Putting the provider arg more front'n'center (and other tweaks) (#1114

)

README.md

@@ -27,7 +27,7 @@ await uploadFile({
   }
 });
 
-// Use HF Inference API
+// Use HF Inference API, or external Inference Providers!
 
 await inference.chatCompletion({
   model: "meta-llama/Llama-3.1-8B-Instruct",
@@ -39,6 +39,7 @@ await inference.chatCompletion({
   ],
   max_tokens: 512,
   temperature: 0.5,
+  provider: "sambanova", // or together, fal-ai, replicate, …
 });
 
 await inference.textToImage({
@@ -146,16 +147,16 @@ for await (const chunk of inference.chatCompletionStream({
 
 /// Using a third-party provider: 
 await inference.chatCompletion({
- model: "meta-llama/Llama-3.1-8B-Instruct",
- messages: [{ role: "user", content: "Hello, nice to meet you!" }],
- max_tokens: 512,
- provider: "sambanova"
+  model: "meta-llama/Llama-3.1-8B-Instruct",
+  messages: [{ role: "user", content: "Hello, nice to meet you!" }],
+  max_tokens: 512,
+  provider: "sambanova", // or together, fal-ai, replicate, …
 })
 
 await inference.textToImage({
- model: "black-forest-labs/FLUX.1-dev",
- inputs: "a picture of a green bird",
- provider: "together"
+  model: "black-forest-labs/FLUX.1-dev",
+  inputs: "a picture of a green bird",
+  provider: "fal-ai",
 })
 
 
@@ -169,14 +170,10 @@ await inference.translation({
   },
 });
 
-await inference.textToImage({
-  model: 'black-forest-labs/FLUX.1-dev',
-  inputs: 'a picture of a green bird',
-})
-
+// pass multimodal files or URLs as inputs
 await inference.imageToText({
+  model: 'nlpconnect/vit-gpt2-image-captioning',
   data: await (await fetch('https://picsum.photos/300/300')).blob(),
-  model: 'nlpconnect/vit-gpt2-image-captioning',  
 })
 
 // Using your own dedicated inference endpoint: https://hf.co/docs/inference-endpoints/
@@ -188,9 +185,9 @@ const llamaEndpoint = inference.endpoint(
  "https://api-inference.huggingface.co/models/meta-llama/Llama-3.1-8B-Instruct"
 );
 const out = await llamaEndpoint.chatCompletion({
- model: "meta-llama/Llama-3.1-8B-Instruct",
- messages: [{ role: "user", content: "Hello, nice to meet you!" }],
- max_tokens: 512,
+  model: "meta-llama/Llama-3.1-8B-Instruct",
+  messages: [{ role: "user", content: "Hello, nice to meet you!" }],
+  max_tokens: 512,
 });
 console.log(out.choices[0].message);
 ```

packages/inference/README.md

@@ -42,15 +42,15 @@ const hf = new HfInference('your access token')
 
 Your access token should be kept private. If you need to protect it in front-end applications, we suggest setting up a proxy server that stores the access token.
 
-### Requesting third-party inference providers
+### Third-party inference providers
 
-You can request inference from third-party providers with the inference client.
+You can send inference requests to third-party providers with the inference client.
 
 Currently, we support the following providers: [Fal.ai](https://fal.ai), [Replicate](https://replicate.com), [Together](https://together.xyz) and [Sambanova](https://sambanova.ai).
 
-To make request 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.
+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
-const accessToken = "hf_..."; // Either a HF access token, or an API key from the 3rd party provider (Replicate in this example)
+const accessToken = "hf_..."; // Either a HF access token, or an API key from the third-party provider (Replicate in this example)
 
 const client = new HfInference(accessToken);
 await client.textToImage({
@@ -63,14 +63,19 @@ await client.textToImage({
 When authenticated with a Hugging Face access token, the request is routed through https://huggingface.co.
 When authenticated with a third-party provider key, the request is made directly against that provider's inference API.
 
-Only a subset of models are supported when requesting 3rd party providers. You can check the list of supported models per pipeline tasks here:
+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)
 - [HF Inference API (serverless)](https://huggingface.co/models?inference=warm&sort=trending)
 
-#### Tree-shaking
+❗**Important note:** To be compatible, the third-party API must adhere to the "standard" shape API we expect on HF model pages for each pipeline task type. 
+This is not an issue for LLMs as everyone converged on the OpenAI API anyways, but can be more tricky for other tasks like "text-to-image" or "automatic-speech-recognition" where there exists no standard API. Let us know if any help is needed or if we can make things easier for you!
+
+👋**Want to add another provider?** Get in touch if you'd like to add support for another Inference provider, and/or request it on https://huggingface.co/spaces/huggingface/HuggingDiscussions/discussions/49
+
+### Tree-shaking
 
 You can import the functions you need directly from the module instead of using the `HfInference` class.