cjones/WhisperLive
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
591 Commits 2 Branches 20 Tags
ced4bdb7373866dbb8da0e3f59bea35d68532963
Commit Graph

93 Commits

Author SHA1 Message Date
Aaron Boxer ced4bdb737 feat: add Prometheus metrics instrumentation 
- New whisper_live/metrics.py with Counter, Gauge, Histogram metrics
- Track connections (opened/closed/rejected), transcription latency,
  audio processed, segments emitted, REST requests, and errors
- All metric helpers are no-ops when prometheus_client not installed
- --metrics_port CLI flag to expose /metrics endpoint (0 = disabled)
- Metrics integrated into server.py, base.py at key instrumentation points
- 17 new tests in tests/test_metrics.py (178 total passing)
 2026-05-13 10:45:40 -04:00
Aaron Boxer 4210697ca6 Add custom vocabulary / hotwords support 2026-05-13 10:32:28 -04:00
Aaron Boxer 18bce1864a Validate server parameters on startup 
- max_clients must be >= 1
- max_connection_time must be > 0
- batch_max_size must be >= 1 (when batch enabled)
- batch_window_ms must be >= 0 (when batch enabled)
- Added 5 new tests for parameter validation
 2026-04-17 09:30:19 -04:00
Aaron Boxer 9e5e4a9970 Bound transcript memory and translation queue size 
- Add MAX_TRANSCRIPT_LENGTH (500) and MAX_TRANSLATION_QUEUE_SIZE (100)
  class constants to ServeClientBase
- Trim transcript and text lists after each update_segments() call
- Create translation queue with maxsize to prevent unbounded growth
- Added tests for _trim_transcript()
 2026-04-17 09:29:38 -04:00
Aaron Boxer 81cdbbca95 Add thread safety to ClientManager with threading.Lock 
- All ClientManager methods (add_client, get_client, remove_client,
  get_wait_time, is_server_full, is_client_timeout) now protected by
  a threading.Lock
- cleanup() called outside the lock to avoid holding it during I/O
- is_server_full() computes wait time inline under lock instead of
  calling get_wait_time() to avoid nested lock acquisition
- Added concurrent thread safety tests for add/remove and get operations
 2026-04-17 09:27:37 -04:00
Aaron Boxer f5340ddf1e audio: add support for raw pcm input via server flag 
fixes #
 2026-04-17 09:21:06 -04:00
ianwh02 3508b39584 Fix missing batch_config init causing CI test hang 2026-03-09 11:42:15 +00:00
ianwh02 e8bd4fd532 Add cross-client GPU batch inference for faster_whisper backend 2026-02-26 00:00:51 +00:00
Aaron Boxer 29ee640409 api: add support OpenAI REST transcription api 2026-02-05 22:31:59 -05:00
makaveli10 f5256fc62f feat: support HuggingFace model IDs for faster_whisper_custom_model_path 
Previously, the server only accepted local file paths for custom Faster Whisper
models. This change allows passing HuggingFace repo IDs which are automatically
downloaded and converted to CTranslate2 format by the backend if not already in
CTranslate2 format.

Signed-off-by: makaveli10 <vineet.suryan@collabora.com>
 2026-01-13 16:34:04 +05:30
makaveli10 04db67170b ServeClientTranslation import only when enable_tranlsation is True 
Signed-off-by: makaveli10 <vineet.suryan@collabora.com>
 2025-07-22 15:47:06 +00:00
makaveli10 2b8b245fa8 Add translation backend 
Translate from any language to any language with alirezamsh/small100
running in a thread and reading from a queue shared with transcription thread.

Signed-off-by: makaveli10 <vineet.suryan@collabora.com>
 2025-07-22 08:54:10 +00:00
makaveli10 8d6ddd4f7b Change max_clients max_connection_time from server only 
Signed-off-by: makaveli10 <vineet.suryan@collabora.com>
 2025-07-21 22:52:24 +05:30
Kiran Lonikar 40edd25468 remove commented code 2025-07-12 22:29:30 +05:30
Kiran Lonikar 9954548075 issue 371 
model name is of form namespace/repo_name and not os path.
 2025-07-06 16:10:49 +05:30
makaveli10 74abf66d48 Make cache path configurable to save auto converted ct2 models 
Signed-off-by: makaveli10 <vineet.suryan@collabora.com>
 2025-06-01 09:04:22 +00:00
giubots 275ed4e45b Merge branch 'main' into configure-more-params 2025-05-02 12:23:45 +02:00
giubots a2271806c3 feat: client sends new parameters to server 2025-04-28 17:21:33 +02:00
makaveli10 47ee035f65 Upgrade tensorrt_llm to v0.18.2 
Signed-off-by: makaveli10 <vineet.suryan@collabora.com>
 2025-04-22 12:33:03 +00:00
makaveli a62495b090 Integrate OpenVINO backend 
Signed-off-by: makaveli <vineet.suryan@collabora.com>
 2025-03-31 12:57:19 +05:30
makaveli10 c1ac71ada0 Refactor 🔨 
Signed-off-by: makaveli10 <vineet.suryan@collabora.com>
 2025-03-24 16:48:33 +05:30
makaveli10 b6b73730fb Fix: typo 
Signed-off-by: makaveli10 <vineet.suryan@collabora.com>
 2025-01-13 11:35:22 +00:00
makaveli10 182b5cbd6d Fix skipped audio chunk by recording the time of the first repition of a segment 
Signed-off-by: makaveli10 <vineet.suryan@collabora.com>
 2025-01-08 13:58:34 +00:00
makaveli10 c936e5f727 Add lock to thread shared variables updates/reads 
Signed-off-by: makaveli10 <vineet.suryan@collabora.com>
 2025-01-06 06:34:02 +00:00
makaveli10 2eff360b9e Support loading hf models 
Signed-off-by: makaveli10 <vineet.suryan@collabora.com>
 2024-11-26 11:14:18 +05:30
makaveli10 a6523b6b71 Minor fixes for better punctuations 
Signed-off-by: makaveli10 <vineet.suryan@collabora.com>
 2024-11-19 02:13:11 -05:00
makaveli10 778a9c5903 Upgrade faster-whisper 1.1.0rc0 
Signed-off-by: makaveli10 <vineet.suryan@collabora.com>
 2024-11-14 10:07:37 -05:00
makaveli10 81c57ae40c Send completed bool with each segment 
Completed bool represents if the segment is completely processed by the server

Signed-off-by: makaveli10 <suryanvineet47@gmail.com>
 2024-11-05 18:11:32 +05:30
makaveli10 0d74790c67 Expose ClientManager arguments to be passed from client 
Signed-off-by: makaveli10 <vineet.suryan@collabora.com>
 2024-10-10 08:37:45 -04:00
makaveli10 380f07394b Set compute_type based on device capability 
Signed-off-by: makaveli10 <vineet.suryan@collabora.com>
 2024-09-03 00:47:38 -04:00
makaveli10 01c6bc1ecd Initialize last_segment to None 
Signed-off-by: makaveli10 <vineet.suryan@collabora.com>
 2024-08-30 07:47:03 -04:00
makaveli10 fac62aaccc Fix hallucinations with no_speech_thres 
Signed-off-by: makaveli10 <vineet.suryan@collabora.com>
 2024-08-08 06:05:12 -04:00
berkaybilik 2f1c934ea2 always use the BackendType enum to reference the backend inside the TranscriptionServer 2024-06-27 00:20:17 +01:00
berkaybilik b220ccb330 fixed reference before assignment error/warning 2024-06-27 00:11:06 +01:00
berkaybilik 5e3906fc7b use enum to validate backend validity in server.run 2024-06-26 23:59:07 +01:00
Andreas Peldszus 3a96f60006 Raise error for invalid model paths 2024-05-31 15:06:54 +02:00
Andreas Peldszus 3c09289dea Add single model mode for custom models 
- Use a threadlock around the model in single model mode
 2024-05-31 15:06:47 +02:00
makaveli10 18b6168807 fix: cpu usage issue 2024-02-28 13:55:37 +05:30
makaveli10 a41e714801 fix: lanuguage, task prefix in decoder start ids 2024-02-26 23:31:19 -05:00
makaveli10 b42ced9816 fix: tests for end of speech message while mocking pyaudio 2024-02-16 20:31:38 +05:30
makaveli10 01dc69e068 close when end of audio from client 2024-02-15 18:07:19 +05:30
makaveli10 3cd96367fb make vad an option 2024-02-15 14:59:43 +05:30
makaveli10 7bcb92c266 create new method for handling a new connection; expcetion handling 2024-02-15 12:16:18 +05:30
makaveli10 170ba22e5b update method docstrings 2024-02-09 16:08:18 +05:30
makaveli10 eaec0ead08 add: handle_transcription_output method 2024-02-09 14:19:49 +05:30
makaveli10 9fbff47126 🔨 refactor whisper_live according to flake8 2024-02-09 13:45:13 +05:30
makaveli10 9ccf940f51 remove debug import excpetion tensorrt llm 2024-02-08 14:05:46 +05:30
makaveli 6071cc1cc5 Revert "Test pypi upload" 2024-02-08 12:23:15 +05:30
makaveli10 8db063ee33 update log level to info 2024-02-07 23:31:58 +05:30
makaveli10 cc3ed74c0e remove unused imports 2024-02-07 23:26:42 +05:30