Priority Reorder Addon

Reorder your new Anki cards so the ones you care about come first.

Changelog

See CHANGELOG.md.

Overview

Instead of learning new cards in plain frequency order, you build a priority queue from searches like these:

• common words first, using your frequency field
• words from the VN, book, game or show you're reading now or plan to, using occurrence dictionaries
• cards you added recently
• words whose kanji, or kanji readings, you don't know yet
• specific decks, tags or note types

View Example

View Example Config

```
{
  "normal_search": "deck:日本語::Mining",
  "priority_search": [
    "deck:日本語::Mining occurrences:[9-nine-ここのつここのかここのいろ,9-nine-そらいろそらうたそらのおと,9-nine-はるいろはるこいはるのかぜ,9-nine-ゆきいろゆきはなゆきのあと]>=10",
    "deck:日本語::Mining occurrences:[9-nine-ここのつここのかここのいろ,9-nine-そらいろそらうたそらのおと,9-nine-はるいろはるこいはるのかぜ,9-nine-ゆきいろゆきはなゆきのあと]>=3",
    "deck:日本語::Mining occurrences:穢翼のユースティア>=3 added:14",
    "deck:日本語::Mining occurrences:穢翼のユースティア>=10 added:14",
    "deck:日本語::Mining occurrences:魔法少女ノ魔女裁判>=10 added:14",
    "deck:日本語::Mining occurrences:[うたわれるもの,うたわれるもの2,うたわれるもの3]>=20 added:14",
    "deck:日本語::Mining occurrences:穢翼のユースティア>=7",
    "deck:日本語::Mining occurrences:穢翼のユースティア>=5"
  ],
  "priority_search_mode": "sequential",
  "sort_field": "FreqSort",
  "sort_reverse": false,
  "tuning": {
    "priority_cutoff": null,
    "normal_prioritization": null,
    "priority_limit": null,
    "shift_existing": true
  },
  "sync_behavior": {
    "reorder_on_sync": true,
    "auto_update_dicts": false
  },
  "word_fields": {
    "expression_field": "Expression",
    "expression_reading_field": "ExpressionReading"
  }
}
```

My first searches pick the most frequent words in the VN I'm reading now. Later searches pick frequent words from VNs I plan to read next.

Installation

1. Install from AnkiWeb.
2. Restart Anki.

You need a note type with frequency data. I recommend Lapis. To add frequency data to existing cards, see backfill-anki-yomitan.

Quick Start

The default config prioritizes cards added in the last 3 days. To change it:

1. Go to Tools → Add-ons, select Priority Reorder and press Config.
2. Edit the settings you want. For cards added in the last 5 days:

   {
        "priority_search": [
            "deck:日本語::Mining added:5"
        ],
        "normal_search": "deck:日本語::Mining",
        "sort_field": "FreqSort"
   }
   

   A deck name with spaces needs escaped quotes: "\"deck:日本語::Mining Deck\" added:5". Occurrence dictionary folder names can't contain spaces at all, so rename the folder.

3. Set sort_field to the frequency field of your note type (e.g. "FreqSort" or "Frequency").
4. Press OK.

The addon reorders your new cards after every sync, then syncs once more so your other devices get the new order. Press Ctrl+Alt+` to reorder by hand.

The searches and sorting sit at the top level of the config. Everything else is grouped into the matching, tuning, sync_behavior, today_new_limit and word_fields sections.

How it Works

The addon splits your new cards into two queues:

1. Priority queue: cards matching priority_search, shown first.
2. Normal queue: cards matching normal_search, shown after.

Each queue is sorted by sort_field. A card matching both goes in the priority queue, so in the example above, recent cards are placed first even though they also match the normal search.

Everything lives under Tools → Priority Reorder: Reorder Cards (Ctrl+Alt+` ), Show Summary, and Update Jiten Occurrence Dictionaries.

Summary Window

Tools → Priority Reorder → Show Summary shows what each priority search did in the latest reorder this session. If nothing has run yet, press Run reorder or sync.

The top shows how many cards were prioritized, when, and a bar splitting the priority queue by search. Below, each search gets a row with its kept and matched counts, its place in the queue, and a thin bar showing where its matches went: kept, taken by an earlier search, over its limit=, or cut by priority_cutoff. Click a row for the numbers and for buttons that open those notes in the Browser. Edit config and Run reorder are at the top, so you can change a setting and check the result straight away.

In "mix" mode the searches are pooled before sorting, so each row shows only its matched count.

In "cycle" mode, a search that takes several turns shows two ranges, such as 1–5, 2981–3400 ↻: its first turn, then the stretch where its later turns alternate with the other searches. Hover the range for the number of turns. The bar shows each first turn in place, then a striped stretch for the later turns; hover it to see which searches placed how many cards there.

Search Terms

Mix these into any Anki search:

All of them except limit= also work in the Browse search bar and through the collection API (and so AnkiConnect), which makes the Browser a good place to try a search before you put it in your config. They use the same matching settings, word_fields and sort_field as the reorder. Comparisons can be =, !=, <, <=, > or >=, and a leading - negates a term (-seen:30).

The short forms are the same terms, and you can mix the two. limit= has no short form, because l belongs to length. If one of your note types has a field named o, k, s or l, the short form hides searches on that field, so use the long form there. Short forms also work only while this addon is enabled, so write the long form in a search you plan to share.

Frequency (f)

f<10000 or f>=30000 compares the number in your sort field. It's most useful combined with other terms, e.g. to keep only the common words from an occurrence search.

Length (length)

length>=3 matches expressions of 3 or more characters, length=1 single-character words. The raw field value is counted in Unicode characters, with no HTML stripping, so markup counts toward the length. An empty field has length 0.

Occurrences (occurrences:)

Prioritize words from specific media, using Yomitan occurrence dictionaries.

• occurrences:銀色、遥か>=5 matches words that appear 5 or more times in 銀色、遥か.
• occurrences:[銀色、遥か,穢翼のユースティア]>=10 adds up the counts from both.
• occurrences:all>=10 adds up every dictionary in user_files, for words common across all your media.

Setting up occurrence dictionaries

Download them from Jiten: each media page has them under Download deck -> Yomitan (occurrences).

1. Go to Tools → Add-ons, select Priority Reorder and press View Files.
2. Open the user_files folder.
3. Create a folder named after the dictionary, e.g. 銀色、遥か. The folder name is what you type after occurrences:, so it can't contain spaces.
4. Unzip the Jiten download into it. The addon reads every term_meta_bank_*.json, and the updater needs index.json:

   user_files/
   ├── 銀色、遥か/
   │   ├── index.json
   │   └── term_meta_bank_1.json
   └── 穢翼のユースティア/
       ├── index.json
       └── term_meta_bank_1.json
   

5. Set word_fields to match your note type. For Lapis:

   "word_fields": {
       "expression_field": "Expression",
       "expression_reading_field": "ExpressionReading"
   }
   

Updating occurrence dictionaries

The addon can update dictionaries that came from Jiten.

• Tools → Priority Reorder → Update Jiten Occurrence Dictionaries checks all of them now.
• "auto_update_dicts": true in the sync_behavior section checks once a day, after a sync.

Jiten's API allows roughly 10 requests per minute, so with more than 10 dictionaries an update pauses to wait out the limit. If that makes syncing feel slow, leave auto_update_dicts off and update from the menu.

Matching options

By default a card counts only the entries that exactly match its expression and reading. The options in the matching section also credit it with related entries:

"matching": { "prefix_matching": true, "variant_matching": true }

They apply wherever occurrences: and seen: run, in the reorder and in the Browse bar. All are off by default, and they combine without counting an entry twice.

The full rules follow. Every option only adds counts, so the worst a loose match can do is push a card over a threshold early.

Kanji (kanji:)

Prioritize words by the kanji you already know. A kanji is known once it appears in the expression of a card you've learned (a review or relearning card that isn't suspended). Each example also works with k:.

• kanji:new=0: every kanji is known.
• kanji:new=1: exactly 1 unknown kanji.
• kanji:new>=2: 2 or more unknown kanji.
• kanji:num=1: exactly 1 kanji.
• kanji:num>=3: 3 or more kanji.

A kanji stops being new as soon as one learned word contains it. To raise the bar, add a target: with kanji:new[3], a kanji stays new until 3 learned words contain it (kanji:new is kanji:new[1]). This helps keep practising kanji you've met in only one or two words. The target works on new and new_reading, not on num.

New readings (kanji:new_reading)

kanji:new asks whether you've met the kanji. kanji:new_reading asks whether you've met the reading it has in this word, which is what the card tests.

Once you've learned 食事 (しょくじ), kanji:new counts 食べる (たべる) as fully known, though 食=た is new to you. That's the card you're about to fail.

• kanji:new_reading>=1: at least 1 kanji in a reading you haven't learned.
• kanji:new_reading>=1 kanji:new=0: words made entirely of kanji you know, in a reading you don't.
• kanji:new_reading[3]>=1: a reading stays new until 3 learned words use it.

Readings are tracked per kanji, so 生活 (せいかつ) doesn't teach 生きる (いきる). Inflections share a reading: 上がる and 上げる are both 上=あ, while 上る (のぼる) is different. Rendaku isn't a new reading either: 血 (ち) covers 鼻血 (はなぢ).

Words whose reading doesn't split across their kanji, like 火傷 (やけど), 今日 (きょう) and gikun readings, count as new for every kanji they can't explain. Those are the least predictable readings, so that's usually what you want. When only part of a word is irregular, only that part counts: 眼鏡/めがね credits 眼=め and flags only 鏡.

Set word_fields.expression_reading_field to the field with the kana reading. Cards without a reading never match. Readings are checked against a bundled table built from KANJIDIC2. If this term matches almost every card, the reading field is usually the cause, and the console prints a warning.

Example: occurrences:銀色、遥か>=5 kanji:new=0 kanji:new_reading>=1 finds words common in the VN you're reading that look fully known but will trip you up.

Recently seen words (seen:)

Experimental: seen: may change or be removed in a future version.

Prioritize words from your recent immersion, using daily occurrence dictionaries. The matching options apply to seen: too.

• seen:N (short s:N) matches words in the daily dictionary of any of the last N days, however many times they appear. -seen:30 negates it.
• seen:1 is today. "Today" follows Anki's "Next day starts at" setting.
• Keep the window small. Cost grows with the number of days: seen:1 to seen:3 are cheap, while seen:30 and up get noticeably slower, more so with prefix_matching or variant_matching on. Use the smallest window that still means "recent". Reusing the same window in several searches costs nothing extra.

Setting up seen dictionaries

My Daily Occurrences addon writes one for each day, from text that arrives over a websocket (the usual VN setup).

To build them yourself, put them in a _seen folder under user_files, one subfolder per day named YYYY-MM-DD:

user_files/
└── _seen/
    ├── 2026-06-11/
    │   └── term_meta_bank_1.json
    └── 2026-06-12/
        └── term_meta_bank_1.json

Each term_meta_bank_*.json is an ordinary Yomitan occurrence dictionary and uses the same word_fields. The _seen folder is reserved: it isn't an occurrence dictionary, so occurrences:all skips it and occurrences:_seen can't reach it. Only seen:N reads it.

Multiple searches

Make priority_search a list, and priority_search_mode decides how the searches combine:

• "sequential": all of the first search, then the second, and so on. ["added:3", "tag:ノベルゲーム::銀色、遥か"] puts recent cards first, then the 銀色、遥か cards.
• "mix": every match in one pool, sorted together.
• "cycle": the searches take turns, each placing its next limit= cards, until all run out. ["deck:A limit=10", "deck:B", "deck:C limit=5"] places A's top 10, all of B, C's top 5, then A's next 10, C's next 5, and so on. Without any limit= this is the same as "sequential".

Limits and cutoffs

• limit=X in a search keeps only its top X cards (added:3 limit=20: the 20 most frequent recent cards). In "cycle" mode it's the number of cards per turn, and the rest wait for the next turn.
• tuning.priority_limit caps the whole priority queue.
• tuning.priority_cutoff sends priority cards whose sort value is past this number (the rarer words, with a frequency-rank field) to the normal queue.
• tuning.normal_prioritization does the opposite: normal cards with a sort value under this number join the priority queue, after all your searches.

Fitting the day's new cards to the queue

With today_new_limit on, a deck's new card limit rises for the day to fit the priority queue: 26 priority cards on a 10-a-day preset gives you 26 that day. It uses Anki's Today only limit, so the preset is never changed.

"today_new_limit": { "enabled": true, "decks": ["日本語"], "max": 60 }

Name the deck you click to study. The limit is recalculated on each reorder until you study your first new card of the day, then holds. config.md has the details.

Credits

Kanji reading and variant data is derived from KANJIDIC2, Copyright © the Electronic Dictionary Research and Development Group, used under the CC BY-SA 4.0 licence. The bundled kanji_readings.txt and kanji_variants.txt are modified extracts (readings and variant groups only, re-encoded) and are likewise CC BY-SA 4.0.