What we store

• Your account. You sign in with Google (through Firebase Authentication). That gives us your name, email address and profile picture.
• Your library. The feeds you subscribe to, which articles you've read or saved, your notes, highlights, tags and thumbs up/down.
• Your settings. Mutes, smart filters, feed priorities, topics you marked "not interested", what you told Discover you want to read about, and the headlines you picked in Discover's taste test.
• Reading activity. Which feeds and topics you open, how long you spend and how far you scroll, daily reading totals for your stats and streaks, and which screens you use (with your device type). Plume uses this to rank articles for you and to improve the app.
• How accurate Jev's scores are for you. For articles Jev scored, whether you finished, abandoned or cleared them, kept as daily counts per score range, not as a list of articles. We use it to check the match % is honest.
• Discover searches that come up short. When "what do you want to read about?" finds few matching feeds, we keep the words you typed, without your account, so we know which feeds to add.
• Feedback you send. Your message, plus your email so we can reply.

This is kept in Google Cloud Firestore. Shared data, such as articles fetched from public feeds and thumbs-up counts, isn't tied to you. Scores and filter results are also cached in your own browser's local storage, along with a copy of your reading profile and your recent Top Stories scores so Discover can compare feeds against them. If you take the taste test signed out, your picks and the sample reader you chose stay in your browser until you sign in.

Discover's catalog forecast (how timeless each catalog feed is, which stories it shares with other feeds, and estimates for four sample readers) is worked out once a week from public feeds only. It doesn't use anything about you.

What we send to Jev

Top Stories, Catch me up, smart filters, same-story grouping and Discover's "what do you want to read about?" are powered by Jev, a model from Typesafe AI. Jev answers yes/no questions with probabilities. It doesn't write text, and Plume never shows you anything it wrote. When you use these features, Plume sends Jev:

• the articles being ranked or checked: title, source, age, and a short excerpt (on Discover, recent posts from catalog feeds you're looking at);
• a reading profile:
  • titles of articles you liked, disliked or saved, and headlines you picked in the taste test;
  • the names of the feeds you open most, favourite or deprioritise;
  • your most-read topics and minutes spent on each;
  • your muted keywords, "not interested" topics and hot keywords;
  • and what you said you want to read about;
• the text of your smart filters, when checking articles against them.

Plume does not send Jev your name, email, account ID, notes or highlights. Typesafe processes these requests under its own terms. Each account has a daily limit on these requests.

Analytics

We use Vercel Analytics and Speed Insights for page views and performance, and Google Analytics for Firebase for in-app usage events. Neither is used for advertising.

What we never do

• Sell or rent your data.
• Show ads, or share your data with advertisers.
• Rewrite or summarise the articles you read. You always read the original.

Your choices

• Leaving? Export your subscriptions any time as OPML from Import / Export.
• Turn Jev off. Settings → Ranking & filtering → "Use Jev". With it off, nothing is sent to Jev and your feed is in plain time order. Mutes and feed priorities still work.
• Deleting your account. Send us a request through the Feedback button in the app and we'll delete your account and the data stored for it.