output — CLI output behaviour

Schema: OutputSettingsConfig in src/thesis/core/config/validators.py. Configures the structured CLI output system (EventBus + OutputRenderer). Every value can be overridden at runtime with CLI flags.

Field	Type	Default	Allowed values	Description
verbosity	str	"normal"	quiet / normal / verbose	Default verbosity. normal shows errors, warnings, and key milestones; verbose adds full event/log detail; quiet shows only errors and the final result.
summary	str	"compact"	off / compact / full	End-of-run summary detail. compact is a headline + up to 6 bullets; full renders all bullets uncapped and always appends the per-result detail block (steps completed/failed, files changed, commands run, warnings) — without requiring -v; off suppresses the summary entirely.
progress	str	"auto"	auto / on / off	Progress bars/spinners. auto enables for TTY, disables for pipes/CI.
output_format	str	"human"	human / json	Output format. human produces styled text with [OK]/[FAIL] labels; json emits machine-readable summaries.

CLI overrides

CLI flag	Effect on this section
-v / --verbose	verbosity = "verbose" (also bumps log level to DEBUG).
-q / --quiet	verbosity = "quiet".
--summary off|compact|full	summary (case-insensitive).
--no-progress	progress = "off" (otherwise auto-detection from TTY).

-v and -q are mutually exclusive — passing both raises click.UsageError.

Example

output:
  verbosity: normal
  summary: compact
  progress: auto
  output_format: human

Auto-disabling progress

Progress UI is automatically disabled when:

• The CI environment variable is set.

• Output is piped (not a TTY).

• The NO_COLOR environment variable is set.

Notes

• Verbose mode also raises the loguru console level to DEBUG; quiet drops it to ERROR; normal sits at WARNING.

• The summary is rendered after the event stream is detached, so the structured event log and the summary never interleave.

• output_format: json produces a single JSON object on stdout containing the RunSummary / BatchSummary — useful for scripted pipelines that want to consume the result programmatically.