unopenAI

What is unopenAI? unopenAI is a privacy-first, model-agnostic AI operating layer by CYBER STARLINK. It is not another ChatGPT clone. The model is the brain. unopenAI is the operating layer.

Why does it exist? Modern AI often hides infrastructure behind a single API. unopenAI puts model choice back in the user's hands. Providers are replaceable. Inference can be local. Your computer. Your model. Your data.

UI (Next.js on Vercel)
 ↓
Router (Task Classifier → Model Selection)
 ↓
Provider Abstraction (AIProvider interface)
 ↓
Ollama Local | OpenAI-Compatible | Mock | Future
 ↓
Qwen / Llama / DeepSeek / Custom
 ↓
Streaming Response

Supported Providers: OllamaProvider (default http://localhost:11434, configurable via OLLAMA_BASE_URL), OpenAICompatibleProvider (any OpenAI-compatible API: vLLM, LM Studio, LocalAI, OpenRouter, Together, etc.), MockProvider (demo mode, clearly labeled, allows Vercel deploy without API keys).

Security: Input validation (zod), request size limits, rate limiting hooks, CORS policy, secure headers (X-Content-Type-Options, X-Frame-Options, Referrer-Policy, Permissions-Policy), no secret exposure, server-side env vars, sanitized markdown rendering, no API keys in client bundles.

Roadmap

• V1: Web interface, provider abstraction, Ollama, OpenAI-compatible, model discovery, local-first memory, Vercel deployment
• V2: Authentication, persistent cloud memory, AI agents, web tools, file tools, code execution
• V3: Bring Your Own GPU, Desktop application, Advanced model routing, Model benchmarking
• V4: Distributed inference, Compute marketplace, unopenAI Network