One interface.
Every AI.
AIR routes requests across Gemini, Groq, xAI, and OpenRouter. When a provider fails, AIR switches to another one and keeps the conversation going.
v0.1.0-beta.1, Python
Routed to Gemini
Four providers, configured once.
Each provider is set up independently and can be prioritized, disabled, or left as a fallback.
AIR picks a provider. If it fails, AIR moves on.
AIR matches each request to a configured provider based on availability, priority, and health. If a provider fails (rate limits, quota, an outage), AIR switches to the next option on its own.
The application sends one request. It doesn't catch provider-specific errors or track which provider is healthy. AIR does that.
STEP 1
Gemini
Handling requests
STEP 2
Quota exhausted
AIR detects the failure
STEP 3
Groq
AIR reroutes automatically
STEP 4
Conversation continues
No interruption to the application
The conversation doesn't reset when the provider does.
Context lives in AIR, not in any single provider. When AIR switches providers mid-conversation, the new provider gets the full history. The application and the user see no gap.
Understood. The secret word is NixGOLD.
The secret word is NixGOLD.
Full history stored. Reduced history sent.
AIR keeps the full, original conversation in storage and trims what actually gets sent to a provider on each request. Nothing is discarded. The stored session stays the source of truth.
STORED SESSION
SENT TO PROVIDER
Every routing decision is observable.
As AIR selects, fails over, or exhausts a provider, it emits structured, machine-readable routing information. Applications can react to these events programmatically instead of inferring what happened from the response alone.
Use AIR in bots, IDEs, backends, and AI apps.
These are illustrative use cases, not existing products. AIR needs no GUI, terminal, or audio device to run.
AI IDE
Route model requests without depending on one provider.
Multi-model platform
Offer several providers behind one integration.
AI assistant
Keep a conversation alive through a provider outage.
Backend service
Add provider redundancy without app-level retry code.
Discord bot
Handle usage bursts without hitting one quota.
Automation system
React to routing events as part of a pipeline.
Developed by
NISPAX InfoTech
AIR is built and maintained by NISPAX InfoTech as a routing layer for applications working across multiple AI providers.