Projects
Electron.jsNode.jsOllamaChromaDBGPT-SoVITSWebSockets

Aiko – AI Desktop Companion

Local-first AI companion with offline LLM, TTS, and persistent memory

Overview

Aiko is a local-first AI desktop companion built with Electron.js, designed to run entirely on-device with no external API calls. Every component — LLM inference, text-to-speech, and long-term memory — is self-contained, making it completely private and available offline.

Highlights

  • Engineered offline LLM inference via Ollama and real-time TTS via GPT-SoVITS, delivering a fully private, on-device AI experience with zero cloud dependency.
  • Built a retrieval-augmented memory system using ChromaDB, enabling contextual recall across sessions, personality consistency, and long-term interaction patterns.
  • Developed a synchronized streaming pipeline over WebSockets, aligning LLM token generation with TTS playback for natural, low-latency conversation flow.
  • Optimized local inference and memory retrieval latency to run smoothly on consumer hardware without GPU acceleration.

Stack

Electron.jsNode.jsOllamaChromaDBGPT-SoVITSWebSockets