Type Alias CircularDependenciesOutput 

pub type CircularDependenciesOutput = CheckOutput;

Concrete circular-dependency output contract returned by typed runs.

Aliased Type

pub struct CircularDependenciesOutput {

    pub schema_version: SchemaVersion,
    pub version: ToolVersion,
    pub elapsed_ms: ElapsedMs,
    pub total_issues: usize,
    pub entry_points: Option<EntryPoints>,
    pub summary: CheckSummary,
    pub results: AnalysisResults,
    pub baseline_deltas: Option<BaselineDeltas>,
    pub baseline: Option<BaselineMatch>,
    pub baseline_staleness: Option<BaselineStaleness>,
    pub regression: Option<RegressionResult>,
    pub gate_outcomes: Option<GateOutcomes>,
    pub request_outcomes: Option<RequestOutcomes>,
    pub meta: Option<Meta>,
    pub workspace_diagnostics: Vec<WorkspaceDiagnostic>,
    pub next_steps: Vec<NextStep>,
}

Fields

schema_version: SchemaVersion

Dead-code output schema version; currently [CHECK_SCHEMA_VERSION].

version: ToolVersion

Fallow CLI version that produced this output.

elapsed_ms: ElapsedMs

Wall-clock analysis duration in milliseconds.

total_issues: usize

Total findings across all issue arrays; excludes next_steps.

entry_points: Option<EntryPoints>

Entry-point totals per source, when the analysis recorded them.

summary: CheckSummary

Per-category finding counts.

results: AnalysisResults

Full analysis results, flattened so each issue array sits at the envelope root.

baseline_deltas: Option<BaselineDeltas>

Count deltas against the matched baseline, in baseline runs.

baseline: Option<BaselineMatch>

Which baseline snapshot was matched, in baseline runs.

baseline_staleness: Option<BaselineStaleness>

This run’s view of the loaded baseline, present only in baseline runs. Carries the staleness counts, the advisory verdict and gate_trips, the same boolean --fail-on-stale-baseline exits on, so a CI integration reads one field instead of restating the rule. Read change_scoped before dividing matched_entries by baseline_entries: a narrowed run can report matched_entries: 0 on a healthy baseline.

regression: Option<RegressionResult>

Regression verdict against the baseline, in --fail-on-regression runs.

gate_outcomes: Option<GateOutcomes>

The verdict of every gate this run evaluated, keyed by name. The CLI always emits it, with the command’s default exit rule in it also when no flag armed a gate, so a CI integration reads the verdict instead of guessing from a process status it usually cannot see. A gate fails the build when status is fail AND enforced is true. The typed programmatic API runs no CLI gate and leaves it absent. See [crate::GateOutcomes].

request_outcomes: Option<RequestOutcomes>

Every narrowing or shaping request this run RECEIVED, keyed by name, absent when it was asked for nothing. An entry whose status is not applied means the run could not do what it was asked and reported something WIDER instead, so what follows is a valid report of a scope nobody requested. Honoured requests are published too, with status: "applied", so an absent object means “nothing was asked for”, never “nothing failed”. See [crate::RequestOutcomes].

meta: Option<Meta>

_meta block with docs and rule definitions, when --explain was passed.

workspace_diagnostics: Vec<WorkspaceDiagnostic>

Non-fatal diagnostics about the project itself, from all three stages that record them (issue #473):

• workspace discovery, at config load: undeclared-workspace, malformed-package-json, glob-matched-no-package-json, malformed-tsconfig, tsconfig-reference-dir-missing;
• source discovery, during the file walk: skipped-large-file, skipped-minified-file, skipped-source-dotdir, excluded-by-default-ignore, source-read-failure, source-parse-degraded;
• dead-code analysis, from the dependency-catalog and override detectors: malformed-pnpm-workspace-yaml, bun-lockb-override-resolution-skipped;
• framework plugins, while they read their own build configs: plugin-config-unreadable, plugin-effect-not-modeled.

Analysis-stage and plugin-stage kinds therefore reach only the envelopes whose run includes a dead-code analyze pass, never a standalone fallow dupes --format json. path is project-root-relative with forward slashes; the array is omitted when empty. The same list is repeated on each top-level command’s envelope so single-command consumers see it without having to look at a separate top-level field.

A diagnostic here is advisory and never withholds a finding. Where an entry reports a source file this run never fully analyzed (source-parse-degraded, source-read-failure, skipped-large-file, skipped-minified-file, skipped-source-dotdir) it can distort a verdict, so the affected unused_files[], unused_exports[], and dependency entries additionally carry the caveat themselves in their own optional reachability_caveats[] array, and a reader who never scrolls back up to this list still sees it. fallow fix reads the same array and withholds the removal while a caveat stands.

excluded-by-default-ignore is the one source-discovery kind that reports unseen files WITHOUT raising a caveat. It names a built-in ignore pattern (**/dist/**, **/build/**, **/coverage/**, or one of the four minified-bundle globs) that removed candidate source files from the walk, which is designed behavior on generated output rather than a degraded run, so it is advisory only and no finding inherits it. One entry per pattern, never per file, so the array stays bounded on a project of any size. Gitignored trees are pruned before the walk sees them and count zero, and **/node_modules/** is never reported: installed dependencies are not the first-party source the kind is about.

next_steps: Vec<NextStep>

Read-only follow-up commands computed from this run’s findings, emitted at the JSON root so an agent acting on the output is pointed at fallow’s adjacent verification capabilities (trace, complexity breakdown, audit, workspace scoping). Each command is runnable as-is and never mutating; see [NextStep] for both contracts. Omitted when empty or when FALLOW_SUGGESTIONS=off; does NOT contribute to total_issues.