Requests

This section explains how to make one model request reliable.

1. Overview: the basic request chain.
2. Prompt Management: slots, scope, and YAML/JSON prompts.
3. Schema as Prompt: authoring structured output schemas.
4. Output Control: ensure, validate, retry, and events.
5. Model Response: reuse one response and read text/data/meta or streams.
6. Session Memory: Session, multi-turn history, windowing, import/export.
7. Context Engineering: where session, info, KB, and tool results belong.
8. Workspace: durable records, artifacts, search, links, and checkpoints for multi-turn tasks.