CodeSensei

Native Android learning app that gives beginners fast, explainable feedback on Kotlin code.

Watch the demo

CodeSensei is deliberately local-first: its current analyzer uses deterministic rules to identify common beginner mistakes, explain what happened, and suggest a concrete fix. That makes every result inspectable instead of hiding feedback behind an opaque remote model.

Engineering proof

Area	Implementation
UI	Kotlin and Jetpack Compose with light/dark themes
State	ViewModel-driven screen and analysis state
Persistence	Room for analysis history and DataStore for settings
Navigation	Navigation Compose
Analysis	Local, deterministic rules with explanations and suggested fixes
Platform	Compile/target SDK 36; minimum SDK 24

Current capabilities

• Compose-based editor and results experience
• Local checks for common beginner errors, including misspelled output calls, missing entry points, and unbalanced parentheses
• Human-readable explanation and suggested correction for each finding
• Learning points and saved analysis history
• Persistent settings and theme support
• No account or network connection required for the current analyzer

Architecture

flowchart LR
  Editor[Compose editor] --> VM[Analyzer ViewModel]
  VM --> Rules[Local rules]
  Rules --> Result[Explanation and fix]
  VM --> Room[(History)]
  VM --> Store[(Settings)]

Loading

Screens

Home	Results
History	Settings

Dark-mode variants are also implemented in the app.

Run locally

1. Clone the repository.
2. Open it in a current Android Studio release.
3. Install Android SDK 36.
4. Sync Gradle and run the app configuration on an emulator or device running API 24+.

Command-line verification:

./gradlew assembleDebug
./gradlew test

Repository map

CodeSensei/
├── app/src/main/       Compose UI, state, storage, and analyzer logic
├── app/src/test/       Local unit-test source set
├── gradle/             Gradle wrapper configuration
└── README.md           Product and engineering overview

Scope and next steps

The current rule engine is intentionally small and deterministic; it is not a Kotlin compiler or production static-analysis replacement. The test source set currently contains starter scaffolding. Strong next steps are rule-level unit coverage, parser-backed diagnostics, accessibility tests, and opt-in AI explanations with clear provenance.

Author

Built by Muhammad Imran — portfolio · LinkedIn