NekoTask

NekoTask is an Android to-do app with an AI assistant, designed around low cognitive load for ADHD-friendly task capture and follow-through.

The app is currently a functional prototype. It is useful for local development and experimentation, but it is not a medical tool and should not be treated as production-ready.

Screenshots

Images will be added here as the current UI stabilizes.

Task List	AI Chat	Settings

What It Does

• Manual task capture with a compact Material 3 bottom sheet, priority selection, and a two-step date/time picker.
• Hierarchical tasks and subtasks with completion state, ordering, and progress display.
• AI chat for natural-language task management, shown either as peek bubbles or a fullscreen overlay.
• AI subtask division for turning a larger task into smaller executable steps.
• Long-term memory, stored locally, for persistent assistant context managed from Settings.
• Debug-only sample data seeding and reset tools for local development.

Why

Many task apps make users decide too much before they can start. NekoTask aims to keep the default workflow small:

• capture quickly,
• split vague work into concrete steps,
• keep the main task list scannable,
• let AI help without replacing direct manual controls.

The cat theme is intentionally lightweight: it should make the app approachable without adding visual noise.

Current Implementation

• UI: Jetpack Compose + Material 3, single MainActivity.
• State: MainViewModel coordinates feature-level state and side effects through section coordinators.
• Data: Room is the source of truth for tasks and long_term_memories.
• AI: JetBrains Koog AIAgent with tool-based task and memory actions.
• Providers: OpenAI and DeepSeek are supported through the app's Settings screen.
• Networking: Ktor is adapted into Koog through KtorKoogHttpClient.
• Time and serialization: kotlinx.datetime and kotlinx.serialization.

Some visible drawer entries, such as Notes and About, are placeholders.

Setup

Open the project in Android Studio or build from the command line with the included Gradle wrapper.

For debug builds, you can provide a default OpenAI key in local.properties:

OPENAI_API_KEY=your_api_key_here

Release builds do not compile the developer's API key into the app. Users are expected to enter their own provider key in Settings. AI features require a valid provider key and network access.

Useful build check:

./gradlew :app:compileDebugKotlin

Documentation Split

• English documentation is preferred for project-level truth. If English and non-English docs disagree, use the English project docs and the code as the source of truth, then update stale docs.
• README.md explains what the app is and why it exists.
• AGENTS.md is the implementation and maintenance guide for agents and developers.
• docs/SUBTASK_DIVISION.md documents the current subtask division module.

When the implementation and documentation disagree, treat code as the source of truth and update the docs.

Near-Term Direction

• Keep the task flow low-friction and manual-first.
• Make AI actions more reliable and transparent.
• Reduce orchestration weight in MainScreen as features grow.
• Tighten repository write APIs where UI needs explicit success or failure.
• Replace placeholder drawer entries with real features or remove them.