Zero-Cloud Privacy
Modern productivity agents rely heavily on cloud APIs (OpenAI, Anthropic) to process context. This fundamentally compromises privacy when dealing with screen content, private messages, and personal work habits. AXIOM OS enforces a strict zero-cloud constraint.
- No external API keys required.
- All inference runs locally via Ollama.
- Telemetry never leaves the device loopback interface.
Local Model Selection
Running an LLM constantly in the background can monopolize system resources. We evaluated several models before settling on the 3-billion to 8-billion parameter class.
| Model | VRAM Reqm. | Reasoning Speed | Verdict |
|---|---|---|---|
| Llama-3-8b | ~6 GB | Medium | Good, but heavy on laptops. |
| Qwen-2.5-3b | ~2.5 GB | Very Fast | Selected for core logic. |
| Phi-3-Mini | ~2 GB | Fast | Context window issues. |
Determinism vs Hallucination
A critical design decision was separating the factual measurement of behavior from the AI's interpretation. The Go daemon writes strict facts to SQLite (e.g., "Chrome was open on youtube.com for 5 minutes"). The LLM is only given read-access to this data to form opinions. This guarantees the AI cannot hallucinate past user actions—it can only judge them.