-
0.5.1
- httpServer.js is now renamed voskjshttp.js. It accepts GE and POST endpoints.
- voskjshttp now reply to an HTTP POST request receiving the speech WAV
file as binary data attached in the HTTP request:
curl -X POST 'http:localhost:3000/transcript' \ --header "Content-Type: audio/wav" \ --data-binary "@speech.wav"
-
0.4.0
- httpServer interface changed. Now accepting HTTP GET requests and query string arguments.
-
0.3.14
- added latency tests (with/without grammars)
- added function
transcriptFromBuffer
that transcripts audio as buffer instead of as a WAV file - added
lib/toPCM.js
module function to manage ffmpeg input/output Buffers, through stdin/stdout, avoiding disk I/O
-
0.3.11
- added some tests to verify ffmpeg transcoding times (using stdout pipe / without producing wav file)
-
0.3.8
- Integrate CLI command
voskjs
with grammar argument - global installation of the package allow to run
voskjs
cli program - socketio server program example
- Integrate CLI command
-
0.3.6
- delivered @solyarisoftware/voskjs (added index.js)
- httpServer now accepts grammar attribute in client API calls
- added test program tests/sequentialRequests.js
- function
transcript
can be called with aliasrecognize
- function
transcript
now accept options argument, allowing to manage Vosk grammar format. - documenation improved (README, examples/README, tests/README)
-
0.2.11
- added com/abtest.sh, apache bench httpServer stress test script
- httpServer improved:
- better validation of body request
- accept optionally client request id as attribute in the request body
- takes model, port and debug attributes as command line args.
- documenation improved (README, examples/README, tests/README)
-
0.2.1
- added function logLevel
- httpServer error responses improvements
-
0.2.0
- function initModel is renamed loadModel
- function loadModel and Transcript return an object containing also the processing latency
- httpServer reviewed. The transcript endpoint return a different JSON data structure, containing latency time.
-
0.1.0 Added a simple HTTP sever
-
0.0.15 Added tests directory, containing some stress tests results
-
0.0.14 Transcript function updated to integrate Vosk version 0.3.25 (
npm install vosk@latest
), where the functionrec.acceptWaveformAsync
now run on a separated external thread!