chore: merge master → dev/video-fer (live camera FER mode)

fix: route video files through async polling to restore FER display

fix: route /api/transcribe-stream to Next.js for SSE

chore: merge master → dev/video-fer (SSE transcribe-stream)

fix: detect Git LFS pointer in FER model, auto-download real binary

fix: pin onnxruntime==1.24.2 + cache-bust Dockerfile for FER

fix: async job polling for timeout + FER ONNX ort>=1.19.0

docs: update README with Studio features and API format

chore: merge master + streaming emotions, inline badges, hidden scrollbar

feat: streaming valence/arousal + inline bracket badges + hidden scrollbar

feat: char-level text highlight + click-to-seek + inline bracket badges

feat: ElevenLabs-style transcript highlight + timeline playhead

feat: bidirectional timeline ↔ transcript sync with scroll highlight

fix: remove play/pause overlay from RightPanel video preview

feat: per-second FER timeline for live emotion panel

fix: restore output:standalone in next.config.ts for Docker build

feat: add delete button to recent sessions in sidebar

fix: prevent double API call on upload; replace frontend with redesigned UI

feat: replace frontend with redesigned Ethos Studio UI + update READMEs

fix: extend proxy timeouts for CPU inference (30/60 min)

fix: correct FER model path and STT error propagation

fix: add required language= arg to apply_transcription_request()

debug: add /debug-inference endpoint to smoke-test local model

fix: use bfloat16 on CPU to halve memory usage (~6 GB vs ~12 GB)

chore: update web/package-lock.json for Node 22

refactor: restructure repo into api/ proxy/ web/ training/ docs/

feat: local Voxtral inference + FER ImageNet norm fix

fix: ImageNet normalization in FER preprocessing

fix: RefObject<HTMLVideoElement|null> for React 19 TS build

fix: RefObject type for React 19 (TS build fix)

feat: add video FER pipeline with full-stack support

chore: sync master to HF Spaces (skip binary history)

docs: update READMEs for evoxtral external API proxy

fix: always convert audio to WAV before forwarding to evoxtral API

feat: switch to external evoxtral Modal API (no local model)

fix: bypass apply_chat_template, use processor(audio) directly

fix: remove add_generation_prompt (not supported by MistralCommonTokenizer)

fix: pin transformers==4.54.0 to avoid mm_load_kwargs bug

debug: expose /api/debug-inference via Node proxy

debug: add /debug-inference endpoint

debug: add traceback logging + fix dtype cast + add_generation_prompt

debug: add detailed timing logs to _transcribe

feat: switch to Voxtral-Mini-3B-2507 + evoxtral-lora adapter

fix: use bfloat16 on CPU to halve memory usage (16GB→8GB)

fix: remove resemblyzer dep (unused, webrtcvad needs gcc)

feat: add HuggingFace Spaces Docker deployment

Merge pull request #3 from aytoast/feat/model

docs: update READMEs to reflect single-speaker VAD segmentation

fix: distribute text by sentence boundaries instead of character proportion

refactor: replace speaker clustering with VAD sentence segmentation