ENTERPRISE DEPLOYMENT
Sageio inside your own infrastructure.
For organizations that can't put meeting data in a multi-tenant cloud. Sageio runs in your VPC, your data center, or your private cloud — under your security boundary, on your audit trail.
WHY SELF-HOST
When on-premise is the only answer.
Data residency
Regulators, customer contracts, or internal policy require meeting data to stay inside a specific country or network. Self-hosting puts the data exactly where it has to be — and never anywhere it shouldn't.
Compliance obligations
Industries under FINRA, HIPAA, ITAR, or equivalent local regimes face audit requirements that SaaS rarely satisfies. Running Sageio in your environment keeps you in control of evidence — and the auditor's questions.
Existing security perimeter
Some organizations have invested years in network segmentation, internal CA, and zero-trust tooling. Self-host lets Sageio inherit that perimeter rather than punch holes through it.
ARCHITECTURE
What runs where.
Self-hosted Sageio is a containerized application that runs end-to-end inside your environment. AI inference, transcription, storage — all within your network boundary.
The core deployment includes the Sageio application server, a PostgreSQL database, object storage for audio and transcripts, and an inference layer for transcription and translation. Customers may bring their own STT and LLM endpoints (Deepgram, Azure OpenAI, Google Vertex, or comparable) or run Sageio's reference inference stack on local GPUs. Outbound network access is required only for license validation and optional telemetry, both of which can be tunneled through a customer-controlled proxy.
Architecture diagram
Reference architecture for on-premise deployment. Adapted per customer environment during scoping.
REQUIREMENTS
What your environment needs to have.
- Orchestration
- Kubernetes 1.27 or later. Helm 3.12+. Docker Compose supported for proof-of-concept and single-host deployments.
- Compute (application tier)
- 8 vCPU, 16 GB RAM minimum per node. Two nodes recommended for high availability.
- Compute (inference tier, optional)
- NVIDIA GPU with 24 GB VRAM (A10G, L4, or equivalent) for self-hosted STT. Not required if using managed STT / LLM endpoints.
- Database
- PostgreSQL 14 or later. Customer-managed or Sageio-deployed. Minimum 100 GB storage, scaled to retention policy.
- Object storage
- S3-compatible (AWS S3, MinIO, Ceph, Azure Blob, GCS). Used for transcript exports and audio retention windows.
- Network egress
- Outbound HTTPS to a Sageio license endpoint required. All other traffic configurable, including through customer proxy.
- Identity
- SAML 2.0 or OIDC identity provider. Tested integrations: Okta, Microsoft Entra ID, Google Workspace, Auth0, Keycloak.
- TLS
- Customer-provided certificate, or integration with internal CA. Let's Encrypt supported for internet-facing deployments.
INSTALLATION
From kickoff to first meeting.
A typical self-host deployment runs four to six weeks from contract signature to the first translated meeting in production. Sageio assigns a dedicated solutions engineer for the duration.
Scoping & environment review
Joint working session covering your infrastructure, identity provider, network architecture, compliance requirements, and rollout plan. Output: a deployment design document, signed off by both teams before any installation work begins.
Staging install
Sageio installed in your staging or pre-production environment. Helm values configured for your storage, identity provider, and inference endpoints. End-to-end test meeting validated.
Security & compliance review
Your security team runs penetration testing, SBOM review, and any internal change-management process. Sageio supplies documentation, attestations, and engineering responses to findings.
Production deployment & cutover
Production install, identity federation enabled, monitoring connected to your observability stack. First production meetings run alongside your existing tooling until your team is confident.
Handover & enablement
Admin training, runbook walkthrough, and documentation handover. Ongoing support transitions to your assigned customer success engineer.
SUPPORT
We're with you through production.
Self-hosted deployments include solutions engineering before, during, and after rollout. Not a Slack channel and a wiki — a person who knows your environment.
Every self-host customer is assigned a named solutions engineer for deployment, and a customer success engineer for the duration of the contract. Standard support covers business-hours response on weekdays, with a four-hour acknowledgment SLA on production-down issues. Extended-hours and 24×7 coverage are available as add-ons. Quarterly review calls cover platform updates, security advisories, and the upgrade roadmap. Major version upgrades are scheduled jointly and run against your staging environment first.
Need this on your terms?
Tell us about your environment and we'll send a scoping questionnaire. Most evaluations complete in two weeks, including a working proof-of-concept in your staging environment.