices: Array<{
+		message: { content: string };
+	}>;
+}
+
 /**
  * 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).
  */
@@ -43,6 +50,30 @@ export async function textGeneration(
 		return {
 			generated_text: completion.text,
 		};
+	} else if (args.provider === "hyperbolic") {
+		const payload = {
+			messages: [{ content: args.inputs, role: "user" }],
+			...(args.parameters
+				? {
+						max_tokens: args.parameters.max_new_tokens,
+						...omit(args.parameters, "max_new_tokens"),
+				  }
+				: undefined),
+			...omit(args, ["inputs", "parameters"]),
+		};
+		const raw = await request<HyperbolicTextCompletionOutput>(payload, {
+			...options,
+			taskHint: "text-generation",
+		});
+		const isValidOutput =
+			typeof raw === "object" && "choices" in raw && Array.isArray(raw?.choices) && typeof raw?.model === "string";
+		if (!isValidOutput) {
+			throw new InferenceOutputError("Expected ChatCompletionOutput");
+		}
+		const completion = raw.choices[0];
+		return {
+			generated_text: completion.message.content,
+		};
 	} else {
 		const res = toArray(
 			await request<TextGenerationOutput | TextGenerationOutput[]>(args, {

package/src/types.ts

@@ -29,14 +29,18 @@ export interface Options {
 export type InferenceTask = Exclude<PipelineType, "other">;
 
 export const INFERENCE_PROVIDERS = [
+	"black-forest-labs",
 	"fal-ai",
 	"fireworks-ai",
-	"nebius",
 	"hf-inference",
+	"hyperbolic",
+	"nebius",
+	"novita",
 	"replicate",
 	"sambanova",
 	"together",
 ] as const;
+
 export type InferenceProvider = (typeof INFERENCE_PROVIDERS)[number];
 
 export interface BaseArgs {

package/src/utils/delay.ts

@@ -0,0 +1,5 @@
+export function delay(ms: number): Promise<void> {
+	return new Promise((resolve) => {
+		setTimeout(() => resolve(), ms);
+	});
+}