🤖 AI Developer Tools
A modern developer workstation is incomplete without intelligent assistance. We have integrated Ollama, allowing you to run state-of-the-art AI models 100% locally on your hardware.
No cloud, no API keys, and no data leaves your machine.
The Models
Our setup defaults to providing three carefully curated models that target specific developer workflows:
1. qwen3:30b-a3b (The flagship reasoner)
- Architecture: Mixture of Experts (MoE)
- Parameters: 30 Billion total (only ~3 Billion active during inference).
- RAM Requirement: 16 GB minimum, 32 GB recommended.
- Disk Space: ~18 GB.
- Use Case: This is your flagship model. Because it uses MoE architecture, it produces reasoning quality comparable to a dense 30B model (like GPT-4 class logic), but generates tokens at the blistering speed of a 7B model. Perfect for architectural debates, strategy, and deep reasoning.
2. deepseek-r1:7b (Chain-of-thought engine)
- Architecture: Dense
- Parameters: 7 Billion
- RAM Requirement: 8 GB minimum
- Disk Space: ~5 GB
- Use Case: DeepSeek-R1 forces itself to output "chain-of-thought" logic before answering. Excellent for highly complex, multi-step math or logical puzzles where you need to see the AI's "work" before its conclusion.
3. qwen2.5-coder:7b (The syntax generator)
- Architecture: Dense
- Parameters: 7 Billion
- RAM Requirement: 8 GB minimum
- Disk Space: ~5 GB
- Use Case: Hardwired directly into the
ai-tunerscript. This model is exceptionally good at reading raw bash outputs (liketoporvmstat) and writing shell scripts, Python, or SQL. It is fast, lightweight, and focused purely on code.
How to Access
You have three primary methods to interact with these models:
1. The Nexus Dashboard
Press Super+X to launch Nexus. You'll see options like:
AI Chat — Reasoning (qwen3) AI Chat — Logic (deepseek)Choosing any of these instantly drops you into a floating terminal session interacting with the model.
2. The AI Auto-Tuner (ai-tuner)
Also accessible via Nexus, this tool runs background telemetry, feeds the raw data silently to qwen2.5-coder:7b via local cURL requests to the Ollama API port 11434, and presents optimization advice.
3. Terminal Interface
Just type normal Ollama commands anywhere:
ollama run qwen3:30b-a3b
ollama run qwen2.5-coder:7bEditor Integration (Antigravity)
Module 07 installs Antigravity, an AI-powered code editor (see antigravity.google for official docs). If Antigravity is not available on Arch Linux via AUR, the installer will attempt to extract it from the official Debian package or fall back to Cursor (cursor-bin from AUR) as an alternative AI-powered editor.
Both editors support AI-assisted coding features. For local model integration, start Ollama separately and configure the editor to connect to http://localhost:11434.