Android AI Text Command Skill -- 2

I’m building an Android-only mobile app and I need a self-contained AI “command skill” that works purely through typed input. The goal is simple: when a user enters a natural-language request in a text field, the skill should recognise intent, extract any required parameters, and hand the result back to the main app so we can trigger the correct feature or workflow.

Here’s what matters most to me:
• Accurate intent recognition on common, short queries as well as longer, conversational text.
• Fast response times; round-trip latency under a second when using a cloud NLP service, or comparable speed when running locally with a lightweight on-device model.
• A clean Android interface (Java or Kotlin—your choice) that exposes one callable function: sendText(String userInput) → returns a JSON object containing intent, entities, confidence score, and a suggested reply string.
• Straightforward extensibility so I can add or modify command definitions without touching core code; a JSON or YAML config file is fine.
• Robust error handling so the module never crashes the host app—even with malformed or empty input.

You’re welcome to leverage Dialogflow, Wit.ai, ML Kit, TensorFlow Lite, or another NLP stack if it keeps the footprint reasonable and can run under current Play-store requirements. My one constraint is that the final artefact must compile cleanly in Android Studio and include clear setup notes for any external keys or model downloads.

Deliverables:
1. Android Studio project (module or library) with the AI text command skill implemented.
2. Sample Activity demonstrating usage.
3. Brief README covering build steps, how to add new commands, and where to plug credentials if a cloud API is involved.
4. One short video or set of screenshots showing the skill correctly identifying at least three sample commands.

I’ll test the module against my existing app as the acceptance step; once it plugs in and parses the demo commands reliably, we’re good to go.