Review and Decks

The review session

stateDiagram-v2
  direction LR
  [*] --> Prompt: next due card
  Prompt --> Reveal: try to recall
  Reveal --> Rated: Forgot · Hard · Got it · Easy
  Rated --> Prompt: more due and under the daily limit
  Rated --> Done: none left
  Done --> [*]: forecast for the coming days

Due cards are computed on the device by replaying logs (see Learning Engine).
Each rating appends a row to reviews and the outbox in one transaction.
The next due date is shown immediately; nothing waits for the network.

Review timing

Setting	Effect	Stored in
Daily limit	Caps reviews per day; overflow rolls forward in due order	`user_settings`
Memory target	Desired retention for FSRS; higher means shorter intervals	`user_settings`
Forecast	Bar chart of cards due each day for the next week	computed
Recall rate	Real first-try recall over recent reviews, shown next to the target	computed

success Showing the honest recall rate next to the target lets learners see whether their target is realistic and adjust it themselves.

Decks

Decks group boxes for focused review. There are three kinds:

Kind	How it's made	Sync
Own	Learner picks boxes from the Shelf	Synced as a record; newest change wins
Topic	Follow a whole topic; new boxes in that topic join automatically	Stores only the topic id
Shared	Opened from a link	A snapshot fetched from `/v1/decks/shared/:code`, which can be saved as an own deck

sequenceDiagram
  participant A as Sharer
  participant API as API
  participant B as Friend
  A->>API: POST /v1/decks/share (deck snapshot)
  API-->>A: short code → link
  A->>B: sends the link
  B->>API: GET /v1/decks/shared/:code
  API-->>B: name + box ids
  B->>B: preview · save as own deck

Deleting a deck writes a tombstone (deleted), so the deletion syncs to every device like any other change.

${sch.nav()}