Explanation¶

Explanation pages connect the SDK's abstractions to the constraints they solve. Use them when you need a mental model before choosing an extension point or debugging an interaction.

Documents¶

Document	Summary
Client layers and ownership	Explain why the SDK separates transport, sessions, manual operations, and simple polling clients.
Typed request and response design	Explain how Pydantic models bind Python types, wire aliases, endpoint metadata, and compatibility behavior.
Generation parameters and templates	Explain why reusable generation intent is separated from API payload and backend representations.
Worker jobs, generations, and state	Explain how worker jobs coordinate popped requests while generations own individual backend results.
Package architecture	Explain package boundaries from generic API contracts through AI Horde models, workers, and backend adapters.
Backend normalization	Explain why backend-native names and shapes are converted at dispatch and parsing boundaries.
Dispatch normalization	Explain how feature requirements, worker advertisements, and parameter conversion select executable work.

Planned¶

Not yet documented in full. Each page describes its subsystem briefly and names the code that holds the behavior.

Document	Scope
Operation chaining	Explain why worker operations are represented as dependency graphs with explicit shared context.
Ratings integration	Explain the separate ratings-service boundary and how its client relates to generation results.
Model-reference integration design	Explain how static model metadata complements live worker availability without replacing it.