Question Box
AI-powered Q&A tool that turns any question into visual, interactive explanations.
$ cat story.md
Built this for 10-15 minute daily learning sessions. Kids ask questions like 'Why is the sky blue?' and instead of a wall of text, they get labeled diagrams, step-by-step breakdowns, and quizzes to check understanding.
The key insight from json-render: constrain AI output to a vocabulary of components you define. Claude can only generate what we've built renderers for. Safe, predictable, visual.
Phase 1 is a personal instance with Claude + Serper search. Phase 2 (if needed) makes providers pluggable so others can use OpenAI, Ollama, etc.
$ git status
Phase 1 built: 6 components, Claude integration, Serper search, chat UI. Ready for testing with API keys.
$ ls ./components
- -Structured output from LLMs (JSON generation)
- -Component-constrained AI rendering
- -Search augmentation for current events
- -Building for kids (age-appropriate UX)
?Add streaming for progressive rendering?
?Worth adding the actual json-render library or keep it simple?