TT-Studio

A web UI for deploying and chatting with AI models on Tenstorrent hardware. It wraps TT Inference Server packaging and TT-Metal execution behind a Django + React + agent stack.

No Tenstorrent hardware? You can still use it — point it at remote endpoints running on cards elsewhere.

---

Table of Contents

• Before you start
• Quickstart
• Documentation
• Community & License

---

Before you start

You'll need:

• Python 3.8+ and Docker installed
• Your user in the docker group so you don't need sudo — sudo usermod -aG docker $USER, then log out and back in
• A Hugging Face token for any gated models you want to run (Llama, etc.)
• First time on Tenstorrent hardware? Do the Getting Started Guide first.

Full prerequisites are in the detailed setup guide.

---

Quickstart

git clone https://github.com/tenstorrent/tt-studio.git
cd tt-studio
python3 run.py

run.py handles the rest — the tt-inference-server artifact, your .env, the right Docker overlays for your hardware, and all the containers. It asks for your Hugging Face token along the way. When it finishes, open http://localhost:3000.

A few flags are worth knowing:

• python3 run.py --dev — development mode: mounts your local source so the backend and frontend hot-reload as you edit.
• python3 run.py --purge-all — tear everything down and wipe the persistent volume and .env for a clean slate. (Use --stop instead to stop the containers but keep your data.)
• python3 run.py --report-bug — bundle your logs into a ZIP and open a pre-filled GitHub issue (also offered automatically if setup errors out).
• python3 run.py --install-shortcut — add a tt-studio shell shortcut so you can launch from any directory without typing python run.py.

Hardware is auto-detected. On a QB2 (QuietBox) you can set IS_QB2=true in .env to have startup verify the board via tt-smi; it's off by default so dev laptops and cloud runs aren't held to that check. See the run.py guide for the full behavior. Run python run.py --help for all flags and python run.py --help-env for every environment variable.

That's all most people need. Everything else — hardware modes, environment variables, the dev workflow, remote access, and troubleshooting — lives in the detailed setup guide.

---

Documentation

• Detailed setup & usage — hardware modes, env vars, dev workflow, remote access, troubleshooting
• run.py reference — every flag and environment variable, explained
• Troubleshooting and FAQ
• Remote endpoints — use TT-Studio without local hardware
• Contributing — branching strategy and PR standards

---

Community & License

• Issues / feature requests — GitHub Issues
• Contributing — CONTRIBUTING.md
• License — Apache-2.0 (© Tenstorrent AI ULC)