Who contributed to Hugging Face on August 11, 2026?

8 developers shipped this update, including andimarafioti, imstevenpmwork, pkooij, Nicholas Broad, HuggingFaceInfra, Bharadwaj Pendyala, ovuruska, and huggingface-hub-bot.

What were the notable Hugging Face updates?

Preserve ordered assistant text and tool output, Fix demo status after empty STT transcription, and [Download] Fix tqdm_class ignored by the Xet transfer bar (#4647).

@huggingface

Transformers, Datasets, and the open AI-model layer

github ↗

Pick a date

Topics: Python AI / ML Full archive →

The Wire · Showcase

SPEECH-TO-SPEECH SHIPS ORDERED ASSISTANT FIX, HUB SDK PATCHES PROGRESS BAR BUG

By RepoJournal · Filed 06:59 UTC on August 11, 2026 · About Hugging Face

8 people shipped this

andimarafioti @andimarafioti 2 cited

pkooij @pkooij 1 cited

Bharadwaj Pendyala @bharadwaj-pendyala 1 cited

ovuruska @ovuruska 1 cited

imstevenpmwork @imstevenpmwork

Nicholas Broad @nbroad1881

HuggingFaceInfra @HuggingFaceInfra

huggingface-hub-bot @huggingface-hub-bot

Hugging Face speech-to-speech preserves assistant text and tool output in a single ordered sequence, eliminating split-queue complexity across parsing, Realtime events, and conversation history.

The fix [1] routes ordered assistant events and their matching audio through one FIFO path, removing response ordinals and per-output audio bookkeeping that created ordering races. A companion fix [2] resolves a demo client state bug where empty STT transcriptions left WebSocket and WebRTC clients stuck in processing mode instead of returning to listening. On the SDK front, huggingface_hub v1.27.0 landed yesterday, and the team is already shipping critical fixes: the Xet transfer bar now respects a caller's custom tqdm_class instead of hardcoding the built-in tqdm [3], ensuring progress bars honor documented customization. DeepInfra inference provider gains text-to-speech support [4], targeting their OpenAI-compatible audio endpoint. LeRobot fixes a dataset loading crash where recipes failed on frames carrying multi-camera VQA events that the default binding could not disambiguate, even when the recipe never referenced that binding [5]. Across 4 repos, 20 commits and 10 PRs shipped in the last 24 hours.

Action items

→ Pull speech-to-speech fixes to any instance using Realtime assistant responses huggingface/speech-to-speech [immediate]
→ Upgrade huggingface_hub to 1.27.0+ if you use Xet-backed repos with custom progress bars huggingface/huggingface_hub [plan]
→ Update LeRobot if using multi-camera datasets with VQA events huggingface/lerobot [plan]

References

[1] Preserve ordered assistant text and tool output ↗ huggingface/speech-to-speech
[2] Fix demo status after empty STT transcription ↗ huggingface/speech-to-speech
[3] [Download] Fix tqdm_class ignored by the Xet transfer bar (#4647) huggingface/huggingface_hub
[4] [Inference Providers] deepinfra: add text-to-speech support ↗ huggingface/huggingface_hub
[5] fix(datasets): resolve only recipe-referenced bindings ↗ huggingface/lerobot

Quick answers

What shipped in Hugging Face on August 11, 2026?: Hugging Face speech-to-speech preserves assistant text and tool output in a single ordered sequence, eliminating split-queue complexity across parsing, Realtime events, and conversation history. In total, 20 commits and 10 pull requests landed.
Who contributed to Hugging Face on August 11, 2026?: 8 developers shipped this update, including andimarafioti, imstevenpmwork, pkooij, Nicholas Broad, HuggingFaceInfra, Bharadwaj Pendyala, ovuruska, and huggingface-hub-bot.
What were the notable Hugging Face updates?: Preserve ordered assistant text and tool output, Fix demo status after empty STT transcription, and [Download] Fix tqdm_class ignored by the Xet transfer bar (#4647).

OPTIMUM-INTEL ADDS META MUSE GLIMMER SUPPORT VIA OPENVINO

OpenVINO now exports Meta's Muse Glimmer multimodal model with int4 quantization, bringing image-text-to-text inference to the optimization stack.

python 20 shipped

@huggingface Aug 10

PYTORCH-IMAGE-MODELS FIXES POSITIONAL EMBEDDING RESAMPLING BUG

A shortcut in positional embedding resampling was silently skipping necessary transformations when token counts matched but grid shapes differed.

python 2 shipped

@huggingface Aug 9

MOON-IDE SHIPS RELIABILITY OVERHAUL ACROSS MESSAGING AND SESSIONS

Hugging Face's IDE tooling just got significantly more resilient, with five commits landing stale response handling, markdown rendering fixes, and a pending message queue for network hiccups.

python 11 shipped

@huggingface Aug 8

SPEECH-TO-SPEECH PROTOCOL OVERHAUL SHIPS, DIFFUSERS FIXES LORA STATE CORRUPTION

The speech-to-speech realtime pipeline just rebuilt its entire response lifecycle to keep audio, transcripts, and tool calls synchronized across WebSocket and WebRTC, while diffusers patched a critical bug where partially unfused LoRA adapters were marked as clean but remained fused in silent components.

python 47 shipped

Elsewhere on the wire

AI Agents Aug 12

LANGCHAIN SHIPS THREE CRITICAL RELEASES IN 12 HOURS, CLOSES SETUPTOOLS VULNERABILITY

LangChain pushed langchain-core 1.5.4, langchain 1.3.15, and langchain-anthropic 1.5.5 overnight while patching a setuptools supply-chain risk that gated the Hugging Face integration.

ai-agents 59 shipped

Local LLMs Aug 12

OLLAMA CUTS REPEAT PENALTY BLOAT, LLAMA.CPP SHIPS POCKET-TTS SUPPORT

Ollama kills a decade of default penalty cruft that was drowning every model, while llama.cpp gains multimodal audio chops and vLLM fixes ROCm speculative decoding corruption.