<!-- Generated by scripts/generate-aeo-artifacts. Do not edit. -->

# Subprocessors

Version 1.3 · Last updated 9 Nov 2026 · Source: https://eturns.app/policy/subprocessors/

Eturns uses the following subprocessors to provide the Services. We disclose them publicly so merchants and customers can review where data is processed before they install Eturns. Everything Eturns persists is stored by Supabase in Singapore (AWS ap-southeast-1), and the application and its queue also run on Render in Singapore. See section 5 of the Privacy Policy and the Data Processing Agreement at /policy/dpa for what that means for international transfers.

## Current subprocessors

| Processor                 | Region                        | What we send                                                                      | DPA                       |
| ------------------------- | ----------------------------- | --------------------------------------------------------------------------------- | ------------------------- |
| **Vercel AI Gateway**     | Global                        | Decision context (sentinel IDs, structured fields, and PII-redacted message text) | Vercel DPA                |
| **xAI / OpenAI / Google** | Global (via Gateway)          | Same as above, routed by Vercel AI Gateway                                        | Routed by Gateway         |
| **Supabase**              | Singapore, AWS ap-southeast-1 | Full app data under RLS; photo evidence in private buckets                        | Supabase DPA              |
| **Render**                | Singapore                     | App hosting, background worker, and Redis counters + session IDs (no PII)         | Render DPA                |
| **Resend**                | Global                        | Recipient email + body for the single send call                                   | Resend DPA                |
| **Sentry**                | United States                 | Stack traces + scrubbed extras                                                    | Sentry DPA                |
| **PostHog**               | United States                 | Event properties (PII-scrubbed before send)                                       | PostHog DPA               |
| **Shopify**               | Global                        | Shopify-internal: Admin API, Billing, webhooks                                    | Shopify Partner Agreement |

## AI model vendors (routed via Vercel AI Gateway)

Eturns never calls model vendors directly. All model traffic goes through the Vercel AI Gateway. The current registry routes:

- Medium/high decisions and vision primary to xAI Grok 4.6
- Low-complexity decisions and triage to OpenAI GPT-5.4 Mini (with GPT-5.4 Nano as first failover)
- Vision failover to Google Gemini 3.5 Flash
- Case similarity embeddings to OpenAI text-embedding-3-small

A later stack is in the application and stays off until AI_MODEL_STACK=next or the PostHog flag ai-model-stack is on. This page lists the vendors that receive traffic by default. Update it before that cutover.

The authoritative model list lives in app/lib/ai/models.ts in the Eturns application and is mirrored in docs/07-AI-ARCHITECTURE.md. If we add or change a model vendor, we update this page.

## Changes to this list

We will post a new or changed subprocessor on this page at least 30 days before it starts processing customer data, and we will email the address on file with Shopify. Watch this page or email contact@eturns.app if you want another path.