2. AI
  3. TBT5-AI Software Installation Guide

TBT5-AI Software Installation Guide

Last Update: March 12th, 2026
Article ID: 3724384

Overview

This guide explains how to install, build, configure, and use Plugable Chat, the open source desktop application designed to run locally on systems connected to the Plugable TBT5-AI GPU enclosure.

You should complete the hardware GPU installation procedure first before proceeding with this software guide.

Plugable Chat is currently distributed as source code via GitHub and must be built locally. There are no pre-built signed installer packages at this time.

If you have questions at any point, contact Plugable support → support@plugable.com

Suggested Materials

  • Host computer with Thunderbolt 5
  • Plugable TBT5-AI enclosure with GPU installed
  • Reliable internet connection
  • Administrator privileges on your system
  • At least 20GB free disk space for development tools and dependencies
     

Before You Begin

  1. Confirm the GPU is fully installed in the TBT5-AI enclosure.
  2. Connect the TBT5-AI to your host computer using a certified Thunderbolt 5 cable.
  3. Connect the enclosure to power.
  4. Power on the enclosure.
  5. Boot your computer.

Part 1 – Install GPU Drivers

Plugable Chat requires a properly installed GPU driver.

Windows (NVIDIA Example)

  1. Visit nvidia.com and download the latest driver for your GPU in the TBT5-AI.
  2. Run the installer.
  3. Choose Clean Installation.
  4. Restart your computer.

Part 2 – Download Plugable Chat Source Code

Plugable Chat repository:

https://github.com/PlugableTechnologies/plugable-chat

Step 1 – Install Git (if needed)

Windows:

winget install Git.Git

Step 2 – Clone the Repository

git clone https://github.com/PlugableTechnologies/plugable-chat.git
cd plugable-chat

Part 3 – Run the Platform Bootstrap Script

Plugable Chat provides platform bootstrap scripts that install all required dependencies automatically.

There is no separate installer path. The bootstrap script is the supported installation method.

Windows Installation

From inside the repository folder:

.\requirements.bat

This script will:

  • Validate Windows version
  • Check disk space
  • Verify network connectivity
  • Install Visual Studio Build Tools with C++ workload
  • Install Node.js
  • Install Rust
  • Install Git
  • Install Protocol Buffers
  • Initialize Rust toolchain
  • Run npm install

You may need to open a new command prompt and re-run .\requirements.bat several times to get all components installed.

Diagnostic mode (no installation):

.\requirements.bat --check

Once complete, start the app:

npx tauri dev

Part 4 – First Launch

When running:

npx tauri dev

The following occurs:

  • React frontend builds
  • Rust backend compiles
  • Tauri desktop application launches
  • Local database initializes
  • Model configuration loads

On first launch, you may be prompted to:

  • Select a model
  • Configure local model endpoints
  • Adjust settings
     

Part 5 – Using Plugable Chat

Starting a Chat Session

  1. Select your model from the model menu.
  2. Click New Chat.
  3. Enter your prompt.
  4. Press Enter or click Send.

Streaming responses will appear token-by-token.

Model Support

Plugable Chat supports multiple model families, including:

  • OpenAI-compatible endpoints
  • Gemma
  • Granite
  • Phi
  • Other transformer-based models

Model-specific tool-calling formats are handled automatically by the backend.

Built-In Capabilities

Plugable Chat includes:

  • Streaming response engine
  • Local vector database (LanceDB)
  • Python sandbox execution
  • Tool calling loop
  • MCP server integration
  • Local-first storage of chat history

Troubleshooting

GPU Not Detected

  • Confirm Thunderbolt connection.
  • Confirm GPU drivers are installed.
  • Check Device Manager or run nvidia-smi.
  • Power cycle enclosure and system.

Build Fails on Windows

Common causes:

  • C++ workload not installed in Visual Studio Build Tools.
  • winget not installed.
  • UAC prompt waiting behind other windows.

Run diagnostic mode:

.\requirements.bat --check


Rust Compilation Errors

Ensure Rust toolchain is initialized:

rustup update

Node Errors

Delete node_modules and reinstall:

rm -rf node_modules
npm install

Windows:

rmdir /s node_modules
npm install

Frequently Asked Questions (FAQ)

1. Is there a pre-built installer?

No. Plugable Chat must currently be built from source.

2. Do I need programming experience?

No coding is required for normal use. The bootstrap scripts automate setup.

3. Can I run this without a GPU?

Yes, but performance will be significantly reduced.

  • 12GB VRAM for smaller models (7B class)
  • 16–24GB for mid-sized models
  • 24GB+ for larger models

5. Does this require internet access?

Only for initial dependency installation and model downloads. The app runs locally.

6. Where is chat history stored?

Locally on disk using LanceDB within the project directory.

7. Is the application secure?

All processing occurs locally. The Python sandbox restricts execution to a curated allowlist.

8. Can I deploy this in production?

Yes, but production builds are not yet code-signed. Enterprise deployments may require internal signing.

9. How do I completely uninstall?

Delete the repository folder and uninstall:

  • Node.js
  • Rust
  • Visual Studio Build Tools (Windows)

10. Who do I contact for help?

Plugable Support → support@plugable.com

Include:

  • Operating system
  • GPU model
  • Error messages
  • Output logs

Conclusion

Once the hardware GPU installation is complete and Plugable Chat has been built and launched, your TBT5-AI enclosure becomes a high-performance local AI workstation capable of running advanced models, executing tools, and managing chat workflows entirely on your own system.

For assistance at any stage, contact support@plugable.com.