AI without limits.
On your hardware.
Myaiheye is a local-first AI system designed for chat, coding, reasoning, automation, and optional vision analysis — with full control over models and tools.
- Chat & Code helper
- Automation tools
- Optional image / vision analyzer
- Safe Mode by default, Root Mode opt-in
Who it’s for
Power users who want a local operator: developers, builders, IT / homelab folks, small teams, and creators who want an assistant that runs on their own machine.
Typical uses
• Coding + debugging with tools
• Local automation workflows
• UI / screenshot understanding (Vision tier)
• A private assistant that doesn’t depend on cloud services
Included models
Base distribution is built around a 3-model set:
| Role | Model | Size |
|---|---|---|
| Primary Chat / Code / Reason | qwen2.5-coder:32b | ~19 GB |
| Vision add-on | llava:34b | ~20 GB |
| Lite fallback | llama3.1:8b | ~5 GB |
Chat / Code / Reason are behavior modes, not separate models.
System requirements
Myaiheye runs entirely on customer hardware. Choose the tier that matches the experience you want.
Tier 1 — Lite (fallback-only)
Best for basic assistant, quick Q&A, low cost.
CPU: 4–8 modern cores
RAM: 16 GB min (32 GB recommended)
GPU: optional (8–12 GB VRAM helps)
Storage: 40 GB minimum (OS + model + logs)
Experience: decent on CPU, better on GPU
Tier 2 — Pro (recommended)
Best for “operator” workflows: tools, coding, automation.
CPU: 8–16 modern cores
RAM: 32 GB min (64 GB recommended)
GPU: 24 GB VRAM recommended (ideal for 32B-class quantized)
Storage: 120 GB min (250 GB recommended)
Experience: fast, stable, tool-friendly
Tier 3 — Pro + Vision
For screenshot/UI/photo understanding.
Same as Tier 2, plus:
GPU: 24 GB VRAM strongly recommended
Storage: add +25–50 GB headroom for vision cache/workflows
Universal requirements
OS: Ubuntu 24.04 LTS (or your supported distro list)
Ollama endpoint: 127.0.0.1:11434
Myaiheye UI: configurable (default e.g. 8080)
Security baseline: tool allowlist (or “confirm before sudo”), logs + audit trail
Startup & health validation
ollama list
curl -s http://127.0.0.1:11434/api/tags | head
pm2 status
nvidia-smi # if GPU tier
Performance expectations
First response after boot: < 5–10 seconds (warmup).
Typical responses: near-instant for chat/code tasks.
Heavy reasoning or vision tasks: longer but predictable.
HORUS — Local-First AI Operator
A fully local AI system designed for real execution, verification, and user-controlled behavior. No cloud dependency.
View HORUS →Safe Mode (default)
Myaiheye ships as a product: stable routing, tight tool permissions, restricted filesystem roots, and no risky system access by default.
Safe Mode defaults
• Tool execution via allowlist
• No sudo by default
• Restricted filesystem roots (project/data folders)
• Model routing locked to the supported set
Root Mode (opt-in)
For power users. Root Mode unlocks full customization: models, tools, routing rules, filesystem access, and (optionally) privileged commands — only when explicitly enabled.
Root Mode principles
• Human must enable it (never the AI)
• Enable locally (localhost/terminal), not from remote web sessions
• Audit log every elevated action
• One-click rollback to Safe Mode
Get started
Email [email protected] with subject Myaiheye and include:
• Your tier (Lite / Pro / Pro+Vision)
• Your hardware (CPU/RAM/GPU)
• What you want Myaiheye to do (chat, coding, automation, vision)
Notes
Myaiheye is designed for local operation and customer-owned hardware. For remote access, use secure networking practices and restrict access appropriately.
Product policy suggestion: large 70B+ reasoning models not included by default; offer optional packs later.