Voice-driven cross-pivot ("Show me Tatooine")
- Lingua principale
- TypeScript
- Stelle
- 0
- Fork
- 1
- Metriche di merge delle PR
- Nessuna PR unita negli ultimi 30g
Descrizione
> **Skills:** Web Speech API · NLP intent parsing · permission UX
> **Time:** ~10 hours
> **Good for:** Voice-UI engineers · accessibility advocates · power users
> **Difficulty:** Advanced
---
## Context
The Web Speech API's SpeechRecognition (Chrome / Edge / Safari 14.1+) is good
enough for command-style voice input. Combined with our hybrid search index,
we can resolve "show me Tatooine" → cross-pivot.
## Goal
A voice control mode: hold a key (or click a mic button) → speak a command
→ the system parses it via the existing search + a small intent grammar, and
executes (cross-pivot, plot route, change view, etc.).
## Where to start
- `components/explorer/VoiceControl.tsx` — wraps SpeechRecognition behind a
permission gate; show a clear listening indicator with waveform
- A small command grammar: `show me ` / `pivot to ` / `plot
route from to ` / `switch to (galaxy | timeline | lineage)` / `play
crawl`
- Use `runHybridSearch()` to resolve entity references in the recognized text
- Confirm action visually before executing (show the matched entity for ~600ms,
then pivot)
## Acceptance criteria
- "Show me Tatooine" cross-pivots to Tatooine
- "Plot route from Coruscant to Endor" enters route mode and selects both
- Recognition errors surface graceful inline error
- Mic permission requested politely; state preserved across sessions
- Reduced motion preserved
- Hover on mic button shows the recognized text history
## Reference
https://developer.mozilla.org/en-US/docs/Web/API/SpeechRecognition
Guida per i contributori
Apri la guida per i contributori
Valutazione
Questa issue non è ancora stata valutata.