4.3.11 2026-10-15

Extension and twinny-server release: completions work against llama-server in router mode.

• Fixed: completions failed against llama-server in router mode with 400 model name is missing from the request (#544). A llama.cpp completion request (and the warm-up before it) now names the provider's model, which the router uses to pick the model. A single-model server ignores it, and a provider with no model name set sends none, as before.
• The chat's context size is read from a router-mode llama-server too, by naming the model when asking for its properties. Asking does not load the model.