Skip to content

Feature overview

This page is the canonical answer to "what can Adaptive Learner actually do?". It lists every major user-facing capability of the app, grouped by theme, and is kept current with each release. Other places (README, the individual help pages) link here instead of maintaining their own copies of this list.

Learning core

  • Six learning methods (deductive, inductive, error-based, dialogic, contextual, AI-adaptive) with bespoke AI prompts per method and step.
  • Seven-step session cycle: input, focus, attempt, feedback, refine, transfer, integrate. A dual-prompt evaluator judges every turn and decides whether to advance, repeat, skip ahead, or step back.
  • Auto-loop: when a topic is integrated, the session picks a new subtopic and starts a fresh cycle (capped per session).
  • Method switching: stagnation detection recommends a different method when ratings flatline; one-click accept.
  • Placement assessment (optional, resumable) that computes a six-method learning profile; a two-field quick start works without it.

See Learning sessions and The learning method.

AI tutor chat

  • Session chat built on assistant-ui: streaming token-by-token replies, markdown rendering, theming, and full localization.
  • Voice: microphone dictation into the chat, read-aloud for replies, and a dedicated pronunciation practice mode.
  • Bring your own key: Anthropic Claude, OpenAI GPT, and Google Gemini as separate provider plugins; live model discovery with a recommended/all picker; per-provider key test and a key vault with rollback.
  • Imported conversations continue as tutor sessions, keeping the original topic and analysis context.
  • "Ask AI" on theory blocks and exercises, and AI replies always in the learner's UI language.

Exercise types

Six core types every set can use, plus five extension types a set can ship:

Core type What the learner does
Matching Pair terms across two columns (start from either side)
Picture choice Pick the matching image
Free text Type the answer (typo tolerance, multiple accepted answers, optional AI second opinion)
Cloze Fill blanks by typing, selecting, or multi-select
Word tiles Compose the answer from shuffled tiles (touch drag)
Multiple choice Single or multi answer
Extension type What the learner does
Categorization Sort items into buckets
Error correction Find and fix the mistake in a sentence
Reading comprehension Read a passage, answer questions
Graded quiz A scored mini-quiz
Audio dictation Listen and type what was said
  • Exercises are direction-aware (recognize vs. produce), show a per-exercise difficulty indicator, support code and formula content with syntax highlighting, and offer listen-first audio variants.
  • Wrong answers get token-level diff feedback; hints are staged and cost XP.

See Lessons for the learner view.

Lessons and learning mechanics

  • Seven ways to play a lesson or set: Practice, Exam (delayed feedback, pass/fail verdict, XP bonus), Timed, Reverse, Shuffle, Endless, and a gated "train errors" mode that replays only what went wrong.
  • Spaced repetition (SRS) on per-element error history: due-review queue, direction-aware mastery, exam-mode interval boost, and a configurable review-session length.
  • Adaptive lessons generated on demand from your own error patterns (rule-based, offline, no API key needed).
  • Error replay and a correction round at lesson end drill exactly the words you missed.
  • Lesson flow control: pause, resume at the exact step, autosave, and a paused-lessons widget on the dashboard.
  • 0-3 star ratings, favorites, next-step suggestions, auto-splitting of oversized lessons, and theory back-links from exercises.

Lesson authoring (Create-Lesson)

  • A no-API-key wizard builds a complete, shareable lesson: card editor with drag-and-drop and CSV import, image upload per card, templates, draft autosave, and preview in the real lesson player.
  • Every exercise is editable: all core types can be edited after generation, added by hand, and balanced; an extension-authoring wizard covers all five extension types, including an audio-file upload for dictation.
  • Book-text ingestion: paste textbook text, or upload a book file (EPUB, DOCX, TXT, Markdown) with a chapter picker, multi-select of detected sections, an automatic front-/backmatter exclusion heuristic, and batch lesson generation per section.
  • AI exercise generation (bring your own key) with a deterministic quality gate, regenerate-with-feedback, and batch generation for a whole set.
  • Manage your own lessons: edit any lesson of a multi-lesson set via a lesson picker, combine own lessons into a set, and pick a content domain (languages plus knowledge domains).

See Creating lessons.

Import and analysis

  • Chat-history import from ChatGPT, Claude, Gemini, and arbitrary markdown or pasted text.
  • AI analysis extracts topic, weaknesses, error patterns, the recommended method, vocabulary, and a suggested curriculum.
  • One click seeds a curriculum, starts a targeted session, or converts the analysis into a replayable offline lesson.

Content management

  • Content hub with Discover / My content / Import tabs, list or grid view, and a search/filter bar (language, level, domain, trust, AI-checked).
  • Downloadable lesson sets from public GitHub content repositories, cached for offline use; sets can be hidden via a manifest visibility flag.
  • Federated repositories: connect multiple own or third-party content repos (private repos via token), a recommended-repos section, and trust badges per source.
  • Community sharing: a four-step share wizard opens a real pull request against a content repo, with smart placement and duplicate detection; invitation codes support private sharing for coaches.
  • Per-set deep links and QR codes, a learning-path view with per-set mastery, book recommendations per domain, and a book companion section per set.

See Content browser, Discover, and Content repositories.

Gamification

  • XP and levels with a visible XP badge and per-lesson rewards.
  • Tiered badge catalog (bronze/silver/gold, locked badges stay visible with an unlock hint).
  • Streaks with a heatmap, and daily missions (up to three adaptive goals per day).
  • Celebrations: earned, intensity-configurable praise, milestone overlays, optional sounds, all reduced-motion-safe.

Exports and backup

  • Anki: AI-extracted flashcards reviewed in-app, exported as .apkg or .txt.
  • NotebookLM: a ZIP with summary, vocabulary, rules, errors, flashcards, and sessions, plus active-recall questions and a study guide.
  • Learning repository: per-project markdown artefacts (README, stats, cheatsheet, roadmap), downloadable as ZIP or git-committed in server mode.
  • Progress reports as Markdown or PDF; lesson results exportable for AI-assisted practice; native share sheet for results.
  • Backups: .alb ZIP backup covering the full data surface, save-to-disk, first-run restore, online-to-local migration, and a separate passphrase-encrypted .alk export for AI keys.

See Backup and restore.

Platform

  • Progressive Web App: installable, offline-capable, service-worker updates with an update banner, works fully in the browser.
  • Two storage modes: local-first (everything in browser IndexedDB, AI calls go direct to the provider, no server needed) or server mode (FastAPI backend with SQLite, multi-device).
  • Local-network sync between devices with QR-code pairing and conflict resolution.
  • Desktop launcher for Linux, macOS, and Windows: a Docker-based one-click self-hosting setup with context-aware Docker detection and self-diagnosis.
  • Eleven UI languages, fully translated, with a searchable language picker.
  • Open content format: lessons are plain JSON validated against a published schema; the app consumes the content engine as a package.

See Installation.

Accessibility and UX

  • WCAG-AA-verified themes (light, dark, colored presets, follow-OS auto mode), enforced by automated contrast checks.
  • Keyboard-first: global shortcuts with a help overlay, Enter advances lessons, Tab navigates cloze blanks.
  • Screen-reader support: landmarks, ARIA labels and live regions, chart data tables, dialog focus management.
  • Reduced motion is honored everywhere; read-aloud (TTS) for lessons and chat.
  • Context-sensitive in-app help: the help panel opens the article for the current view; every article links to this documentation site.