Quentin Fuxa 3b7a2fcc87 Add Qwen3-ASR MLX SimulStreaming backend ...

New backend 'qwen3-mlx-simul' for Apple Silicon: AlignAtt border
detection via monkey-patched cross-attention on MLX Qwen3-ASR.
Supports 0.6B (RTF 0.236 on M5) and 1.7B models.

- qwen3_mlx_simul.py: full streaming implementation with KV cache,
  alignment head attention extraction, border-distance policy
- core.py: register new backend in TranscriptionEngine + online_factory
- parse_args.py: add qwen3-mlx-simul to CLI choices

2026-03-15 11:00:00 +01:00

..

benchmark

update benchmark results and procedure

2026-03-15 11:16:15 +01:00

diarization

Improve diarization backends

2026-02-15 14:55:00 +01:00

local_agreement

Improve online ASR processor

2026-01-17 09:35:00 +01:00

silero_vad_models

fixes silence detected but never reported by silero

2025-11-23 11:20:00 +01:00

simul_whisper

Fix ctranslate2 encoder conversion (#345) and memory leak in TokensAlignment (#344)

2026-03-10 22:37:00 +01:00

voxtral_mlx

voxtral mlx : improved chunking

2026-03-14 00:13:29 +01:00

web

Fix frontend buffer rendering for slow backends

2026-02-14 09:28:00 +01:00

whisper

fix(whisper/tokenizer): prevent IndexError from crashing multilingual streams

2026-03-02 15:31:43 +08:00

__init__.py

Update package setup and exports

2026-02-21 11:33:00 +01:00

audio_processor.py

Refactor audio processor async pipeline

2026-01-25 13:48:00 +01:00

backend_support.py

mixstral hf v0

2026-02-20 20:49:57 +01:00

basic_server.py

voxtral mlx : improved chunking

2026-03-14 00:13:29 +01:00

cascade_bridge.py

qwen3 simul+kv: optimized streaming with kv cache reuse

2026-03-15 18:30:00 +01:00

cli.py

update benchmark results and procedure

2026-03-15 11:16:15 +01:00

config.py

voxtral mlx : improved chunking

2026-03-14 00:13:29 +01:00

core.py

Add Qwen3-ASR MLX SimulStreaming backend

2026-03-15 11:00:00 +01:00

deepgram_compat.py

Add OpenAI REST API and Deepgram WebSocket

2026-02-08 15:42:00 +01:00

diff_protocol.py

Add snapshot-then-diff WebSocket protocol

2026-02-07 10:17:00 +01:00

ffmpeg_manager.py

isort

2025-11-23 11:20:00 +01:00

metrics_collector.py

Refactor timed objects and data structures

2026-01-11 16:08:00 +01:00

metrics.py

Refactor timed objects and data structures

2026-01-11 16:08:00 +01:00

model_mapping.py

simulstreaming mlx & torch dedup of common base

2025-02-15 23:52:00 +01:00

model_paths.py

Clean up config and model paths

2026-01-10 11:42:00 +01:00

parse_args.py

Add Qwen3-ASR MLX SimulStreaming backend

2026-03-15 11:00:00 +01:00

qwen3_asr.py

voxtral mlx : improved chunking

2026-03-14 00:13:29 +01:00

qwen3_mlx_asr.py

qwen

2026-01-02 23:52:00 +01:00

qwen3_mlx_simul.py

Add Qwen3-ASR MLX SimulStreaming backend

2026-03-15 11:00:00 +01:00

qwen3_simul_kv.py

tune simul-kv: 2s inference interval, configurable min_new_seconds

2026-03-15 21:30:00 +01:00

qwen3_simul.py

qwen3: reuse encoder kv cache

2026-03-15 22:31:39 +01:00

session_asr_proxy.py

Add per-session language proxy

2026-02-01 17:03:00 +01:00

silero_vad_iterator.py

Ruff lint cleanup

2026-01-03 10:23:00 +01:00

test_client.py

Add test harness and test client

2026-02-22 16:19:00 +01:00

test_data.py

Add test harness and test client

2026-02-22 16:19:00 +01:00

test_harness.py

Add test harness and test client

2026-02-22 16:19:00 +01:00

thread_safety.py

Refactor timed objects and data structures

2026-01-11 16:08:00 +01:00

timed_objects.py

Refactor timed objects and data structures

2026-01-11 16:08:00 +01:00

tokens_alignment.py

Improve tokens alignment and silence handling

2026-01-24 10:55:00 +01:00

vllm_realtime.py

voxtral mlx : improved chunking

2026-03-14 00:13:29 +01:00

voxtral_hf_streaming.py

voxtral mlx : improved chunking

2026-03-14 00:13:29 +01:00

voxtral_mlx_asr.py

update benchmark with qwen3 which reuses kv cache

2026-03-15 22:32:01 +01:00

warmup.py

Update package setup and exports

2026-02-21 11:33:00 +01:00