AIBook — AI Language Learning Reader
v1.1.0 · Mobile-first PWA · Next.js 16 · Gemini AI · Supabase Cloud
An intelligent application for learning foreign languages through reading. Open a book, tap on any word, and receive an instant analysis: translation, grammar, context, and examples. Listen to the text with interactive karaoke highlighting and sync your progress across all devices.
Screenshots
| Home | Reader | AI Panel | Flashcards |
|---|
| Dashboard with progress | Book text with highlights | Word/phrase analysis | List of saved words |
New in v1.1.0 🚀
🎙️ Audio & TTS (Karaoke Mode)
- Interactive Audio: Listen to any sentence or paragraph using high-quality AI voices.
- Dual Engine: Support for Google Gemini TTS (natural AI voices) and Native Browser TTS (offline capability).
- Karaoke Highlighting: Real-time visual tracking of spoken words. The text highlights as the audio plays.
- Audio Scrubber: Precise control over playback with seeking, pausing, and repeat mode.
☁️ Cloud Sync (Supabase)
- Multi-device Sync: Your library, reading progress, and flashcards are now synced to the cloud via Supabase.
- Persistent Profile: Save your native language and target language preferences across sessions.
- Progress Tracking: Never lose your place in a book, even if you switch devices.
🧠 Global AI Cache
- Cost Efficiency: Shared global cache for word and sentence analysis. If another user has already analyzed a word, you get the result instantly without an API call.
- Performance: Instant dictionary lookups for frequently used words.
Core Features
Book Library
- Load TXT and EPUB files from the device.
- Automatic parsing and text splitting into paragraphs.
- Book language detection via filename (
_de, -en, -fr, -es).
- Books are stored locally and synced to the cloud for authenticated users.
Reader & AI Analysis
- Context-Aware Analysis: Tap any word to get its lemma, part of speech, grammar (gender, case), and context-specific translation.
- Phrase Detection: Automatic detection of idioms, collocations, and compound words.
- Sentence Breakdown: Full translation and structural analysis of the containing sentence.
- Interactive Examples: 5 usage examples for every word. Tap any word within the examples to explore further!
Flashcards (SRS Ready)
- Save words, phrases, sentences, or fixed expressions with one click.
- Automatic back-side generation using AI analysis.
- Status tracking:
new, due, learning.
Usage Guide & Workflow
1. Setting Up
- API Key: Go to Settings and enter your Google Gemini API Key.
- Languages: Choose your Native Language (for explanations) and Target Language.
2. The Reading Workflow
- Import: Open Library, click "+", and upload your book.
- Read: Tap a book cover to start. Scroll naturally; your progress is saved automatically.
- Explore:
- Tap a word: Opens the bottom panel with quick info.
- Open Modal: Click the expand icon for deep analysis and usage examples.
- Interactive Examples: Tap words inside the examples to see their definitions without leaving the current context.
- Listen:
- Tap the Speaker icon to start playback.
- Watch the Karaoke Highlighting to associate sounds with text.
- Use the Scrubber at the bottom to replay specific parts.
- Save: Click the Save icon in the AI panel to add the word/phrase to your flashcards.
3. Reviewing
- Open the Flashcards tab to review your vocabulary. Your progress here is synced to your account.
Connecting an outside AI agent (MCP)
Settings → Подключение ИИ-агентов issues a personal, HMAC-signed URL
(/api/mcp/<token>) that turns the app into an MCP server. ChatGPT, Claude,
Gemini CLI and anything else that speaks the protocol can then read the
learner's state and write into their app. Every call is scoped to the token's
owner and is plain database work — no Gemini spend.
The agent on the other end is the language model: it writes the texts and picks
the words, and these tools store them exactly the way the in-app generator does.
| Area | Tools |
|---|
| Discovery | get_overview, get_capabilities |
| Flashcards | list_flashcards, add_flashcards, update_flashcard, delete_flashcards, get_study_words |
| Dictionary & batches | list_word_batches, list_batch_words, search_dictionary, add_word_batch, add_words_to_batch |
| Learning quality | get_progress, get_training_summary, get_training_history (docs) |
| Texts | create_lesson, list_texts, get_text, list_catalogue |
| Interactive tests | get_assessment_capabilities, create_assessment, update_assessment, prepare_assessment_audio, get_assessment_status, publish_assessment, list_assessments, get_assessment_results, submit_assessment_review, get_learning_gaps, check_dictionary_words, get_speech_service_status |
Interactive tests are taken by the learner at /test/<id> (and listed on the home
screen): reading, listening (recorded by the app itself with Gemini TTS, with a
server-side play limit), choice, gaps, word order, short answers, translation of a
ready Russian text, writing, and speaking — read aloud, listen-and-repeat and free
spoken answers assessed by Azure Pronunciation Assessment (AZURE_SPEECH_KEY,
AZURE_SPEECH_REGION) — in a learning or a diagnostic mode. The learner can tap
any word for a translation or mark it unknown; the teacher agent sees both. Schemas, rules and examples for the teacher agent:
docs/assessments-mcp.md.
Beyond tools, the server also serves prompts (ready-made flows the learner
can pick from a menu: «рассказ из моих слов», «что я забываю») and resources
(aibook://guide, aibook://state, aibook://progress). All three are built
from one description of the app in lib/mcp/capabilities.ts, because clients
disagree about which of them the model actually sees — ChatGPT's connector UI,
for instance, drops the server's instructions. lib/mcp/tools.test.ts fails
the build if a tool exists that the capability map does not mention, which is
what keeps a newly shipped feature from being invisible to connected agents.
Architecture
aibook/
├── app/
│ ├── layout.tsx # Root layout, PWA metadata, CSS imports
│ ├── page.tsx # Main SPA router (section state machine)
│ ├── auth/ # Auth pages (Login/Signup)
│ └── api/
│ └── ai/
│ └── analyze/
│ └── route.ts # Server-side proxy → Gemini API
│
├── components/
│ ├── ui/
│ │ ├── AppShell.tsx # Application shell with bottom navigation
│ │ ├── AudioScrubber.tsx # Interactive TTS controls
│ │ └── SpeakButton.tsx # TTS trigger
│ ├── auth/ # Authentication components
│ ├── home/ # Dashboard widgets
│ ├── library/ # File upload, book list
│ ├── reader/ # Reader view with word highlighting
│ ├── ai-panel/ # Side panel for AI analysis
│ ├── word-modal/ # Modal window for word analysis
│ ├── cards/ # Flashcards list
│ └── settings/ # Settings & Profile
│
├── lib/
│ ├── ai/ # Gemini API logic & prompts
│ ├── auth/ # Supabase Auth hooks
│ ├── db/ # LocalStorage & Supabase CRUD
│ ├── parser/ # TXT/EPUB parsers
│ ├── tts.ts # TTS Engine (Gemini/Browser)
│ └── config.ts # App constants
│
├── styles/ # Vanilla CSS design system
└── public/ # PWA assets & manifest
Technology Stack
| Technology | Role |
|---|
| Next.js 16 | Framework |
| React 19 | UI Library |
| Supabase | Auth, Database, Cloud Sync |
| Gemini AI | Text Analysis & TTS |
| epubjs | EPUB Parsing |
| Lucide | Icons |
Roadmap (v1.2+)
Privacy
Your data is private. While we use cloud sync, your Gemini API key remains in your local storage and is never stored on our servers.
License
ISC © 2025
