GitHub repository

A free, open-source local vocabulary trainer and Reading and Writing practice tool. Import your own material in documented formats, inspect the preview, and study. No AI account, prompt, API key, server account or paid service is required. The software does not generate questions, guess answers or correct spelling automatically. One original word and one original question are bundled. This is not affiliated with the College Board and does not implement official adaptive scoring.

Created by Revelete9, derived from personal SAT tools with AI-assisted refactoring. Original private tools and study data are kept separately. Released under the MIT License.

Previously named SAT Study Kit. The name has changed, but local data paths, browser storage keys and portable clipboard format IDs have not. Existing libraries, learning progress and shared groups remain compatible.

Unified desktop app (Python 3.11+, Tkinter and pywebview). It reuses the original React website's typography, cards and colors inside a desktop window. Python still owns the local files, import validation, review queues and exam launcher. Nothing is uploaded. On Windows, the window uses the installed Edge WebView2 runtime.

On Windows, double-click Launch StudyWeave.cmd. It uses a project-local .venv, creates it with python if missing, and installs missing dependencies from requirements.txt into that same environment. First setup requires Internet access and Python 3.11+ with Tkinter. Later launches work offline when dependencies are already installed. It does not rely on the Codex runtime or change your global Python packages. Launch StudyWeave.cmd --check checks setup without opening a window. If a Python dependency is missing, run this launcher again rather than installing the unrelated package named webview; the package is pywebview.

If the interface has not been built yet (Node.js 22.13+ is needed to build it, not to run it):

cd web
npm ci
npm run build
cd ..

For a manual Windows launch after setup:

.venv\Scripts\python.exe desktop.py

On other platforms, create and activate a Python virtual environment, install requirements.txt, then run python desktop.py in that same environment.

• Word library: browse Word books → Groups → Words. Create books such as SAT, CET-4 or CET-6, and groups for a day's study or a category. These are user-created books, not bundled official word lists. Open a group to edit, search, paste or export. Study one group or choose Study whole book with Sequential / Random order.
• Question Banks: select or create a bank; rename it; add, edit, search, import and export questions.
• Organize books & groups: build study plans with two independent panes. Create books/groups, paste words, or drag existing material into another container. Back to word library returns to a single pane. No Send step.
• Practice: choose a question group and module durations, then Start exam. The original realistic simulator opens in a separate window. An already-open test is brought forward instead of replacing its answers.

Group creation, renaming, word/question editing and imports stay inside the main workspace. Save results and validation errors appear inline. Import instructions can be expanded when needed; previews must match the current input before saving. Cancel and × close a paste editor immediately without changing the library; the in-app clipboard remains available for another paste. Other changed editors show an inline choice to keep editing or discard changes. Only file selection/export, startup failures and closing an active exam need native dialogs. The standalone legacy simulator keeps its original bank-management UI.

Desktop words and review progress are saved in data/desktop-vocabulary.json. The first launch reads the local vocabulary catalog (or bundled example), but does not migrate or overwrite browser progress. Editing/importing keeps existing word IDs and recall records. Backups are made before library edits and new rounds; each rating is saved atomically. Export full backup includes books, words and progress; a group's Export includes only word content. Exports are written into data/exports, and the exact path appears in the status message. Review resumes at the next ungraded card; a completed round offers Start new round.

Existing desktop v1 groups are migrated into My vocabulary with a backup of the old file. IDs, counters and unfinished queues are preserved. Whole-book study resumes its current unfinished group, skips complete/empty groups and continues to the next unfinished group. Random mode shuffles groups and fresh word queues once, then saves their order. Changing mode does not reshuffle a partially studied group. New words added to a completed group make that group unfinished again. New groups join the book's plan. A new book round resets queues, not lifetime recall counters.

On recall cards, Space reveals/hides meaning; 1 remembers, 2 marks forgotten, 3 skips, Esc returns with progress saved, and ? / F1 toggles inline help. Reveal the meaning before rating 1 or 2. These shortcuts are inactive in library editors and the separate mock-exam window. The ? button shows the same help.

In Word library → Organize books & groups, choose a book in each pane. All groups lists that book's groups; click a group's title or choose it in the second dropdown to see its words. Search and learning-status filters remain available. Use checkboxes or Ctrl/Shift-click for multiple items. Words can be sorted by order, errors, error rate and encounters. Click the same sort again to reverse it.

• Drag groups into the other book's pane to move whole groups (not merge them).
• Drag words into an open group pane, or onto a group row in All groups.
• The Move/Copy selector chooses the operation; holding Ctrl during a drag forces Copy. The destination is highlighted while dragging; release to save. Escape cancels a drag. The lists scroll when dragging near their upper or lower edges.
• Moves keep word IDs and learning counters. Whole-group moves also retain the group's queue. Word moves remove their old queue entries and add one review in the destination queue if it exists. Other words' completed positions are retained.
• Copies get new IDs and fresh learning progress; source material is unchanged.
• Conflicting group names or word terms reject the entire transfer, without overwriting or silently skipping anything. Rename conflicts before retrying.
• Transfers save atomically with backups. Undo transfer restores the last transfer in this workspace; later library edits invalidate that undo.
• + Book / + Group opens an inline creator. Paste words is the only entry for adding words, whether one or many. Existing words still have Edit, with word, meaning, pronunciation and expansion/notes all visible. There is no manual word-number field; ordering is assigned internally for sequential study. The destination is captured when an editor opens and shown above its fields. New containers immediately accept drops.

Select one or more group checkboxes, choose Copy groups in-app, open the destination book, then Paste group. Review the preview and confirm. Inside a group, checked words can be copied individually; with no words checked, Copy copies the entire group. Paste whole groups at book level, and words inside a group. Copying works in both single-pane and two-pane layouts. Ctrl+C / Ctrl+V use the active pane and the selected clipboard mode; text fields keep normal text editing shortcuts.

• In-app is the default. It keeps a content snapshot in the current app window while navigating between books; it never reads or replaces your system clipboard. It is cleared when the window is closed or reloaded.
• System / share uses portable JSON text. Choose it in the toolbar, or use the separate Copy to system button without first changing modes. System copy replaces the system clipboard; the in-app copy is kept separately. Switching the mode alone does not read or write either clipboard.
• Read system clipboard and Use in-app copy are available in the paste editor. Existing input is not overwritten; clear it before switching its source. If clipboard access is unavailable, paste the shared text into the input manually. Cross-device transfer requires your own clipboard sync or text/file transfer; the app does not connect devices or upload anything.
• Group copies keep names, terms, pronunciation, meanings and notes, but not IDs, source paths or study history. They get fresh progress. Existing groups stay intact; duplicate names receive a (copy) suffix. The preview shows the final names.
• Whole-group pastes are atomic and can be undone with Undo in the organizer until a later library edit. Limits are 2 MB, 100 groups and 5,000 words per paste.

The older Tk interface is retained with python desktop.py --classic; it uses the same desktop data. Do not edit the same library in both interfaces at once. The new host detects external changes and rejects stale edits instead of replacing them. Browser-only vocabulary records are still separate and are not overwritten.

Inside a book choose Paste group, enter its name and paste all words. Inside an existing group choose Paste words to append. Click Preview, inspect the parsed contents, then save. The Example button fills only an empty input.

concise: brief and clear
obsolete: no longer used

For all word properties, use four fields; empty optional fields are allowed:

concise | /kənˈsaɪs/ | brief and clear | a concise explanation
obsolete | | no longer used | obsolete equipment

Numbers are unnecessary. Older numbered imports remain supported.

In Question Banks → Paste questions, append to the selected group or supply a title and choose Save as new bank. Plain text supports numbered question starts, A.–D. options, optional Passage:, Module:, Answer: and Explanation: labels. Unknown answers stay unknown; no answer or missing option is invented. Indent continuation lines that start with a number to distinguish them from a new question.

1. Which word means brief and clear?
A. Concise
B. Obsolete
C. Ambiguous
D. Contradictory
Answer: A

2. Which word means no longer used?
A. Concise
B. Obsolete
C. Ambiguous
D. Contradictory
Answer: B

Plain text is a defined format, not arbitrary document/OCR interpretation. Existing JSON, CSV, TSV, pipe, pairs and tagged formats remain available as applicable below.

Existing web vocabulary trainer (kept separately; Node.js 22.13 or later):

cd web
npm ci
npm run dev

Open http://127.0.0.1:47832. The bundled example works without an import script. In development, an existing data/vocabulary.json is used instead of the example. Production builds bundle only the example, never the private data directory. Previous browser records remain in sat-study-kit.web.v1:<pack_id>.

Question browser, editor and timed practice (Python 3.11+ with Tkinter):

python simulator.py

No third-party Python packages are required. Choose a bank to practice, or open Browse / Edit / Import Question Banks. Create an empty bank, write questions with the editor, or import a file. Search and browse questions without starting a timed test. Edit a selected question and export a bank as JSON. Edits to the bundled example or old data/questions.json create a separate editable bank; they do not overwrite those sources.

Use Word library → Paste group / Paste words on the desktop, or Word library in the browser-only version, to paste material or upload a UTF-8 .csv, .tsv, .txt or .json file. Select the format if auto-detection picks the wrong separator. Click Preview, inspect the words and expandable format corrections, then save. Imports append to the selected group, not replace it.

Without a header, table rows must have exactly one of these layouts:

• 2 columns: term,meaning
• 4 columns: term,phonetic,meaning,expansion
• 5 columns: sequence,term,phonetic,meaning,expansion

With a header, columns can appear in any order. term and meaning are required; phonetic, expansion and sequence are optional. Aliases are word, definition, pronunciation, notes and number, respectively. Quote an entire field containing a separator or line break. Inside quoted fields, double ".

JSON word shape:

{"term":"concise","meaning":"brief and clear","phonetic":"","expansion":""}

• Remove a leading BOM; normalize CRLF/CR line endings; trim outer field whitespace.
• Normalize header case and the documented aliases.
• In pipe mode, convert full-width ｜ separators to |.
• Assign missing, invalid or conflicting sequence numbers without changing existing IDs.
• Skip identical repeated words within an import. Conflicting definitions for the same new word are rejected so the author can combine senses explicitly.
• Skip words already present in the destination group, preserving their existing content and learning history. Use Edit for intentional changes.

Corrections appear in the preview or final import notice. No spell-check guesses, automatic translations, missing definitions or unsupported image fields are added. A malformed row rejects the import rather than silently dropping that row. Limits: 2 MB and 5,000 input entries. Supplied study content is preserved; the application interface, documentation and generated messages are English.

The manager supports UTF-8 JSON, CSV, TSV and tagged text. Imports require a preview before Append to Selected Bank or Save as New Bank. The preview shows normalized questions, answers and corrections. New banks are stored under data/banks/; bank selection discovers every compatible JSON file there.

JSON accepts an array of questions or this envelope:

{
  "schema_version": 1,
  "title": "My practice",
  "questions": [{
    "id": "q1", "module": 1,
    "passage": "", "question": "Which word means brief and clear?",
    "options": {"A":"Concise","B":"Obsolete","C":"Ambiguous","D":"Contradictory"},
    "answer": "A", "explanation": ""
  }]
}

Required: nonempty question and all four nonempty options. Optional: id, module (default 1; allowed 1 or 2), passage, answer, explanation. A blank or null answer remains unknown and is not graded. An explanation is never generated.

CSV/TSV require headers. Required columns: question,A,B,C,D. Optional columns: id,module,passage,answer,explanation. Headers are case-insensitive. Aliases: prompt/stem -> question, text -> passage, correct -> answer, rationale -> explanation. Use standard CSV quoting for multiline text or embedded separators. Do not change the original option order.

Tagged text:

Module: 1
Passage: Optional passage.
Question: Which word means brief and clear?
A: Concise
B: Obsolete
C: Ambiguous
D: Contradictory
Answer: A
Explanation: Optional explanation.

Separate questions with a line containing ---. Unlabelled lines continue the previous field. Indent a continuation line if it begins with a field-like label. The ID, Module, Passage, Answer and Explanation labels are optional.

Automatic corrections normalize BOM, line endings, whitespace, headers/aliases, option-label case and lowercase/full-width answer letters. Missing or conflicting IDs are assigned. Exact duplicate questions (same content, module, answer and explanation, ignoring ID) are skipped. Conflicting answers are NOT silently merged. Unknown fields, incomplete options, invalid answer letters and malformed rows are rejected as a whole. Limits: 2 MB / 200 questions per bank. These are text-only formats: no OCR, images, audio, remote paths or math rendering.

• Vocabulary has recall cards, self-rating, groups, editing, searching, statistics, export and saved review position. An exported progress backup currently has no in-app restore command; it is not a standard word-import file.
• Missed words return until three consecutive successful self-ratings at distinct queue positions. When enough cards remain, the gaps are 3, 8 and 20 intervening cards. This is a same-session heuristic, not a validated optimal schedule or a calendar-based scheduler. No algorithm change is included in this release.
• Timed practice supports one or two modules, custom durations, navigation, review marks, option elimination, font controls and a line reader. Module 1 must contain a question to start a test. The answer summary shows supplied answers; it does not predict a SAT score. Test answers are held in memory, not saved for resuming after closing the app.
• Existing editable bank files are copied to data/backups/<timestamp>/ before a save. New bank filenames are generated IDs, never user-supplied paths.
• All imports are local. Materials, generated banks, backups, browser records, .env files, dependencies and build output must not be published.
• This repository contains only original demonstration material and synthetic test fixtures. Personal libraries, study histories, local backups and credentials are excluded. Commits use the maintainer's GitHub no-reply address.
• The old AI-specific prompt and command-line study-pack workflow have been removed. Legacy local banks are read for continuity; newly saved banks use the plain format.

Repeated retrieval and spacing have support, but a fixed card gap is not a universal optimum. Cepeda et al. (2008) found that useful spacing depends on the target retention interval. Their study does not validate this app's 3/8/20-card rule.

Rawson and Dunlosky (2022) review successive relearning: reach a recall criterion and revisit the material in later spaced sessions. They caution against treating one session count as universally optimal. Three correct self-ratings in this app are therefore not evidence of permanent mastery. A future scheduling change should distinguish within-session recovery from between-day retention, with feedback after errors.

python -m unittest discover -s tests -v
cd web
npm test
npm run build