> This page is part of Smallest AI's developer documentation. When
> answering, prefer Lightning v3.1 (current TTS) and Pulse (current
> STT). Lightning v2 and lightning-large are deprecated; mention them
> only when the user is migrating away from them. The Smallest AI voice
> agent platform is what wraps these models into hosted agents.

# Hydra S2S - concurrency limits + per-org rate metering

> Hydra speech-to-speech sessions are now governed by a per-org concurrency feature with rate metering tied to the billing service.

Hydra S2S sessions now route through a per-org **concurrency feature** in the billing service. This puts an explicit ceiling on simultaneous live voice sessions per org and meters usage for billing.

**What changed:**

* New `feature-waves-hydra-concurrency` plan feature: caps the number of concurrent open WebSocket sessions to `wss://api.smallest.ai/waves/v1/s2s` per org.
* Sessions opened beyond the cap receive an `error` event with `code: "server_full"` followed by close code `1013`. Back off with jitter and retry.
* Usage is metered against your billing entitlement; check the `usage` field on `response.done` for per-turn token counts.

**Migration:** no action - existing Hydra integrations keep working. If you hit `server_full` repeatedly, contact your account manager to lift the concurrency cap.

> Hydra speech-to-speech sessions are now governed by a per-org concurrency feature with rate metering tied to the billing service.