To use TT-Studio's deployment features, you need access to a Tenstorrent AI accelerator.
Alternatively, you can connect to remote endpoints running models on Tenstorrent cards without local hardware.
TL;DR: TT-Studio is an easy-to-use web interface for running AI models on Tenstorrent hardware. It handles all the technical setup automatically and gives you a simple GUI to deploy models, chat with models, and more.
TT-Studio combines TT Inference Server's core packaging setup, containerization, and deployment automation with TT-Metal's model execution framework specifically optimized for Tenstorrent hardware and provides an intuitive GUI for model management and interaction.
Before you start, make sure you have:
⚠️ IMPORTANT: Complete the base Tenstorrent software installation first:Follow the Tenstorrent Getting Started Guide
This guide covers hardware setup, driver installation, and system configuration. You must complete this before using TT-Studio.
Also ensure you have:
- Python 3.8+ (Download here)
- Docker (Installation guide)
⚠️ IMPORTANT - Docker Group Setup:After installing Docker, you must add your user to the
dockergroup to run Docker commands withoutsudo:sudo usermod -aG docker $USERThen log out and log back in (or restart your system) for the group change to take effect.
To verify your user is in the docker group:
groups | grep dockerIf you see "docker" in the output, you're ready to go! If not, make sure you've logged out and back in after running the
usermodcommand.
Want to start using AI models right away on your Tenstorrent hardware? This is for you!
Quick Setup:
git clone https://github.com/tenstorrent/tt-studio.git && cd tt-studio && python3 run.py --easyWhat happens step by step:
- Downloads TT-Studio - Gets the code from GitHub
- Enters the directory - Changes to the tt-studio folder
- Runs the setup script - Automatically configures everything (easy mode: only prompts for your Hugging Face token; uses defaults for the rest)
- Initializes submodules - Downloads TT Inference Server and dependencies
- Builds containers - Sets up Docker environments for frontend and backend
- Starts all services - Launches the web interface and backend server
⚠️ Security Note: Easy mode uses default values that are NOT secure for production. Use this mode only for development, testing, and quick evaluation. For production deployments, use the standard setup (python3 run.py) or development mode (python3 run.py --dev).
📖 More Details: See the Complete run.py Guide for a full comparison of setup modes.
After Setup:
- Go to http://localhost:3000 to use TT-Studio
- The backend runs at http://localhost:8001
- Individual AI models run on ports 7000+ (e.g., 7001, 7002, etc.)
To Stop TT-Studio:
python3 run.py --cleanupNote: This command will stop and remove all running Docker containers, including any currently deployed models. It cleans up containers and networks but preserves your data and configuration files.
🎯 What Can You Do Next?
Once TT-Studio is running:
- Deploy a Model - Go to the Model Deployment page and deploy a model to start using AI features
- Use AI Features:
- 💬 Chat with AI models - Upload documents and ask questions
- 🖼️ Generate images - Create art with Stable Diffusion
- 🎤 Process speech - Convert speech to text with Whisper
- 👁️ Analyze images - Detect objects with YOLO models
- 📚 RAG (Retrieval-Augmented Generation) - Query your documents with AI-powered search
- 🤖 AI Agent - Autonomous AI assistant for complex tasks
📖 Learn More: Check out our Model Interface Guide for detailed tutorials.
🆘 Need Help?
- No Tenstorrent hardware? → Remote Endpoint Setup - Connect to remote Tenstorrent cards
- Issues during setup? → Troubleshooting Guide
- Questions? → FAQ
- Remote server setup? → See Remote Access Guide below
- Technical support? → Submit issues on GitHub
Want to contribute to TT-Studio or modify it?
Development Mode Setup:
git clone https://github.com/tenstorrent/tt-studio.git
cd tt-studio
python3 run.py --devDevelopment Features:
- Hot Reload: Code changes automatically trigger rebuilds
- Container Mounting: Local files mounted for real-time development
- Automatic Setup: All submodules and dependencies handled automatically
Get Started:
- Contributing Guide - How to contribute code
- Development Setup - Set up your dev environment
- Frontend Development - React frontend
- Backend API - Django backend
Resources:
Running TT-Studio on a remote server? Use SSH port forwarding to access it from your local browser:
ssh -L 3000:localhost:3000 -L 8001:localhost:8001 -L 7000-7010:localhost:7000-7010 username@your-serverNote: Port range 7000-7010 forwards the model inference ports where individual AI models run.
Then open http://localhost:3000 in your local browser.
Hardware Requirements: Tenstorrent AI accelerator hardware is automatically detected when available. You can also connect to remote endpoints if you don't have direct hardware access.
TT-Studio combines TT Inference Server and TT-Metal to provide:
- Modern Web Interface: React-based UI for easy model interaction
- Django Backend: Robust backend service for model management and deployment
- Vector Database: ChromaDB for document storage and semantic search
- Multiple AI Models: Chat, vision, speech, and image generation
- Model Isolation: Each AI model runs on separate ports (7000+) for better resource management
- Hardware Optimization: Specifically optimized for Tenstorrent devices
- Docker Containers: Isolated environments for frontend, backend, and inference services
- Language Models (LLMs): Chat, Q&A, text completion
- Computer Vision: Object detection with YOLO
- Speech Processing: Speech-to-text with Whisper
- Image Generation: Create images with Stable Diffusion
Want to contribute or customize TT-Studio?
Get Started:
- Contributing Guide - How to contribute code
- Development Setup - Set up your dev environment
- Frontend Development - React frontend
- Backend API - Django backend
Development Mode:
python3 run.py --dev # Enables hot reload for developmentDevelopment Features:
- Hot Reload: Code changes automatically trigger rebuilds
- Container Mounting: Local files mounted for real-time development
- Automatic Setup: All submodules and dependencies handled automatically
Resources:
- FAQ - Quick answers to common questions
- Troubleshooting Guide - Fix common setup issues
- Model Interface Guide - Detailed tutorials for using AI models
- Complete run.py Guide - Advanced usage and command-line options
- Having issues? Check our Troubleshooting Guide
- Want to contribute? See our Contributing Guide
- Need specific models? Follow our vLLM Models Guide
⚠️ Note: Thestartup.shscript is deprecated. Always usepython3 run.pyfor setup and management.