Residency & regions

You choose the region your companion thinks in, you can see where every single provider processes your data, and one switch — EU only — holds everything weside chooses inside the EU.

What it is

There are two settings here, and they do different amounts of work.

Where your companion thinks decides the inference region: which language model answers you, and where it runs. You pick a scope. Today that is EU, US, Global and Your key; the list is drawn from what the provider registry actually holds, so it follows what is available rather than a fixed menu. Each scope carries its own legal line — the EU scope says GDPR · DPA in place · processed in the EU, the US scope names the US CLOUD Act, Global promises no region at all.

This choice governs the model, and only the model. Web search, the platform tools your companion calls, picture generation, speech and the channels you connected run where they run, no matter which region you pick.

EU only is the second setting, and it is the one that governs the rest. While it is on, nothing weside chooses leaves the EU: a tool whose processor sits outside the EU is not offered and is not silently substituted, pictures are only generated on an EU endpoint, and speech is off. A source whose location cannot be established counts as outside — the switch promises what it can prove.

Data locations is the third thing on this screen, and it is where the question “where is my data processed” is actually answered. It lists, row by row with a country: the model, the database, embeddings, speech, the Claude Code sandbox, every tool and web-search provider, picture generation, and each channel you connected.

Why it matters to you

Your companion’s memory is the most personal data you hold here. A region choice is a routing rule rather than a sentence in a policy — but it is a rule about one thing, and a claim that it covers everything would be false. That is why EU only exists as its own switch, why it costs you capabilities, and why the screen shows you what it costs before you flip it.

Data locations is the part you can check. Every provider that touches anything of yours is named there with the country it processes in, including the ones weside would rather not have to list.

How you use it

Open DATA RESIDENCY from the pill on the Home surface.
Pick a scope under Where your companion thinks. The legal line for that scope appears under the row.
Pick a quality tier: balanced or deep. (A lighter, system-internal swift tier exists but isn’t one you pick — it is chosen automatically.)
Use Change to see every model inside the scope you chose, family by family, and pick one.
Scroll to Data locations to read where each part of the system processes your data.

Turning EU only on

EU only sits under the scope row and can only be switched on while your companion thinks in the EU. Before it flips, a sheet shows you exactly what stops working — each blocked source with the country it processes in — and, separately, your connected channels with their locations, under the heading that says they stay connected. The switch governs what weside picks for you; a channel you connected yourself is your choice and keeps running. Turn on stays disabled until that channel list has loaded, because a preview that can’t show what it promises to show isn’t a preview.

From the command line the same switch is weside provider eu-only on and weside provider eu-only off.

When something is blocked

You don’t discover this by a failure. Ask your companion for something that would need a blocked tool and it tells you, once per conversation, that EU only is the reason and names both ways out: switch it off, or carry on without that tool.

Limits & edge cases

EU only needs the EU region. Switch your companion to another scope, or activate your own key, and EU only is switched off in the same step. The app says so where you make that change, and the API refuses the combination outright.
Speech is off under EU only, not degraded. Speech synthesis runs inside the voice pipeline where no tool filter can see it, so the server refuses to open a voice session at all rather than route it outside the EU. See voice.
Place names go, coordinates stay. Turning coordinates into a place name uses a processor outside the EU, so under EU only your companion works with the coordinates and skips the lookup.
What keeps working under EU only: notes, memory, files, the sandbox, and anything weside runs itself. So do the channels and integrations you connected — the switch never disconnects something you chose.
Changing region doesn’t move what is already stored. It changes where new processing happens.
Your own key is your own responsibility. With bring your own key, your companion’s model calls go wherever your provider runs them, and residency is yours to reason about rather than weside’s.
Which models exist in which region changes. Model families, presets and their data locations are operational data and are updated without a release, which is why this page names none of them. The in-app picker and Data locations are always the current answer.