Privacy & security

How Tosin handles your data

Tosin is designed to keep your designs in your hands. This page summarizes the data path, retention policy, and provider boundaries.

Short version

Your screenshots are sent over HTTPS to the model provider you configure, processed for the analysis, and dropped from worker memory once the batch result is returned. Tosin does not run analytics on the contents of your designs.

Core privacy guarantees

No persistent design storage

Screenshots you submit are processed in-memory, forwarded to your chosen model provider, and discarded after the analysis batch completes. Tosin does not retain image bytes after results are delivered.

TLS in transit

All requests between the browser, the worker, and the model provider travel over HTTPS/TLS 1.2+. Self-hosted deployments inherit the TLS configuration of your reverse proxy or platform.

Bring-your-own keys

Tosin uses the OpenAI key you configure on the worker. We do not multiplex traffic across tenants or proxy through Anthropic-owned infrastructure. Your provider relationship is direct.

Minimal telemetry

The worker logs request IDs, timings, and provider error codes for observability. No image bytes, prompt content, or user PII is shipped to any third-party analytics destination.

Data flow

  1. STEP 01

    You submit a batch

    Your browser base64-encodes the images and sends them with your goal, user, and problem context to the worker over HTTPS.

  2. STEP 02

    Worker forwards to the model provider

    The FastAPI worker calls the OpenAI Chat Completions API (or your configured provider) with the encoded image and the analysis prompts.

  3. STEP 03

    Results stream back to you

    Per-stage outputs (T1–T4) return to the worker, are merged into the batch response, and streamed to your browser as they complete.

  4. STEP 04

    In-flight state is discarded

    Once the batch is delivered, the worker drops the in-memory batch record. Redis (if configured) is used only for short-lived job coordination, never as long-term storage.

Model provider boundaries

OpenAI

By default, the OpenAI API does not use API-submitted data to train models. Refer to https://openai.com/policies/api-data-usage-policies for the current terms.

Self-hosted models

If you point the worker at a self-hosted or compatible API (Azure OpenAI, vLLM, Together, etc.), data handling is governed by that provider’s contract, not Tosin.

What you should not send

Treat screenshots like model input

Do not include personally identifiable information, secrets, customer data, or unreleased product secrets you would not paste into any external API. Mask sensitive regions before uploading; the analysis quality does not depend on real data.

Your data subject rights

Deletion: there is nothing persistent to delete on the Tosin side once a batch result is delivered. Provider-side retention is governed by your model provider account.

Access / export: the batch result you receive in the browser is the complete record. Save the JSON output if you need an archive; Tosin will not keep one for you.

Self-hosting: Tosin is fully runnable on your own infrastructure. Set OPENAI_API_KEY and WORKER_URL to keep the entire pipeline inside your perimeter.

This page is a plain-language summary. It is not a contract and does not replace your organization’s data processing agreement.