Trust
Residency on PeachDesk is not a region checkbox. It is a sourcing decision you make per deployment and per pipeline stage, from fully hosted to fully self-hosted voice AI, with per-stage mix and match in between.
We run the platform for you on managed AWS infrastructure behind a web application firewall. Your data lives in the hosted environment. We do not publish a region list, so if residency specifics matter to you, talk to us before you commit.
The full platform runs from Docker Compose on your infrastructure: database, object storage, voice engine, all of it. Recordings, transcripts, and inference stay inside your environment, in the region you choose. Billing and telemetry are off by default in the self-hosted stack.
Mix and match per stage, per agent. Stages routed to Local mode run on your GPUs inside your environment, so that stage’s audio and text stay with you. Precisely: stages on managed providers still send that stage’s data to the provider. Residency on PeachDesk is exact, stage by stage, not a vague platform-wide promise.
Data residency claims fail in the details. Here is exactly where each part of a call is processed, per sourcing choice.
| Pipeline piece | Where its data is processed |
|---|---|
| Stage on a managed provider (Frontier mode) | Sent to the managed provider you selected for that stage. |
| Stage on your own provider keys (BYOK) | Sent to your provider, under your own agreement with them. |
| Stage on self-hosted inference (Local mode) | Processed on your GPUs and stays inside your environment. |
| Recordings and transcripts | Object storage the deployment controls, behind signed, time-limited URLs. |
| The phone call itself | Your telephony carrier carries it, on every deployment. |
| Product telemetry | Off by default; traces route to your own Langfuse project. |
If EU residency is a hard requirement, self-hosting gives it to you today: deploy on infrastructure in your chosen EU region and recordings, transcripts, and Local mode inference stay inside your environment, with product telemetry off by default. Configurable retention windows with automated purge let you enforce storage limits per agent.
Your counsel decides what the regulation requires of you. Our job is to give you the deployment control to meet it, and to say exactly which stages touch which provider. For hosted-cloud residency specifics, talk to our team. For the controls underneath, see security and privacy controls.
That depends on the layer you choose. PeachDesk-hosted runs the platform on our managed AWS cloud. Self-hosted puts the full platform, including object storage for recordings and transcripts, inside your cloud and region. Per-stage sourcing splits it: stages routed to Local mode keep that stage’s data in your environment, while stages on managed providers send that stage’s data to the provider you selected.
Yes, through self-hosting. Deploy the full platform on infrastructure in your chosen EU region and recordings, transcripts, and Local mode inference all stay inside your environment, with product telemetry off by default. For the PeachDesk-hosted cloud, residency specifics are a conversation: talk to our team about your requirements before you commit. We do not publish a hosted region list, so we will not quote one here.
Local stages do not. When speech-to-text, the language model, and text-to-speech run in Local mode, that stage’s audio and text are processed on your GPUs inside your environment on leading open models. The precision matters: any stage you route to a managed provider still sends that stage’s data to that provider, and your telephony carrier always carries the call itself.
None, unless you turn it on. Product telemetry is off by default in every deployment, and the self-hosted stack ships with telemetry inert. If you want observability, you point it at your own tools: call and LLM traces can route to your own Langfuse project per organisation. Your operations data stays yours by default, not ours.
Tell us which data may leave your environment and which may not. We will map it to a deployment and sourcing plan, stage by stage.