Can AI replace Speak?

Build a closest honest personal substitute for Speak in an empty repository.
Use Next.js 15, TypeScript, SQLite, Drizzle ORM, and a service worker for offline use; do not offer alternative stacks.
The core loop is: run scripted language roleplay with clear lesson objectives using a user-supplied model, schedule retrieval practice, and show progress without claiming accredited instruction.
Make the first run work locally with one documented command.
Store all user data locally by default and make export straightforward.
Put secrets in .env, ship .env.example, and never commit credentials.
Import Markdown, text, or user-authored notes and split them into editable learning units.
Create flashcards and short quizzes only after showing the generated items for approval.
Implement a transparent spaced-repetition schedule with due, again, hard, good, and easy actions.
Add goals, study sessions, streak-free progress charts, notes, and weak-topic review.
For language practice, record the learner and compare timing or pronunciation without presenting medical claims.
Support offline use, full export, reset, backup, and deletion of all learning data.
Include clear empty, loading, success, and recoverable error states.
Add input validation, safe filenames, and graceful handling of unavailable APIs.
Write focused tests for the core transformation and one end-to-end happy path.
Create a README with setup, architecture, permissions, data location, and backup steps.
Do not add accounts, billing, telemetry, analytics, or a hosted control plane.
Do not claim to reproduce proprietary data, network liquidity, regulated access, or frontier infrastructure.
Deliberately leave out licensed premium courses.
Deliberately leave out accredited certificates and grading.
Deliberately leave out live tutors, large classrooms, and a learner marketplace.
Finish by running the tests and listing the exact commands used.