All work
PersonalPersonal project

Local multi-agent LLM system

Routes coding and vision tasks across self-hosted models with Ollama.

Problem
I wanted coding and vision tasks handled by the model best suited to each, running on my own hardware.
My role
Personal project, built end to end.
Stack
  • Python
  • Ollama

Architecture

  1. Incoming task

    Coding or vision.

  2. Router

    Picks a model for the task type.

  3. Self-hosted models

    Served locally through Ollama.

  4. Result

    Returned to the caller.

Architecture of Local multi-agent LLM system, in order

Result

  • Runs fully on self-hosted hardware.