Self-Hosting
Reson8 is available as a self-hosted deployment for teams that need audio to stay within their own network. The same speech-to-text engines that power our API run on your own Kubernetes cluster and GPUs, managed by you.
What You Get
Section titled “What You Get”- The same API - the self-hosted deployment exposes the same Prerecorded, Realtime and Turns endpoints as our API, with these exceptions: requests need no authentication (access control is handled by your own network), and phrases and custom models are not available.
- Your data stays with you - audio and transcripts are processed entirely inside your cluster and never leave it. The deployment only contacts Reson8 to report aggregated usage counters (seconds of audio processed) and to renew its license; no audio, transcripts or request content is sent.
- Pick what you need - realtime, turn detection and diarization can each be enabled separately, so you only run the components your use case needs.
Requirements
Section titled “Requirements”- A Kubernetes cluster with Helm
- NVIDIA GPU nodes
- Outbound HTTPS access to Reson8 for pulling images, reporting usage and periodic license checks
Versioning
Section titled “Versioning”Self-hosted releases follow our API: new features and model improvements ship as new versions of the Helm chart. You decide when to upgrade.
Chart versions follow semantic versioning. Breaking changes only ship in a major version, so minor and patch upgrades can be applied with your existing configuration. The API itself carries the same compatibility guarantees described in Versioning.
Getting the Helm Chart
Section titled “Getting the Helm Chart”Self-hosting is available on request. Contact us to discuss your use case and set up an agreement. Once onboarded, you’ll receive credentials for the Reson8 registry, which hosts the Helm chart and container images.
The chart ships with its own installation guide covering configuration, sizing, upgrades and monitoring.