Flagship · Second brain & voice AI
Araba — my “Momento Media World” second brain
A local-first app that renders my entire creative studio as a living knowledge graph — 3,900+ nodes and 4,100+ links — and fronts it with a voice assistant I named Araba. Ask it anything about my work out loud and it answers grounded in the actual files, naming the source it used.
What it does:
- Indexes the whole vault into a searchable 2D / 3D force-graph — focus a node, trace the path between two, and watch it rebuild live as the studio changes.
- Araba, a British-butler voice assistant that hot-swaps models (Fable 5 / Opus / Sonnet / Haiku) mid-conversation and never goes silent (ElevenLabs → local → browser fallback).
- An “Agentic-OS” layering — Core, Skills, Memory, Routines, Applications — plus a governed Job Search Mode that scans boards and match-scores roles behind a hard anti-fabrication guard.
- Node (zero-build)
- Force-graph 2D/3D
- ElevenLabs TTS
- Speech I/O
- SSE live-sync
- Claude CLI bridge


