DetectorRegistry#

class scikitplot.cleanprompt.DetectorRegistry(detectors=None)[source]#

An ordered, immutable-by-convention collection of detectors.

Parameters:

detectorsiterable of Detector, optional: Initial detectors.

Raises:

PolicyError: If two detectors share a name.

Parameters:

detectors (Iterable[Detector] | None)

Notes

Developer notes. add returns self for chaining but mutates, so a registry is a builder, not a value. The engine never mutates one; it reads the registry once per pass. Names must be unique because arbitration uses the detector name as its final, deterministic tie-break — two detectors sharing a name would make the outcome depend on insertion order.

add(detector)[source]#

Append a detector.

Parameters:

detectorDetector: The detector to add.

Returns:

DetectorRegistry: This registry, for chaining.

Raises:

PolicyError: If a detector of the same name is already registered.

Parameters:

detector (Detector)

Return type:

DetectorRegistry

detect_all(text, policy, detectors=None)[source]#

Run detectors and collect every span.

Parameters:

textstr: The original text.
policyRedactionPolicy: The active policy; supplies Limits.max_spans and RedactionPolicy.min_confidence.
detectorssequence of Detector, optional: Override the registry’s own detectors.

Returns:

list of Span: Unresolved spans, which may overlap.

Raises:

DetectorError: If a detector raises. The detector name is carried on the error.
LimitExceededError: If more than Limits.max_spans spans are produced.

Parameters:

text (str)
policy (RedactionPolicy)
detectors (Sequence[Detector] | None)

Return type:

list[Span]

Notes

Developer notes. The span cap is checked while collecting, not afterwards, so a pathological detector cannot exhaust memory before the bound is noticed.

kinds()[source]#

Return every kind this registry can produce.

Returns: