FAQ and Troubleshooting

FAQ and Troubleshooting

Common questions, each pointing to the page with the full answer.

Do I have to rewrite my app to a new SDK?

No. Applications speak the OpenAI, Anthropic, or Gemini SDK shape they already use; you change only the base URL and the credential. The gateway translates your request to whatever provider and model routing selects, and translates the response back into your shape. See API and Client Integration.

Which providers and models can I use?

Twenty in-tree adapters ship today — eleven with full bidirectional translation (OpenAI, Anthropic, Gemini, Vertex, Azure, Bedrock, Cohere, MiniMax, GLM, Replicate, Voyage) and nine OpenAI-compatible passthrough providers (DeepSeek, Moonshot, Mistral, Groq, Fireworks, Together, Perplexity, xAI, Hugging Face). See Core Concepts and Connecting a Provider.

What is a virtual key and how do I get one?

A virtual key is the SDK-layer credential that authenticates a caller to the AI Gateway and scopes which models it may use. Create one in the console under Virtual Keys (or your Personal Virtual Keys); the secret is shown once at creation, so copy it then. Gateway virtual keys are prefixed nvk_. See Connecting a Provider.

Why was my request blocked?

The compliance pipeline ran a hook — for example PII detection, content safety, or a keyword filter — and blocked the request at the request stage. The decision is recorded on the request's audit row, so you can see which hook fired. See Using the Control Plane.

How is traffic intercepted — must I integrate with the gateway directly?

Not necessarily. Traffic comes under governance through three independent paths: the AI Gateway for apps that call it with a virtual key, the Compliance Proxy for traffic captured at the network layer (including browsers and desktop AI tools), and the Desktop Agent for traffic intercepted on the endpoint. You can use any combination. See Architecture Overview.

Can I run it on-premises or without internet access?

Yes. Nexus runs self-hosted on your own infrastructure, and in an air-gapped model with no outbound internet where updates and credentials are brought in out of band. See Installation and Deployment.

Where do I see what happened to a request?

Every request handled by any intercept path is recorded as one traffic-event row, visible in the Control Plane's traffic view with its routing, tokens, latency, and any compliance decisions. See Using the Control Plane.

How do I run Nexus locally?

Bootstrap with ./scripts/dev-start.sh, run each service, and send a first request. See Getting Started.

See also

• Getting Started — local bring-up and first request.
• Core Concepts — the vocabulary behind these answers.
• Home — what Nexus is.