# System Facts

These are the durable facts that should be loaded when working with the user's local model setup.

## Machine

- OS: Ubuntu 26.04
- User: `wildlama`
- Home: `/home/wildlama`
- GPU: NVIDIA GeForce RTX 5090
- Approx VRAM: 32 GB

## Local AI services

- ComfyUI path: `/home/wildlama/comfy/ComfyUI`
- ComfyUI port: `8188`
- Ollama endpoint: `http://127.0.0.1:11434/v1`
- Ollama model: `qwen3-coder:30b-a3b-q4_K_M`

## Hermes profiles

- Default profile: hosted ChatGPT-backed profile
- Local profile: `local-ollama`
- Local profile wrapper: `local-qwen`
- Hosted profile wrapper: `chatgpt-mini`

## Network access

- Tailscale is available on the machine.
- Use the Tailscale IP for private access instead of exposing ports publicly when possible.