FSRS Dynamic Preset Selection

Anki add-on that maps cards to add-on-defined FSRS presets using ordered search rules. The add-on writes Anki's synced fsrsPresetOverlay collection config and lets Anki core resolve the effective FSRS preset during scheduling, search, browser, and stats operations.

This add-on requires an Anki build that exposes:

• Collection.set_fsrs_preset_overlay()
• anki.collection.FsrsPresetOverlay

Configuration

The add-on config contains add-on FSRS presets and ordered rules:

{
  "presets": [
    {
      "id": "addon:example:medical",
      "name": "Medical",
      "fsrs_version": "seven",
      "params": [],
      "desired_retention": 0.9,
      "historical_retention": 0.9,
      "deck": "Medical",
      "search": "firstgrade:1"
    }
  ],
  "rules": [
    {
      "search": "tag:medical",
      "preset_id": "addon:example:medical"
    }
  ]
}

The add-on config screen edits the preset list directly. If a preset has a deck selected and/or a search filter, the add-on writes one combined rule like deck:"Medical" firstgrade:1 before the advanced rules entries. Preset ids are generated from the preset name. Use Move Up and Move Down in the dialog to change preset precedence.

The config screen can refresh card counts for the current draft rules. Each preset row shows how many cards its generated query selects. The deck summary below the preset table shows, for every normal deck, how many cards are not selected by any generated or advanced rule. It also shows the same unselected count with is:new cards removed.

Use the Same-day Reviews checkbox on an FSRS-7 preset row to control whether same-day reviews are included when optimizing that preset. Use Optimize on an individual preset row to recompute only that preset's parameters. Use Optimize All to recompute every visible preset row in current order. While optimization is running, each row's optimize button is replaced by a progress bar: pending rows show Pending, the active row shows Optimizing, and completed rows show Done. The active row is filled from Anki's backend optimizer progress.

For the common first-answer split workflow, check Grade Split on a preset row. The row is replaced with four grade rows for Again, Hard, Good, and Easy. They inherit the row's name, deck, and search filter, but keep separate desired retention, same-day optimize, and parameter values. Their generated queries are ordered as deck:"..." firstgrade:N <search>.

The interval split checkbox creates Young and Mature rows using stored card intervals. In Days mode, Young uses prop:ivl<N and Mature uses prop:ivl>=N. In Percent mode, the add-on reads the current matching cards' stored intervals, computes a day cutoff for the bottom percentage, then writes the same prop:ivl rules. If both grade and interval split are checked, the tree shows eight leaf rows with both Grade and Interval labels. Interval split rows also emit simulator rules that keep the stable deck, grade, and search selector separate from the simulated interval bound.

Rules are evaluated by Anki core in order. The first matching rule assigns the card's FSRS preset. Cards that do not match any rule use their deck-config FSRS preset.

Development

Anki loads the installed add-on from ~/Library/Application Support/Anki2/addons21/1798968356, not directly from this source repo. Copy changed runtime files into that folder and reload or restart Anki before testing in the app.

When reading deck options through collection.decks.config_dict_for_deck_id(), remember that Anki returns a legacy Python dict. Extra deck-option fields stored in Rust as DeckConfig.inner.other are flattened into the top level, so fsrs7IncludeSameDayOptimize is read as config["fsrs7IncludeSameDayOptimize"]. This add-on stores the FSRS-7 same-day optimize choice on each add-on preset; the deck option is only used to initialize older presets that do not yet have the field.

Run tests with:

python3 -m pytest