PERSONAL READING, ON YOUR TERMSONLINE WORKSPACE
A PERSONAL TOOL

Settings & privacy

Your library is stored in your hosted database. You choose how recommendations are made.

Scoring provider

Use the fake provider to explore, or OpenAI to assess articles against your rule.

Use a model that supports Responses structured outputs, for example gpt-4.1-mini. Pricing is user-configured; verify current rates for your chosen model. A provider or model change makes existing results stale.

Where your data goes

Fetching connects to the public publications you follow. Real scoring sends article text, article metadata, and your recommendation rule to OpenAI. The app requests store=False; your provider’s data policies still apply.

The fake provider and the bundled demo need no network. API keys stay in the server environment and are never written to the database or included in backups.

OpenAI key: configured on the server. API credentials are managed in your Vercel project settings.

Keep a copy of your library

Your hosted Postgres database contains your library, preferences, reading history, and scoring results. Use its backup tools or pg_dump to keep a copy.

Source removal preserves reading history. Backups contain personal reading data, but never API secrets.