Single-pane local dashboard
System and GPU health, disk, uptime, host OS/kernel/driver versions, and live telemetry for your local AI workstation or lab box.
Public repo · active development
Cortex
A command center for your local AI stack.
Cortex is a self-hostable web UI for local AI systems: dashboard, GPU/system telemetry, installed model inventory, model discovery and pull, services catalog, and tool management from one trusted-LAN command surface.
Models dashboard
Shows the installed local Ollama inventory on the dashboard, including resident models and context-window information, so the box is easier to understand at a glance.
Model discovery + pull
Search Ollama Library and Hugging Face, manually refresh for same-day model drops, and pull selected models into local Ollama with live progress. Discovery only; Cortex does not host remote models.
Services catalog
A configurable grid for local tools such as Ollama, Open WebUI, ComfyUI, Jupyter Lab, Synapse, and more, each with health checks and direct links.
Built for DGX Spark / GB10
Developed on NVIDIA DGX Spark and tuned for GB10 unified-memory telemetry, where GPU power draw is the reliable activity signal.
Runtime theming
Rebrand the name, logo, subtitle, and color palette through runtime config without rebuilding the app.
Trusted-LAN controls
Optional tool installs, Ollama updates, and Tailscale controls are off by default and gated behind deliberate local opt-in.
Screenshots
Cortex in active development.
Current v0.5.0 surfaces include the live dashboard, local services catalog, and model discovery + pull page for local Ollama.
Status
Active development, not an internet-exposed hosted service.
Cortex is public and installable for trusted local environments. It is designed for a single trusted machine or private LAN, not direct public internet exposure.
The Cortex repository is now tagged v0.5.0 on main at 87162d9. Current capabilities include dashboard telemetry, Server-Sent Events performance streaming, installed Ollama model inventory, the new model discovery + pull page, config-driven services, auto-discovery, runtime theming and theme selector, settings, security docs, opt-in tool installs, opt-in Ollama updates, and opt-in Tailscale control.
Model discovery searches the Ollama library and Hugging Face, supports manual refresh for same-day model drops, and pulls selected models to the local Ollama instance with live progress. It is discovery and local pull orchestration, not hosted remote-model execution.
The security model is explicit: Cortex has no built-in authentication and is intended for a trusted LAN or local lab environment. Put authentication in front of it before exposing it beyond that boundary.