Skip to main content

Chat middleware

Request middleware transforms normalized chat parameters before dispatch. Response observers run after streaming finishes and receive an immutable summary.

How it works​

Middleware can use cluster embeddings through its extension context. Handle unavailable embedding capacity gracefully. Observers can log or index; they cannot rewrite an already-completed stream.

Complete contract​

Read the Skulk™ 2.0 development guide for the complete parameters, constraints, examples and operational behavior.