Skip to content

Concepts

Concept pages map the framework's mental model to the shipped code. Each page matches what's actually in the crates rather than the aspirational framework spec — if a concept page and the code disagree, the code is authoritative.

v0.1 shipped

  • App context — the App struct, Arc-shared services, cancellation flow, why there's no App<C> yet.
  • Configuration — typed layered config via figment, precedence rules, atomic reload.
  • Error diagnostics — thiserror + miette, tool-specific footer, the edge-rendering pipeline.
  • MCP exposure — how Command::mcp_exposed surfaces a CLI command as an MCP tool.

Planned (per roadmap)

  • assets-overlay.md — with the v0.2 rtb-docs + hot-reload work.
  • command-authoring.md — once the #[rtb::command] macro lands.
  • telemetry-events.md — with the OTLP sink in v0.2.
  • credentials-resolution.md — alongside rtb-cli credential subcommands.
  • ai-providers.md — covering rtb-ai's multi-provider layout alongside the streaming and structured-output paths.

Until a concept page exists, the per-crate spec is the authoritative source.

Authoritative docs

For engineering requirements that span crates (security rules, testing discipline, documentation expectations), see docs/development/engineering-standards.md.