Skip to content
Filippo Galbo

Applied AI / active

Exocortex Local AI System

A local desktop console for voice interaction, a room display, source collection, and reviewed publishing.

Back to work

The problem

Voice, local-model routing, source collection, and publishing ran as separate programs. Their settings, errors, and current state were difficult to inspect together.

My role

Desktop console, voice routing, source collectors, and publishing controls.

What I built

  • Desktop console with settings, logs, and diagnostics
  • Voice interaction and room-display states
  • Source collectors and private publishing drafts

The result

A local system with tested components for bringing voice, source status, and publishing review into one operator console.

Architecture illustration of the Exocortex console coordinating voice, source collectors, the room display, and reviewed draft publishing.
Architecture illustration of the Exocortex console coordinating voice, source collectors, the room display, and reviewed draft publishing.

Statusactive

Evidencetested

Record reviewed

Implementation

How it works

A Tauri console manages settings, logs, collectors, and publishing controls. A Python room process handles wake phrases, microphone and speaker selection, speech output, command routing, and display states. A Node sidecar gathers approved sources and creates private drafts for review.

Technical context and methods

The implementation combines a React/Tauri console, shared TypeScript schemas, a Node collector, and a Python room process. Commands use named action routes, while collected material stays in draft form until reviewed.

  • Tauri
  • React
  • TypeScript schemas
  • Node sidecar
  • Python Presence agent
  • Piper TTS
  • Windows SAPI
  • Ollama / local LLM routing
  • Room-display rendering
  • Scoped Git publishing