CONTROL PLANE

Memory & settings

Review the durable context that shapes Skynet's decisions and control where operator and scheduled model work runs.

MEMORY · LOADINGSECRETS · ROOT MANAGED

OPERATOR MEMORY

What Skynet has been told

This combines explicit instructions, guest notes, comments, and resolved or dismissed issue context. It guides analysis but never authorizes execution by itself.

Loading durable memory…

MODEL ROUTING

Models & automatic fallback

OPENAI KEY · CHECKING

Choose a preferred route for operator and scheduled work. If a model fails, Skynet can try the other provider and then your remaining local endpoints in order.

OPERATOR ROUTELoading…

WEB

Provider Local endpoint Fast model Fast reasoning Deep model Deep reasoning

SCHEDULED ROUTELoading…

NIGHTLY / CLI

Provider Local endpoint Fast model Deep model Deep reasoning

RESILIENCEKeep working when a model fails

Automatically try other configured models OpenAI output-token limit

Preferred model → other provider → remaining local endpoints, in the order below. Handles output limits, invalid JSON, quota errors, outages, and timeouts. Each endpoint/model is tried once. Cancellation stops the chain. OpenAI uses the installed API key.

Local endpoints & models

Add LM Studio or another OpenAI-compatible server. URLs must end in /v1/.

Output limits include the model's reasoning where the server counts it. Set 0 to omit Skynet's output limit and use the server default; the server context window still applies. Endpoint URLs are reached from the Skynet controller.

Loading durable settings…