Deepgram released a new version of its on-premises solution.

On-Premises Release 221130: Docker Hub Images

  • deepgram/onprem-api:1.75.1

  • deepgram/onprem-engine:3.39.2

    • Minimum required NVIDIA driver version: >=450.80.02

  • deepgram/onprem-license-proxy:1.2.2

  • deepgram/onprem-billing:1.5.0

  • deepgram/onprem-metrics-server:2.0.1

Changes

  • For streaming requests, interim_results now defaults to false. This parameter does not impact batch. For normal streaming requests, Deepgram recommends leaving this set to the default value. To change the default to true, modify your api.toml with:
    [features]
      interim_results_on_by_default = true  # default is false

  • For streaming requests, endpointing now defaults to true. This parameter does not impact batch. For normal streaming requests, Deepgram recommends leaving this set to the default value. To change the default to false, modify your api.toml with:
    [features]
      endpointing_on_by_default = false  # default is true

  • Improved handling of punctuation when redaction is also enabled.

  • Added support for audio data that uses the the A-law algorithm.

  • Error message formatting is now consistent with Deepgram Cloud:

    • The error field has been renamed err_code.

    • The reason field has been renamed  err_msg.

    • The transaction_key field is now deprecated.

  • Added a new DG-Error HTTP response header to provide information on failing HTTP 101 WebSocket upgrades.

  • Resolves the following CVEs in supporting libraries for deepgram/onprem-billing:

  • Resolves the following CVEs in supporting libraries for deepgram/onprem-metrics-server:

  • Other minor stability improvements.

We welcome your feedback. Please share it with us at Product Feedback.

Stop building work-arounds for STT systems that don't work.

Start FreeTalk to an expert