Module config

Structs§

AiConfig
AssertionFreeTestOptions: [rules.options.no-assertion-free-test].
ClassifierConfig: The optional System One classifier (Jev / TypeSafe), on a separate axis from the generative LLM. Off by default: opt-in strict, since Jev is a proprietary SaaS and slopguard’s deterministic AST mode stays the offline socle.
Config
EscalationConfig: Severity escalation: when one rule fires repeatedly in a single file, its warnings become errors. Opt-in so existing CI results do not change.
FixConfig: scan --fix policy. Builtin rules marked autofix_safe are always rewritten; a rule from rules.custom_dirs or [[rules.sources]] is rewritten only when its id is listed in allow_external, so a third-party ruleset cannot rewrite source files on its own say-so. User-only: an untrusted repo slopguard.toml cannot set it.
GitConfig: [git]: how external git rule sources are fetched. User config only.
LoadedConfig: A resolved config plus the warnings raised while loading it (reserved keys dropped from an untrusted repo file).
OutputConfig
RuleOptions: The options of the builtin rules that take some, keyed by rule id.
RuleSource: One external rule source declared as [[rules.sources]].
RulesConfig
RulesetsConfig
ScanConfig

AiTransport: Transport used to reach the LLM.
AiVendor: Vendor used when provider = "api".
ClassifierTransport: How the System One classifier (Jev) reaches the TypeSafe API.
ConfigError
OutputFormat
ProjectTrust: Whether the scanned repository’s slopguard.toml is trusted.

MAX_AI_CONCURRENCY: Upper bound of [ai].concurrency. Keeps the AI pass far below the semaphore permit limit and the provider’s rate limits.

load_config: Load configuration with hierarchical resolution.
load_config_file: Load configuration from a single file, skipping hierarchical resolution.
load_config_from: Load config with explicit paths (for testing).