Aether-Key

The "Connect Wallet" of the AI era—a universal, zero-knowledge gateway that allows end-users to securely bring their own API keys or local Ollama models to any AI app.

Description

Aether-Key solves the biggest scaling problem for indie AI developers: API costs. Agentic AI and RAG pipelines burn through tokens rapidly, making it financially unviable to scale free apps.

Aether-Key acts as the "Connect Wallet" for intelligence. It is a drop-in React component and stateless FastAPI proxy that allows users to instantly connect their own cloud API keys (OpenAI, Anthropic, Gemini, Groq, etc.) or local AI models (via Ollama). Instead of developers paying for inference, users pay with their own keys.

Core Features include:

  • Absolute Zero-Knowledge: Keys are never stored on a server. They travel securely in request headers for a single inference call and are immediately wiped from memory. A custom middleware redacts all sensitive headers from logs.

  • Client-Side Encryption: If users choose to "Remember" their keys across sessions, Aether-Key uses the native browser Web Crypto API to AES-256 encrypt the key with a user-defined passphrase before saving it to localStorage.

  • First-Class Local Models: The React component features automatic Ollama discovery. It pings localhost:11434 directly from the browser to surface local models in the UI. When selected, requests route directly to the local model, bypassing the proxy for zero latency.

  • Universal Routing: The stateless FastAPI proxy uses LiteLLM to translate standard OpenAI-format requests into over 100+ different provider formats automatically, supporting full Server-Sent Events (SSE) streaming.

Aether-Key is designed as infrastructure, not a product. It allows developers to write code once and let their users choose how their intelligence is powered.

Issues & PRs Board
No issues or pull requests added.