[Inference Providers] Rework the Get started page #1806
Conversation
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
SBrandeis
requested review from
Copilot,
julien-c,
Pierrci,
Wauplin,
Vaibhavs10,
burtenshaw and
hanouticelina
There was a problem hiding this comment.
Pull Request Overview
This PR overhauls the Inference Providers docs to highlight the new OpenAI-compatible chat completion endpoint and streamline the Getting Started experience for developers.
- Refines the introduction and “Why Choose” section with clearer benefits
- Reworks Getting Started to showcase chat completion examples via Python, JavaScript, and HTTP
- Adds text-to-image quick start and detailed provider selection policy
Comments suppressed due to low confidence (2)
docs/inference-providers/index.md:16
- The table header 'Text to video' is inconsistent with 'Text to Image' casing; consider capitalizing 'Video' for consistency.
| Provider | Chat completion (LLM) | Chat completion (VLM) | Feature Extraction | Text to Image | Text to video |
docs/inference-providers/index.md:284
- The example here is inconsistent with earlier examples that include quote marks around 'G' and 'huggingface'. Update to
"How many 'G's in 'huggingface'?"for clarity and consistency.
"content": "How many G in huggingface?"
There was a problem hiding this comment.
Some really nice improvements, especially the extra snippets for all languages.
- I followed @Vaibhavs10 's comment and tried it make the style a bit less synthetic.
- I found some of the content to detailed for the quickstart, and might be better in guides. But that's a matter of opinion.
🚀
docs/inference-providers/index.md
Outdated
| When using Inference Providers, your requests go through Hugging Face's proxy infrastructure, which provides several key benefits: | ||
|
|
||
| - **Unified Authentication & Billing**: Use a single Hugging Face token for all providers | ||
| - **Automatic Failover**: If one provider is unavailable, requests can be routed to alternatives |
There was a problem hiding this comment.
well it almost is true
When the preferred provider mapping is in error state and the client or server in auto mode, the request is routed to the next best one
There was a problem hiding this comment.
ah yes sorry, I understood it as "if request fails, it's rerouted to the next one"
|
@Wauplin I pushed some changes addrssing your concerns - let me know what you think |
|
LGTM! |
|
|
||
| <a href="https://huggingface.co/playground" target="blank"><img src="https://cdn-uploads.huggingface.co/production/uploads/5f17f0a0925b9863e28ad517/9_Tgf0Tv65srhBirZQMTp.png" style="max-width: 550px; width: 100%;"/></a> | ||
| Inference Providers works with your existing development workflow. Whether you prefer Python, JavaScript, or direct HTTP calls, we provide native SDKs and OpenAI-compatible APIs to get you up and running quickly. |
There was a problem hiding this comment.
this particular sentence feels a bit AI-slop
TL;DR
Update the documentation to mention the new chat completion endpoint
Hopefully makes this more useable and clearer for developers as well