EU Data Residency

Gradium runs inference in both the EU and the US. By default you call a single endpoint, api.gradium.ai, and requests are routed to the nearest cluster. For European customers this already means that, in practice, sessions are processed in the EU. EU data residency is for when “in practice” is not enough. Once enabled on your plan, EU processing becomes a guarantee that is enforced on every request: your sessions are never re-routed outside the EU, and every response tells you so.

Endpoints

Endpoint	Behaviour
`api.gradium.ai`	Default. Routed to the nearest cluster; no residency commitment.
`eu.api.gradium.ai`	EU-pinned. Same API, same keys, same paths. For organizations enrolled in EU data residency, sessions are guaranteed to be processed in the EU.

The base URLs for the EU endpoint are:

https://eu.api.gradium.ai/api
wss://eu.api.gradium.ai/api

Everything else is unchanged: the same API key, the same routes, the same request and response formats.

What is guaranteed

For an organization enrolled in EU data residency:

Inference stays in the EU. Text-to-Speech, Speech-to-Text and Speech-to-Speech sessions, over WebSocket or the one-shot REST endpoints, run on our EU clusters, located in Germany and France.
No cross-region fallback. Under load a session may be served by another EU cluster, but never by a cluster outside the EU.
Fail closed. If a pinned request ever reaches a non-EU cluster, it is refused rather than processed out of region (see Rejections).
Session data is written in region. Whatever session data we retain (see Zero Data Retention to retain none) is written by the cluster that served the session.

Account data, custom voices and pronunciation dictionaries are managed by our control plane, which is hosted in the EU for all customers.

How to opt in

EU data residency is available on our paid plans. It is enabled on your plan by our team; there is no self-serve toggle yet.

1. Ask us to enable it

Contact support@gradium.ai with your organization name and pick one of two modes:

Explicit (default): sessions are pinned to the EU when you call eu.api.gradium.ai. Calls to api.gradium.ai keep today’s behaviour. Choose this if only part of your traffic needs the guarantee.
Implicit: your whole organization is pinned to the EU, whichever endpoint you call. Choose this if all of your usage is in scope.

2. Point your integration at the EU endpoint

import gradium

client = gradium.client.GradiumClient(
    api_key="your-api-key",
    base_url="https://eu.api.gradium.ai/api",
)

# -D - prints the response headers, including x-gradium-residency
curl -L -D - -o bonjour.wav -X POST https://eu.api.gradium.ai/api/post/speech/tts \
  -H "x-api-key: your-api-key" \
  -H "Content-Type: application/json" \
  -d '{"text": "Bonjour !", "voice_id": "YTpq7expH9539ERJ", "output_format": "wav", "only_audio": true}'

wss://eu.api.gradium.ai/api/speech/tts
wss://eu.api.gradium.ai/api/speech/asr

With implicit mode this step is optional, since every endpoint pins your sessions to the EU, but we still recommend it.

3. Verify

Every response reports the residency the request actually got, so you can check it instead of trusting a hostname:

REST: the x-gradium-residency response header.
WebSocket: the residency field of the ready message.

Value	Meaning
`pinned; zone=eu`	The guarantee is in force: this session is processed in the EU and will not leave it.
`best-effort; zone=eu`	The request landed in the EU, but your organization is not enrolled: no commitment is made.
`none`	No region is involved: default endpoint, organization not enrolled.

TTS ready

{
  "type": "ready",
  "request_id": "req_...",
  "model_name": "default",
  "residency": "pinned; zone=eu",
  "sample_rate": 48000,
  "frame_size": 3840
}

If you rely on the guarantee, assert pinned in your client and treat anything else as a configuration error. best-effort on eu.api.gradium.ai means residency has not been enabled on your plan yet.

Rejections

A pinned request that reaches a cluster outside the EU is refused instead of being processed out of region:

REST: 403 Forbidden
WebSocket: close code 1008 (policy violation)

Both carry the reason This request must be sent to the eu region endpoint. In practice this happens when an organization in implicit mode calls an endpoint that is not EU-pinned; the fix is to use eu.api.gradium.ai.

Residency and Zero Data Retention

The two are independent and combine:

Data residency governs where your requests are processed.
Zero Data Retention governs whether request and response payloads are retained at all. See the FAQ to enable it.

For a Data Processing Agreement or details on our privacy and security posture, contact support@gradium.ai.

Getting Started

Text-to-Speech

Speech-to-Text

Speech-to-Speech

Shared

Real-time Recipes

Migrations

Voices

Self-Hosted

Resources

Endpoints

What is guaranteed

How to opt in

1. Ask us to enable it

2. Point your integration at the EU endpoint

3. Verify

Rejections

Residency and Zero Data Retention

​Endpoints

​What is guaranteed

​How to opt in

​1. Ask us to enable it

​2. Point your integration at the EU endpoint

​3. Verify

​Rejections

​Residency and Zero Data Retention

Endpoints

What is guaranteed

How to opt in

1. Ask us to enable it

2. Point your integration at the EU endpoint

3. Verify

Rejections

Residency and Zero Data Retention