[Agustin Torres (torrresagus)]

What changed in the API

Notice huggingface-serverless-inference-api-retired about huggingface.co's API describes this change, published 2025-07-01. It is not signed by huggingface.co: it was read from a file the person running Tiltmeter trusts.

The notice's summary:

The serverless Inference API at api-inference.huggingface.co is gone: its domain no longer resolves
(checked October 27, 2026), so requests fail before they reach Hugging Face. Its successor, HF
Inference, answers at router.huggingface.co/hf-inference with the same requests and responses per
task, and the OpenAI-compatible chat API moved to router.huggingface.co/v1. Sources:
huggingface.co/docs/inference-providers and the huggingface_hub client.

POST https://api-inference.huggingface.co/models/{model}

Severity	Change	Path	Before	After
breaking	endpoint_removed	$	—	(gone), moved to POST https://router.huggingface.co/hf-inference/models/{model}

POST https://api-inference.huggingface.co/pipeline/feature-extraction/{model}

Severity	Change	Path	Before	After
breaking	endpoint_removed	$	—	(gone), moved to POST https://router.huggingface.co/hf-inference/models/{model}/pipeline/feature-extraction

POST https://api-inference.huggingface.co/models/{model}/v1/chat/completions

Severity	Change	Path	Before	After
breaking	endpoint_removed	$	—	(gone), moved to POST https://router.huggingface.co/v1/chat/completions

POST https://api-inference.huggingface.co/v1/chat/completions

Severity	Change	Path	Before	After
breaking	endpoint_removed	$	—	(gone), moved to POST https://router.huggingface.co/v1/chat/completions

The notice's notes:

POST /models/{model}: Same request body and response for the model's task. HF Inference needs a
token (Bearer hf_...) with the Inference Providers permission, and since July 2025 it serves mostly
CPU models such as embeddings and classifiers; large chat models are served through
router.huggingface.co/v1/chat/completions.
POST /pipeline/feature-extraction/{model}: The task now comes after the model. Same body (inputs)
and the same array of embeddings back. HF Inference needs a token (Bearer hf_...) with the Inference
Providers permission, and since July 2025 it serves mostly CPU models such as embeddings and
classifiers; large chat models are served through router.huggingface.co/v1/chat/completions.
POST /models/{model}/v1/chat/completions: Send the model id in the body's model field instead of the
path; the router picks a provider that serves it. An OpenAI client's base URL changes from
https://api-inference.huggingface.co/v1 to https://router.huggingface.co/v1.
POST /v1/chat/completions: Same OpenAI-compatible body and response. An OpenAI client's base URL
becomes https://router.huggingface.co/v1.

What this pull request changes

I moved the .NET HuggingFace connector off the retired api-inference.huggingface.co host. Text generation, image-to-text and embeddings now default to router.huggingface.co/hf-inference, and embeddings use the new path order: models/{model}/pipeline/feature-extraction. Chat completion now defaults to router.huggingface.co/v1/chat/completions, with the model id still sent in the request body. I updated the unit tests, the TelemetryWithAppInsights and F# samples, and the integration testsettings.json to match. Nothing was built or run.

• dotnet/samples/Demos/FSharpScripts/huggingFaceChatCompletion.fsx
• dotnet/samples/Demos/TelemetryWithAppInsights/Program.cs
• dotnet/src/Connectors/Connectors.HuggingFace.UnitTests/Services/HuggingFaceChatCompletionTests.cs
• dotnet/src/Connectors/Connectors.HuggingFace.UnitTests/Services/HuggingFaceEmbeddingGenerationTests.cs
• dotnet/src/Connectors/Connectors.HuggingFace.UnitTests/Services/HuggingFaceEmbeddingGeneratorTests.cs
• dotnet/src/Connectors/Connectors.HuggingFace.UnitTests/Services/HuggingFaceImageToTextTests.cs
• dotnet/src/Connectors/Connectors.HuggingFace.UnitTests/Services/HuggingFaceTextGenerationTests.cs
• dotnet/src/Connectors/Connectors.HuggingFace/Core/HuggingFaceClient.cs
• dotnet/src/Connectors/Connectors.HuggingFace/Core/HuggingFaceMessageApiClient.cs
• dotnet/src/IntegrationTests/testsettings.json

Left for a human

• The embeddings URL now always puts the task after the model (models/{model}/pipeline/feature-extraction), including when the caller passes their own endpoint. Callers who point the embedding service at a server that still uses the old path order would need to adjust.
• HF Inference now needs an hf_ token with the Inference Providers permission, and since July 2025 it serves mostly CPU models. Text-generation models such as the integration test's microsoft/phi-2 may no longer be served there. That can't be fixed with code edits.
• The F# script still pins the Microsoft.SemanticKernel.Connectors.HuggingFace 1.12.0-preview NuGet package. Its chat endpoint now points at router.huggingface.co, but the package version was not raised because no release with this fix exists yet.

---

This pull request was generated automatically by Tiltmeter from the unsigned notice huggingface-serverless-inference-api-retired about huggingface.co's API. An AI agent (Claude Code with claude-opus-5-5) wrote these edits; review them like any other change.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has insufficient remaining quota.

[Agustin Torres (torrresagus)]

@microsoft-github-policy-service agree