Skip to main content
VARTA

Run VARTA inside your own environment

VARTA ships as four Docker services -- the workflow engine (FastAPI), the LiveKit-based voice worker that answers and places SIP calls, the designer/operations web console, and PostgreSQL -- deployable with Docker Compose for a single-server pilot or Kubernetes for horizontal scale. Because the stack runs on infrastructure you choose, call audio, transcripts and workflow data stay inside the environment you deploy it in. Data residency here is a deployment decision you make, not a certification we hold.

Security posture

  • Access is scoped per tenant with an ownership check on every request, behind an audit-before-enforce authentication mode -- you can watch what would be blocked before you switch enforcement on.
  • Enterprise SSO via SAML (through WorkOS), with just-in-time user provisioning configured per tenant -- implemented and unit-tested today; end-to-end sign-in against a live identity provider is still being verified, so treat it as available for evaluation rather than proven in production yet.
  • The session store supports envelope encryption at rest, turned on per deployment by configuring an encryption key; retention windows are configurable, application logs are PII-redacted, and a dedicated endpoint erases a caller's data on request -- controls built directly against India's DPDP Act obligations (storage limitation, the right to erasure, security safeguards), not a claim of certification.
  • All external traffic runs over HTTPS/WSS, the API is CORS-restricted with a per-deployment origin allowlist, and the database and internal services sit off the public network -- only the ports a call actually needs (HTTPS, SIP/RTP, WebRTC media) are exposed.

Deployment & scale

A pilot runs on a single 16 vCPU / 32 GB server. Our sizing guidance for a 1,000 concurrent-call deployment across 100 numbers -- voice workers, media/SIP gateway, API tier and a Postgres primary with a replica -- comes to roughly 20 nodes. These are engineering sizing estimates from our own deployment documentation, not a measured benchmark or an SLA.

Talk to us

Tell us what you're looking to build and we'll get back to you.