Open-Source Desktop AI Agent App Runs Locally on Ollama

The maintainer of the open-source desktop application dodaios demonstrated running local AI agents using Ollama with Qwen models on a 16GB GPU. The demonstration highlighted multi-agent workflows, local voice tools, and the current challenges of local model tool calling. This showcase proves the viability of consumer hardware for running private, multi-agent AI workflows entirely offline without relying on cloud services. It highlights the current capabilities and limitations of running medium-sized open-source models for complex agentic tasks. Tested on an RTX 5070 Ti with 16GB VRAM, the app utilized Qwen models taking 71 to 116 seconds for responses, though local models failed at complex tool-calling routines. The application is built with a Python FastAPI backend and a React frontend, featuring local memory wikis and multi-level risk checks.