Overview

Sage is a demand-driven Rust semantic frontend. Symbols give definitions stable identities. Parsing and expansion create those symbols; resolution discovers them; signature and body queries compute semantic products when requested. Checking a body brings together name resolution, type inference, trait solving, and elaboration into a completed typed tree.

flowchart LR
    Source[Source] --> Symbols[Parse and expand symbols]
    Symbols --> Resolve[Resolve names]
    Resolve --> Signature[Check signatures]
    Signature --> Body[Check and elaborate a body]
    Metadata[Dependency metadata] --> Resolve
    Metadata --> Signature
    Infer[Inference and trait solving] --> Body
    Body --> Typed[Typed IR]

This is a map of demand boundaries, not a mandatory batch sequence. Start with Crates to follow the modeled entities and operations.

• Meta specifies how this book, its architecture language, source correspondence, and review system work.
• Tenets explains the principles behind the design.
• Anchors lists the specification's named claims with generated implementation and test status. Every rule has a page connecting its source definition, implementation excerpts, tests, and review.
• Crates is a guided entry into the compiler model. Its subchapters render .sage declarations and their Rust implementations.
• RFDs record changes, alternatives, rationale, and plans. Their normative changes are expressed directly in this specification.

Markdown and .sage are equally authoritative specification formats. Each anchor states one independently testable claim. A source badge opens the anchor's generated page; its context link returns to the definition. TODO and N/A describe explicit implementation or test entries, not successful verification. Evidence and status are generated from code, tests, and saved reviews rather than maintained as a second specification.

For tool usage, start with the User's Guide.