How-to guides¶
Task-focused recipes. Each guide assumes you have a tool built on the framework (see App context for the entry-point pattern) and walks one job end to end.
If you have never used RTB, start with Build your first tool instead — a how-to assumes you know what you are trying to do. To look a flag or a key up rather than follow a recipe, go to Reference.
Commands & errors¶
- Add a custom command — register a subcommand
via the
Commandtrait andBUILTIN_COMMANDS. - Produce user-facing errors — diagnostics
with
mietteinstead of bareResult.
Configuration & secrets¶
- Configure credentials — env-var, OS keychain, or literal, and the resolution precedence.
- Configure an AI provider — wire
rtb-chatto Claude, OpenAI, Gemini, or Ollama.
Observability¶
- Structured JSON logging — JSON logs
- telemetry events with automatic secret redaction.
Releases¶
- Secure releases & self-update — signed,
checksum-verified release artefacts and the
updateflow.
Scaffolding¶
The scaffolder shipped in 0.6 and has its own reference rather than a how-to guide:
rtb generate— scaffold a project, command, flag, setting or docs page.rtb remove— reverse any of those.rtb regenerate— reconcile a tree with its manifest.
Guides that are missing
There is no how-to yet for migrating an existing hand-written CLI onto the framework, and none for the AI codegen path — the latter needs a live provider key to follow, so it cannot be verified the way the others can.