Running Your First AI Model
Load and run AI models on your Thox.ai device for intelligent code completion.
Your Thox.ai device comes with thox-coder pre-installed and ready to use. This guide shows you how to verify it's working, download additional models, and start using AI-powered code completion.
The Pre-installed Model
Every Thox.ai device ships with thox-coder ready to go. This is our recommended model for most use cases - it provides an excellent balance of speed and quality for code completion tasks.
To verify it's loaded and ready:
You should see "thox-coder" listed as the active model with status "loaded".
Available Models
thox-coder
RecommendedDefault model. Balanced speed and quality for code completion.
thox-coder-fast
Faster responses, slightly lower quality. Good for quick completions.
codestral
Mistral's code model. Excellent for complex generation and review.
llama3-8b
General-purpose. Good for documentation and explanations.
View all available models: thox models list --available
Downloading Additional Models
Download a Model
Download time depends on model size and connection speed. A 7B model takes ~10-20 minutes on typical broadband.
Switch Active Model
The new model loads into memory and becomes active for all requests.
Remove a Model
Free up disk space by removing models you no longer need.
Testing Your Model
Quick Test via CLI
The model will complete the function. First request may take a few seconds as the model loads into memory.
Web Interface Chat
Navigate to http://thox.local:8080/chat for an interactive chat interface. Try asking it to explain code or generate functions.
API Test
Connect Your IDE
For the best experience, connect your development environment to Thox.ai for real-time code completions as you type.
Tips for Best Results
Start with thox-coder
It's optimized for the device and provides excellent results.
Give it context
More code context leads to better completions. Open related files in your IDE.
Use descriptive names
Good function and variable names help the model understand intent.
Iterate and refine
Accept partial completions and continue typing for better results.