zhifu gao 55daa5a193 fix(realtime): batch concurrent websocket decoding (#3529) ...

Fix the realtime WebSocket concurrency regression by removing the process-wide session lock and batching compatible vLLM decode requests. Isolate per-session VAD runtime state while sharing encoder weights, keep batch failures contained, and document reproducible concurrency benchmarks.

Fixes #3528

Signed-off-by: LauraGPT <[email protected]>

2026-08-26 23:30:20 +08:00

..

__init__.py

create

2022-11-26 21:56:51 +08:00

_server_app.py

fix(server): return speaker labels for spk requests (#3433)

2026-07-27 20:01:21 +08:00

compute_audio_cmvn.py

docs: add docstrings to 1950 functions across 276 files for API documentation

2026-05-21 07:52:52 +00:00

export.py

docs: add docstrings to 1950 functions across 276 files for API documentation

2026-05-21 07:52:52 +00:00

inference.py

docs: add docstrings to 1950 functions across 276 files for API documentation

2026-05-21 07:52:52 +00:00

realtime_ws.py

fix(realtime): batch concurrent websocket decoding (#3529)

2026-08-26 23:30:20 +08:00

server.py

fix(server): return speaker labels for spk requests (#3433)

2026-07-27 20:01:21 +08:00

tokenize_text.py

fix: add try-import guards for optional deps (fastapi, humanfriendly, transformers)

2026-05-26 23:11:51 +08:00

train_ds.py

fix: use torch.amp via compatibility shim (#3518)

2026-08-26 20:10:53 +08:00

train.py

fix: use torch.amp via compatibility shim (#3518)

2026-08-26 20:10:53 +08:00