gitwtfhub

wtf is kid-mode-jarvis?

hermes815/kid-mode-jarvis — explained in plain English

Analysis updated 2026-05-18

20TypeScriptAudience · developerComplexity · 4/5Setup · hard

TL;DR

A safety-walled JARVIS style voice assistant tablet app for kids, where a Node server, not the AI, controls ranks, facts, and rewards.

Mindmap

mindmap
  root((repo))
    What it does
      JARVIS voice for kids
      Server graded ranks
      Real call to parent
    Tech stack
      TypeScript
      Node.js
      Gemini Live
    Use cases
      Kid safe AI companion
      Reward ladder system
      Parent voice call
    Audience
      Parents
      Developers

Code map

Detail Auto

An interactive map of this repo's files and how they connect — its source is parsed live in your browser. Click Visualize to build it.

filefunction / class

Why would anyone build with this?

REASON 1

Set up a locked down tablet that gives your child a JARVIS style AI companion for lessons and missions.

REASON 2

Build a server-authoritative reward ladder so an AI can perform a character without controlling real outcomes.

REASON 3

Give a child a safe way to place a real voice call to a parent through an AI-narrated story.

What's in the stack?

TypeScriptNode.jsReactGemini LiveWebRTC

How it stacks up

hermes815/kid-mode-jarvisadrianhajdin/job_pilotanshaneja5/skyline-run
Stars202020
LanguageTypeScriptTypeScriptTypeScript
Setup difficultyhardhardmoderate
Complexity4/54/52/5
Audiencedeveloperdevelopervibe coder

Figures from each repo's GitHub metadata at analysis time.

How do you spin it up?

Difficulty · hard Time to first run · 1h+

Requires HTTPS (mic will not work over plain HTTP/LAN), a Gemini API key, Node 22+, and a tablet meeting specific hardware requirements.

Wtf does this do

This project turns a locked down tablet into a JARVIS style voice assistant for a young child, built on top of Google's Gemini Live example app. The child talks to a British accented, film inspired AI character that can teach lessons, run hands-on missions, and even place a real voice call to the child's dad. A tablet running this app has no browser, no app store, and no video app, only the console itself. The key design idea is a split between performance and truth. The Gemini voice model plays the JARVIS character and can improvise stories, missions, and lessons, but it never decides anything that has to be factually or practically true. A Node.js server, not the AI, is the one that grades whether the child answered a question correctly, awards ranks on a seven step ladder from Cadet up to Chief Engineer, confirms whether a physical building project is actually finished, and controls whether music in the house starts or stops. Any real factual question the child asks gets routed to a separate grounding AI service, or falls back to an in-character answer of asking their dad, so the voice model is never trusted as a source of facts. The top rank on the ladder can only be unlocked by completing a real physical project, something no AI transcript alone can confirm. Other features include an approval gated memory system, where anything the AI wants to remember about the child must be reviewed by a parent first, and a two way voice call to a parent framed inside the story as a satellite search. Optional integration with Home Assistant lets the AI turn on whole house music, again controlled server side so it cannot be faked. Getting it running requires a tablet with a modern browser, a secure HTTPS connection since browser microphones will not work otherwise, a Gemini API key, Node 22 or newer, and optionally a separate AI service to answer factual questions and a kiosk browser app to lock the tablet down. The project's documentation stresses that setup only takes about forty minutes, but writing the actual lesson content, the family details, and the reward story is the real ongoing work.

Yoink these prompts

Prompt 1
Help me set up this Kid Mode JARVIS project on a Samsung tablet with HTTPS via Tailscale.
Prompt 2
Show me how the Node server in this repo grades ranks and keeps the Gemini voice model from deciding real outcomes.
Prompt 3
Walk me through writing my own AUTHORING.md content, persona, and dad dossier for this project.
Prompt 4
Explain how the grounding LLM gateway in this architecture keeps the voice model from inventing facts.

Frequently asked questions

wtf is kid-mode-jarvis?

A safety-walled JARVIS style voice assistant tablet app for kids, where a Node server, not the AI, controls ranks, facts, and rewards.

What language is kid-mode-jarvis written in?

Mainly TypeScript. The stack also includes TypeScript, Node.js, React.

How hard is kid-mode-jarvis to set up?

Setup difficulty is rated hard, with roughly 1h+ to a first successful run.

Who is kid-mode-jarvis for?

Mainly developer.

View the repo → Decode another repo

This repo across BitVibe Labs

Don't trust strangers blindly. Verify against the repo.