Frequently asked questions

A small ecosystem for running your own local language models: a proxy that sits in front of llama-server and speaks the Ollama and OpenAI-compatible APIs, a browser-based WebUI chat client, and a VS Code extension - all sharing the same account and models.

No. Inference happens entirely between your client (WebUI, VS Code, or any OpenAI-compatible tool) and your own LlamaNexus proxy, running on hardware you control. This site's account/billing system never sees your prompts, model responses, or files - see our Privacy Policy for details.

The proxy (this repository's llamanexus binary) is the only piece that actually talks to llama-server and your models - it's required. WebUI and the VS Code extension are two different chat clients that connect to a proxy over its API; you can use either, both, or neither (any OpenAI-compatible client works too).

Depends on the mode. In normal (single-node) router mode, llama-server can keep several GGUF models loaded at once and switch between them per-request with no manual step. In RPC / distributed-inference mode (connecting worker machines to pool memory and compute for one large model), only one model at a time can be loaded across the whole cluster - RPC workers hold that one model's split layers. You can still switch models, just not two loaded simultaneously: unload the current one and select a different one, same as loading any model normally, and keep chatting - nothing needs restarting. See Distributed inference for the full picture.

Every account starts on a 14-day Trial. Paid plans are Basic, Pro, and Enterprise, billed monthly or annually, each with its own proxy/WebUI/VS Code install limits - see your account page for exact numbers and pricing.

It's how a proxy, WebUI, or VS Code install registers itself against your plan - and it's required to send chat requests at all. Leave it blank and chatting is blocked with a reminder to add one, but that makes zero calls to the account API until you do - the block itself is decided entirely on your own machine. Generate a key from your account page to get started (your 14-day Trial works the same as any paid plan here - no payment needed until it ends).

No. Worker nodes (llamanexus worker) have no connection to the license system at all, no matter how many you run - only the serve process counts as a proxy install.

New chat requests are blocked until you upgrade - everything else (settings, past sessions, the app itself) keeps working normally. Upgrading from your account page lifts the block immediately.

From your account page - "Manage billing" opens Stripe's Billing Portal, where you can update your payment method, switch plans, or cancel.

Check the docs for setup and configuration, or see our Usage Policy for how to report a security issue.