@huggingface/inference 3.3.3 → 3.3.5

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)

package/dist/index.cjs

@@ -115,9 +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("/");
@@ -125,7 +134,7 @@ function isUrl(modelOrUrl) {
 
 // package.json
 var name = "@huggingface/inference";
-var version = "3.3.3";
+var version = "3.3.5";
 
 // src/providers/consts.ts
 var HARDCODED_MODEL_ID_MAPPING = {
@@ -135,13 +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
@@ -195,13 +207,10 @@ 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.`);
   }
@@ -218,14 +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}/${version}`;
   headers["User-Agent"] = [ownUserAgent, typeof navigator !== "undefined" ? navigator.userAgent : void 0].filter((x) => x !== void 0).join(" ");
@@ -251,7 +265,7 @@ async function makeRequestOptions(args, options) {
     method: "POST",
     body: binary ? args.data : JSON.stringify({
       ...otherArgs,
-      ...chatCompletion2 || provider === "together" || provider === "nebius" ? { model } : void 0
+      ...taskHint === "text-to-image" && provider === "hyperbolic" ? { model_name: model } : chatCompletion2 || provider === "together" || provider === "nebius" || provider === "hyperbolic" ? { model } : void 0
     }),
     ...credentials ? { credentials } : void 0,
     signal: options?.signal
@@ -264,6 +278,10 @@ function makeUrl(params) {
   }
   const shouldProxy = params.provider !== "hf-inference" && params.authMethod !== "provider-key";
   switch (params.provider) {
+    case "black-forest-labs": {
+      const baseUrl = shouldProxy ? HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", params.provider) : BLACKFORESTLABS_AI_API_BASE_URL;
+      return `${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}/${params.model}`;
@@ -315,13 +333,32 @@ function makeUrl(params) {
       }
       return baseUrl;
     }
+    case "hyperbolic": {
+      const baseUrl = shouldProxy ? HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", params.provider) : HYPERBOLIC_API_BASE_URL;
+      if (params.taskHint === "text-to-image") {
+        return `${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}`;
     }
   }
 }
@@ -768,6 +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) {
@@ -795,17 +839,36 @@ 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, options?.outputType);
+    }
     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 (options?.outputType === "url") {
+        return res.images[0].url;
+      } else {
+        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") {
+      if (options?.outputType === "url") {
+        return `data:image/jpeg;base64,${res.images[0].image}`;
+      }
+      const base64Response = await fetch(`data:image/jpeg;base64,${res.images[0].image}`);
+      return await base64Response.blob();
     }
     if ("data" in res && Array.isArray(res.data) && res.data[0].b64_json) {
       const base64Data = res.data[0].b64_json;
+      if (options?.outputType === "url") {
+        return `data:image/jpeg;base64,${base64Data}`;
+      }
       const base64Response = await fetch(`data:image/jpeg;base64,${base64Data}`);
-      const blob = await base64Response.blob();
-      return blob;
+      return await base64Response.blob();
     }
     if ("output" in res && Array.isArray(res.output)) {
+      if (options?.outputType === "url") {
+        return res.output[0];
+      }
       const urlResponse = await fetch(res.output[0]);
       const blob = await urlResponse.blob();
       return blob;
@@ -815,8 +878,33 @@ async function textToImage(args, options) {
   if (!isValidOutput) {
     throw new InferenceOutputError("Expected Blob");
   }
+  if (options?.outputType === "url") {
+    const b64 = await res.arrayBuffer().then((buf) => Buffer.from(buf).toString("base64"));
+    return `data:image/jpeg;base64,${b64}`;
+  }
   return res;
 }
+async function pollBflResponse(url, outputType) {
+  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") {
+      if (outputType === "url") {
+        return payload.result.sample;
+      }
+      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) {
@@ -911,43 +999,11 @@ 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) => {
@@ -1000,11 +1056,9 @@ 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) {
@@ -1090,6 +1144,27 @@ 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, {
@@ -1302,10 +1377,13 @@ 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

@@ -60,9 +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("/");
@@ -70,7 +79,7 @@ function isUrl(modelOrUrl) {
 
 // package.json
 var name = "@huggingface/inference";
-var version = "3.3.3";
+var version = "3.3.5";
 
 // src/providers/consts.ts
 var HARDCODED_MODEL_ID_MAPPING = {
@@ -80,13 +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
@@ -140,13 +152,10 @@ 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.`);
   }
@@ -163,14 +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}/${version}`;
   headers["User-Agent"] = [ownUserAgent, typeof navigator !== "undefined" ? navigator.userAgent : void 0].filter((x) => x !== void 0).join(" ");
@@ -196,7 +210,7 @@ async function makeRequestOptions(args, options) {
     method: "POST",
     body: binary ? args.data : JSON.stringify({
       ...otherArgs,
-      ...chatCompletion2 || provider === "together" || provider === "nebius" ? { model } : void 0
+      ...taskHint === "text-to-image" && provider === "hyperbolic" ? { model_name: model } : chatCompletion2 || provider === "together" || provider === "nebius" || provider === "hyperbolic" ? { model } : void 0
     }),
     ...credentials ? { credentials } : void 0,
     signal: options?.signal
@@ -209,6 +223,10 @@ function makeUrl(params) {
   }
   const shouldProxy = params.provider !== "hf-inference" && params.authMethod !== "provider-key";
   switch (params.provider) {
+    case "black-forest-labs": {
+      const baseUrl = shouldProxy ? HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", params.provider) : BLACKFORESTLABS_AI_API_BASE_URL;
+      return `${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}/${params.model}`;
@@ -260,13 +278,32 @@ function makeUrl(params) {
       }
       return baseUrl;
     }
+    case "hyperbolic": {
+      const baseUrl = shouldProxy ? HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", params.provider) : HYPERBOLIC_API_BASE_URL;
+      if (params.taskHint === "text-to-image") {
+        return `${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}`;
     }
   }
 }
@@ -713,6 +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) {
@@ -740,17 +784,36 @@ 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, options?.outputType);
+    }
     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 (options?.outputType === "url") {
+        return res.images[0].url;
+      } else {
+        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") {
+      if (options?.outputType === "url") {
+        return `data:image/jpeg;base64,${res.images[0].image}`;
+      }
+      const base64Response = await fetch(`data:image/jpeg;base64,${res.images[0].image}`);
+      return await base64Response.blob();
     }
     if ("data" in res && Array.isArray(res.data) && res.data[0].b64_json) {
       const base64Data = res.data[0].b64_json;
+      if (options?.outputType === "url") {
+        return `data:image/jpeg;base64,${base64Data}`;
+      }
       const base64Response = await fetch(`data:image/jpeg;base64,${base64Data}`);
-      const blob = await base64Response.blob();
-      return blob;
+      return await base64Response.blob();
     }
     if ("output" in res && Array.isArray(res.output)) {
+      if (options?.outputType === "url") {
+        return res.output[0];
+      }
       const urlResponse = await fetch(res.output[0]);
       const blob = await urlResponse.blob();
       return blob;
@@ -760,8 +823,33 @@ async function textToImage(args, options) {
   if (!isValidOutput) {
     throw new InferenceOutputError("Expected Blob");
   }
+  if (options?.outputType === "url") {
+    const b64 = await res.arrayBuffer().then((buf) => Buffer.from(buf).toString("base64"));
+    return `data:image/jpeg;base64,${b64}`;
+  }
   return res;
 }
+async function pollBflResponse(url, outputType) {
+  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") {
+      if (outputType === "url") {
+        return payload.result.sample;
+      }
+      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) {
@@ -856,43 +944,11 @@ 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) => {
@@ -945,11 +1001,9 @@ 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) {
@@ -1035,6 +1089,27 @@ 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, {
@@ -1247,10 +1322,13 @@ 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/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;