Solo builder · 2026
Offline Personal Assistant (Hermes)
Hermes Agent running on a personal PC with a local LLM (qwen3.5:9b).
Most AI assistants send every prompt to a remote API. Hermes goes the other way: a local qwen3.5:9b model on a personal PC, wrapped in an Electron desktop app with a Python backend and SQLite for persistent memory. The goal was a useful daily assistant that stays private, works offline, and costs nothing per query.
Problem
Cloud assistants are convenient but leak context, need connectivity, and rack up API costs for always-on use.
Approach
A Hermes agent loop backed by qwen3.5:9b locally, with YAML-configured tools, SQLite memory, and a React + Electron UI for day-to-day interaction.
Outcome
A working offline assistant demo — private, self-hosted, and runnable entirely on personal hardware. Walkthrough on YouTube.
Stack