U
unopenAI
Operating Layer • No Mandatory API • Open Architecture

unopenAI
AI without mandatory AI APIs.

Use open models. Run AI locally. Bring your own brain. Choose your infrastructure. The model is the brain. unopenAI is the operating layer.

START CHATRUN LOCALLYEXPLORE MODELS
Open-weight Ollama BYO GPU Privacy-first
Architecture • V1 Operating Layer
UINext.js • Vercel
RouterAuto Model Router • Task Classifier
ProviderOllama • OpenAI-Compatible • Mock • Future
ModelQwen / Llama / DeepSeek / Custom
ResponseStreaming • Local-first
Two Primary Modes
Web AI
Configured cloud endpoint. Bring any OpenAI-compatible API.
→ /api/chat → Router
Local AI
Your computer. Your model. Your data. No mandatory provider.
→ Ollama → Qwen / Llama
LOCAL MODE: Your messages are processed by your local model. No mandatory cloud AI provider.
Section 01 — The Problem

Modern AI hides the infrastructure behind an API. unopenAI puts the model choice back in your hands.

Most AI products lock you to a single provider. Your prompts, your data, and your workflow become coupled to someone else's endpoint.

unopenAI is an operating layer, not a wrapper. Providers are replaceable. Models are discoverable. Inference can be local.

Provider abstraction • No vendor lock-in

Your brain, your choice

Open models • Local intelligence
Qwen
Alibaba • Reasoning & coding
Local-ready
Llama
Meta • General & open
Local-ready
DeepSeek
DeepSeek • Reasoning
Cloud/Local
Local Models
llama.cpp • Ollama • BYO
Your hardware
Custom
OpenAI-compatible endpoint
BYO API
Section 03 — Local AI

Run AI on your own machine

01Your PC↓ Ollama ↓ Open Model ↓ unopenAI
No mandatory proprietary AI API. Your computer. Your model. Your data.
Setup GuideTry Chat
Terminal • Local Setup
# 1. Install Ollama
brew install ollama  # macOS
# or https://ollama.com

# 2. Download an open model
ollama pull qwen3
# or llama3.1, deepseek-r1, etc.

# 3. Start Ollama (runs on :11434)
ollama serve

# 4. Connect unopenAI
# set OLLAMA_BASE_URL=http://localhost:11434

# 5. Chat at /chat → Model: Auto → Local
Architecture Principle

No mandatory provider

unopenAI is designed so your AI provider can be replaced. Frontend never directly depends on a specific provider.

Roadmap
V1Web • Ollama • Router • Mock
V2Auth • Agents • Tools • Files
V3BYO GPU • Desktop • Benchmarks
V4Distributed • Compute marketplace
Future • Coming
Local AI → BYO GPU → AI Agents → Distributed Compute → unopenAI Network
Privacy-first • Model-agnostic