Inference
Local AI Brain
Your AI, on your hardware.
Run Ollama models locally when enabled. Local inference reduces external data flow and variable API cost, while keeping the same production surface and approval discipline.
Private early access for founders, operators, and executive teams.
$alabobai init --local
✓Scanning local models...
1.llama3.3:70b active· 40GB · Q4_K_M
2.mistral:7b-instruct · 4.1GB · Q4_0
3.codellama:13b · 7.4GB · Q5_K_M
4.nomic-embed-text · 274MB · embedding
✓4 models detected · Network: offline · Ready
Cloud calls
Optional
Variable cost
Lower
How it works
The surface changes by product, but the operating standard stays the same: clear steps, visible controls, and calmer execution.
01
Install Ollama
Install Ollama locally. Alabobai can detect a running Ollama instance and list available models.
02
Pick your model
Switch between models instantly. Use large models for quality, small ones for speed.
03
Work offline when possible
With local models installed, core chat workflows can run without external LLM API calls. Some workflows (web browsing, cloud models, collaboration) still require network access.
Key features
Every product page should use the same card rhythm, typography contract, and rose-gold operating language as the homepage.
Ollama integration
Native support for Ollama. Pull models with one command, then use them through the same production surface.
Auto-model detection
Alabobai can surface available local models with size and quantization info when Ollama is running.
No internet required
After initial model download, core inference can run locally. Fully air-gapped deployments depend on your infrastructure and workflow scope.
Lower variable cost
Local inference can reduce variable per-token spend for enabled workflows. Commercial access is still governed by your active plan.
Offline-capable core workflows
Core chat and local inference can run without cloud model calls. Web research and external integrations require network access.
Model switching
Switch models mid-conversation. Use llama for chat, codellama for code, mistral for analysis.
They charge per token. We give you the whole model.
The comparison layer should read clearly, not feel like a different design system.
| Feature | Alabobai | Others |
|---|---|---|
| Variable token spend | Reduced with local inference | Always per-token |
| Offline capability | Core workflows (local models) | Cloud required |
| Data flow control | Local + cloud options | Cloud default |
| Model choice | Available Ollama models | Vendor locked |
Own your AI. Run it on your terms.
Local inference reduces external calls and keeps the production loop reviewable. Part of Alabobai — the Digital Production System.