@huggingface/inference - npm Package Compare versions

Comparing version 3.2.0 to 3.3.0

dist/index.js

@@ -60,2 +60,5 @@ var __defProp = Object.defineProperty;
 
+// src/providers/fireworks-ai.ts
+var FIREWORKS_AI_API_BASE_URL = "https://api.fireworks.ai/inference";
+
 // src/lib/isUrl.ts
@@ -68,3 +71,3 @@ function isUrl(modelOrUrl) {
 var name = "@huggingface/inference";
-var version = "3.2.0";
+var version = "3.3.0";
 
@@ -75,4 +78,12 @@ // src/providers/consts.ts
    * "HF model ID" => "Model ID on Inference Provider's side"
+   *
+   * Example:
+   * "Qwen/Qwen2.5-Coder-32B-Instruct": "Qwen2.5-Coder-32B-Instruct",
    */
-  // "Qwen/Qwen2.5-Coder-32B-Instruct": "Qwen2.5-Coder-32B-Instruct",
+  "fal-ai": {},
+  "fireworks-ai": {},
+  "hf-inference": {},
+  replicate: {},
+  sambanova: {},
+  together: {}
 };
@@ -90,4 +101,4 @@
   const task = options.taskHint === "text-generation" && options.chatCompletion ? "conversational" : options.taskHint;
-  if (HARDCODED_MODEL_ID_MAPPING[params.model]) {
-    return HARDCODED_MODEL_ID_MAPPING[params.model];
+  if (HARDCODED_MODEL_ID_MAPPING[params.provider]?.[params.model]) {
+    return HARDCODED_MODEL_ID_MAPPING[params.provider][params.model];
   }
@@ -243,2 +254,9 @@ let inferenceProviderMapping;
     }
+    case "fireworks-ai": {
+      const baseUrl = shouldProxy ? HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", params.provider) : FIREWORKS_AI_API_BASE_URL;
+      if (params.taskHint === "text-generation" && params.chatCompletion) {
+        return `${baseUrl}/v1/chat/completions`;
+      }
+      return baseUrl;
+    }
     default: {
@@ -1222,3 +1240,10 @@ const baseUrl = HF_HUB_INFERENCE_PROXY_TEMPLATE.replaceAll("{{PROVIDER}}", "hf-inference");
 // src/types.ts
-var INFERENCE_PROVIDERS = ["fal-ai", "replicate", "sambanova", "together", "hf-inference"];
+var INFERENCE_PROVIDERS = [
+  "fal-ai",
+  "fireworks-ai",
+  "hf-inference",
+  "replicate",
+  "sambanova",
+  "together"
+];
 export {
@@ -1225,0 +1250,0 @@ HfInference,

dist/src/providers/consts.d.ts

@@ -1,2 +0,3 @@
-import type { ModelId } from "../types";
+import type { InferenceProvider } from "../types";
+import { type ModelId } from "../types";
 type ProviderId = string;
@@ -7,5 +8,7 @@ /**
  * you can add it to the following dictionary, for dev purposes.
+ *
+ * We also inject into this dictionary from tests.
  */
-export declare const HARDCODED_MODEL_ID_MAPPING: Record<ModelId, ProviderId>;
+export declare const HARDCODED_MODEL_ID_MAPPING: Record<InferenceProvider, Record<ModelId, ProviderId>>;
 export {};
 //# sourceMappingURL=consts.d.ts.map

dist/src/types.d.ts

@@ -41,3 +41,3 @@ import type { ChatCompletionInput, PipelineType } from "@huggingface/tasks";
 export type InferenceTask = Exclude<PipelineType, "other">;
-export declare const INFERENCE_PROVIDERS: readonly ["fal-ai", "replicate", "sambanova", "together", "hf-inference"];
+export declare const INFERENCE_PROVIDERS: readonly ["fal-ai", "fireworks-ai", "hf-inference", "replicate", "sambanova", "together"];
 export type InferenceProvider = (typeof INFERENCE_PROVIDERS)[number];
@@ -44,0 +44,0 @@ export interface BaseArgs {

package.json

 {
   "name": "@huggingface/inference",
-  "version": "3.2.0",
+  "version": "3.3.0",
   "packageManager": "pnpm@8.10.5",
@@ -5,0 +5,0 @@ "license": "MIT",

README.md

@@ -49,3 +49,8 @@ # 🤗 Hugging Face Inference
 
-Currently, we support the following providers: [Fal.ai](https://fal.ai), [Replicate](https://replicate.com), [Together](https://together.xyz) and [Sambanova](https://sambanova.ai).
+Currently, we support the following providers:
+- [Fal.ai](https://fal.ai)
+- [Fireworks AI](https://fireworks.ai)
+- [Replicate](https://replicate.com)
+- [Sambanova](https://sambanova.ai)
+- [Together](https://together.xyz)
 
@@ -68,6 +73,7 @@ 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.
 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](./src/providers/fal-ai.ts)
-- [Replicate supported models](./src/providers/replicate.ts)
-- [Sambanova supported models](./src/providers/sambanova.ts)
-- [Together supported models](./src/providers/together.ts)
+- [Fal.ai supported models](https://huggingface.co/api/partners/fal-ai/models)
+- [Fireworks AI supported models](https://huggingface.co/api/partners/fireworks-ai/models)
+- [Replicate supported models](https://huggingface.co/api/partners/replicate/models)
+- [Sambanova supported models](https://huggingface.co/api/partners/sambanova/models)
+- [Together supported models](https://huggingface.co/api/partners/together/models)
 - [HF Inference API (serverless)](https://huggingface.co/models?inference=warm&sort=trending)
@@ -74,0 +80,0 @@

src/lib/getProviderModelId.ts

@@ -33,4 +33,4 @@ import type { WidgetType } from "@huggingface/tasks";
 	// 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];
+	if (HARDCODED_MODEL_ID_MAPPING[params.provider]?.[params.model]) {
+		return HARDCODED_MODEL_ID_MAPPING[params.provider][params.model];
 	}
@@ -37,0 +37,0 @@

src/lib/makeRequestOptions.ts

@@ -6,2 +6,3 @@ import { HF_HUB_URL, HF_ROUTER_URL } from "../config";
 import { TOGETHER_API_BASE_URL } from "../providers/together";
+import { FIREWORKS_AI_API_BASE_URL } from "../providers/fireworks-ai";
 import type { InferenceProvider } from "../types";
@@ -212,2 +213,11 @@ import type { InferenceTask, Options, RequestArgs } from "../types";
 		}
+		case "fireworks-ai": {
+			const baseUrl = shouldProxy
+				? HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", params.provider)
+				: FIREWORKS_AI_API_BASE_URL;
+			if (params.taskHint === "text-generation" && params.chatCompletion) {
+				return `${baseUrl}/v1/chat/completions`;
+			}
+			return baseUrl;
+		}
 		default: {
@@ -214,0 +224,0 @@ const baseUrl = HF_HUB_INFERENCE_PROXY_TEMPLATE.replaceAll("{{PROVIDER}}", "hf-inference");

src/providers/consts.ts

@@ -1,5 +0,5 @@
-import type { ModelId } from "../types";
+import type { InferenceProvider } from "../types";
+import { type ModelId } from "../types";
 
 type ProviderId = string;
-
 /**
@@ -9,8 +9,18 @@ * 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 following dictionary, for dev purposes.
+ *
+ * We also inject into this dictionary from tests.
  */
-export const HARDCODED_MODEL_ID_MAPPING: Record<ModelId, ProviderId> = {
+export const HARDCODED_MODEL_ID_MAPPING: Record<InferenceProvider, Record<ModelId, ProviderId>> = {
 	/**
 	 * "HF model ID" => "Model ID on Inference Provider's side"
+	 *
+	 * Example:
+	 * "Qwen/Qwen2.5-Coder-32B-Instruct": "Qwen2.5-Coder-32B-Instruct",
 	 */
-	// "Qwen/Qwen2.5-Coder-32B-Instruct": "Qwen2.5-Coder-32B-Instruct",
+	"fal-ai": {},
+	"fireworks-ai": {},
+	"hf-inference": {},
+	replicate: {},
+	sambanova: {},
+	together: {},
 };

src/types.ts

@@ -47,3 +47,10 @@ import type { ChatCompletionInput, PipelineType } from "@huggingface/tasks";
 
-export const INFERENCE_PROVIDERS = ["fal-ai", "replicate", "sambanova", "together", "hf-inference"] as const;
+export const INFERENCE_PROVIDERS = [
+	"fal-ai",
+	"fireworks-ai",
+	"hf-inference",
+	"replicate",
+	"sambanova",
+	"together",
+] as const;
 export type InferenceProvider = (typeof INFERENCE_PROVIDERS)[number];
@@ -50,0 +57,0 @@

No alert changes

Improved metrics

Total package byte prevSize

276342

increased by1.55%

Number of package files

193

increased by1.58%

Lines of code

5878

increased by1.96%

Number of lines in readme file

668

increased by0.91%

No dependency changes