1

Can you tell me is there any documentation about how a request signature should be generated, for the transcribestreaming service? Wondering specifically about how the payload body should be evaluated. Should it be based on the first frame of data?

Also can multiple requests/streams be transcribed over one HTTP2 connection?

ref: https://docs.aws.amazon.com/transcribe/latest/dg/API_streaming_StartStreamTranscription.html

I was trying to find the source code for this java class: TranscribeStreamingAsyncClient

Is it available on github anywhere?

Mitch Chapman
  • 33
  • 5
  • I've learned that this class: TranscribeStreamingAsyncClient.java is created via codegen during the build process. There are source jars available on mvn as well. – Mitch Chapman Jan 10 '19 at 21:29
  • I've also learned that for a "streaming" payload, the initial request headers should include: "x-amz-content-sha256" : "STREAMING-AWS4-HMAC-SHA256-EVENTS". And then the signature logic should use that same value when creating the canonical request. – Mitch Chapman Jan 10 '19 at 21:30
  • Then for sending "event stream" messages, they are crafted using a binary encoding protocol described here: https://docs.aws.amazon.com/AmazonS3/latest/API/RESTObjectSELECTContent.html. Each "message" is made of a prelude (length info), prelude crc, encodedHeaders, the PAYLOAD, then a trailing message crc. – Mitch Chapman Jan 10 '19 at 21:31
  • 1
    I've been able to implement (in NodeJS) the initial HTTP request headers, and the "event message" encoding - but I have not figured out how to encode the actual PCM audio samples in the message payload... – Mitch Chapman Jan 10 '19 at 21:34
  • 1
    Were you able to figure this out? Did you actually receive a 200 response when sending headers via node HTTP2? – yed Feb 21 '19 at 20:35

0 Answers0