Strata Supervisor

Strata Supervisor is a web control panel that lets homelab users manage multiple Strata large language model configurations in one self-hosted Docker…

  • AI
  • Tools & Utilities
Strata Supervisor is a web control panel that lets homelab users manage multiple Strata large language model configurations in one self-hosted Docker container on NVIDIA or AMD GPUs. Strata is the local inference server for large mixture-of-experts models such as Qwen and DeepSeek. Built on the official Strata Docker image, it keeps the container running with no model preloaded at startup and lets you keep several model configs side by side, loading or unloading each one by hand from the browser. Strata downloads a model on its first load, and you can also point it at your own GGUF files. Each config gets its own RAM strategy (all experts in RAM, only uncached experts, a fixed RAM budget, or streaming from the model files) and its own multi-GPU role (main, helper, or off, across different cards). A loaded model serves an OpenAI- and Anthropic-compatible API on port 8080.

Platform availability

Available on 1 of 10 platforms

  • Unraid
  • TrueNAS (not listed)
  • Umbrel (not listed)
  • ZimaOS (not listed)
  • Proxmox (not listed)
  • Helm (not listed)
  • Mac (not listed)
  • Windows (not listed)
  • Linux (not listed)
  • Any server (not listed)

Health score?A 0-100 score computed nightly from four factors: maintained (40%: recent commits and steady activity), popular (25%: GitHub stars, log scale), easy to install (20%: platforms, install notes, setup guides, Docker image) and light to run (15%: minimum RAM, ARM support). When the source code has been checked, a fifth factor, security check (20%: automated scan for committed secrets, unsafe desktop-app settings, known-vulnerable dependencies and project hygiene; any serious finding caps it at 20), is added. Factors without data are left out and the rest are rescaled.

38/100 Low

Maintained
75last commit 12 days ago
Popular
00 GitHub stars
Easy to install
351 install platform, install notes, Docker image
Light to run
1032 GB minimum RAM

Availability: platform listings confirmed today

GitHub stars
0
Open issues
0
Last commit
2026-11-09
Latest release
v1.0.1
License
MIT
Activity
rising

Checked 12 days ago - source: GitHub (sognix/strata-supervisor)

Resources & Compatibility

Minimum requirements: 32 GB RAM.

Checked 12 days ago - source: Docker Hub image tags and published docs

First-install notes

  • Run setup first with a safe model size First run the setup script for your OS: ./setup.sh on Linux/macOS or START-HERE.bat on Windows, then use --help to pick a smaller --model (model size) or lower --context (conversation memory) if your machine is modest. Without a model size that fits, startup can fail or the app can crawl. source
  • Use the local address after setup After setup, open http://127.0.0.1:8080 in a browser on the same computer. It is limited to that machine, so opening it from another homelab PC will not work by itself. source
  • Check the saved setup if it won't start If the app does not start, run strata doctor with --profile pointing to your strata-model.json file. It checks the saved launch choices without starting the model, so you can spot a bad setting before spending more time. source
  • Choose the Docker image tag to match your GPU When installing, select the image tag that matches your graphics card. The default tag targets NVIDIA RTX 20-50 with driver 580 or newer; alternate tags cover older NVIDIA cards and AMD RX 7000, RX 9000, and Strix Halo. On NVIDIA you must enable the Nvidia-Driver plugin in Unraid. source
  • Hardware Deploy to Unraid or any Docker host with NVIDIA or AMD GPU support. source

Fetched 5 days ago - each note links to its source

Alternatives

No alternatives collected yet.

Common questions

Which ports should I expose for the web UI and LLM API?

The Docker example exposes port 8081 for the supervisor UI and port 8080 for the LLM API. Port 8080 is the OpenAI- and Anthropic-compatible endpoint.

Does it load or switch models automatically?

No; nothing loads at container start and nothing loads or swaps by itself. You load or unload model configs by hand in the browser, and switching models does not require a container restart.

Can I use my own GGUF files instead of only Strata catalog models?

Yes, you can pick a folder of your own GGUF files. Strata's catalog is also available with download size, disk status, and a fit marker for your cards and RAM.

How do I assign multiple GPUs to a model?

Use the multi-GPU roles option to assign each card a role such as main, helper, or off via --gpu-roles. Card numbers are taken from nvidia-smi, not from the code.

What does the LLM API return when no model is loaded?

The API port answers 503 with the message that no model is loaded in Strata. GET /v1/models lists nothing while no model is running.

Answers sourced from github.com

Community

No community discussions collected yet.