For the complete documentation index, see llms.txt. This page is also available as Markdown.

v1/classify-verbose

Classifies text and returns detailed match-level diagnostics for each candidate, including accepted and rejected matches with scores, regex patterns, and validation results. Use this endpoint when debugging classification results or tuning filters. To classify without diagnostics, use v1/classify.

Body options

Field
Type
Description

content

string

Text to classify. Required.

filters

string[]

Allow-list of element labels to report. Omit to report every supported element.

excludeFilters

string[]

Deny-list of element labels to skip. Takes precedence over filters.

sensitivityLevels

string[]

Restrict results to these tiers: low, medium, high, critical. Elements with no assigned tier are always included.

focusEntity

string

Restrict verbose diagnostics to matches for this element label. Other elements are still classified and appear in classifications.

focusStart

integer

Byte offset into content to begin verbose tracking. Defaults to 0.

focusEnd

integer

Byte offset into content to end verbose tracking. Defaults to end of content.

useML and treatAsJSON are not supported on this endpoint.


Response fields

Field
Type
Description

content

string

The original content submitted.

classifications

Classification[]

Accepted matches, same shape as v1/classify response.

accepted_matches

VerboseMatch[]

All candidates that passed scoring, validation, and overlap resolution.

rejected_matches

VerboseMatch[]

Candidates that were considered but did not make the final result.

ml_classifier_used

boolean

Whether the ML classifier ran.

scanner_version

string

Version of the scanner that produced the result.

focus_entity

string

The focusEntity value echoed back, if provided.

focus_offset_start

integer

The focusStart value echoed back, if provided.

focus_offset_end

integer

The focusEnd value echoed back, if provided.

VerboseMatch fields

Each object in accepted_matches and rejected_matches describes one candidate.

Field
Type
Description

entity

string

Element label for this candidate.

start_byte

integer

Byte offset of the match within content.

end_byte

integer

Byte offset of the end of the match within content.

text

string

Matched text, if available.

score

float

Confidence score.

is_accepted

boolean

true if this candidate is also in classifications.

regex_used

string

The regex pattern that matched, if any.

regex_id

integer

ID of the matched regex rule.

regex_quality

string

Quality tier of the matched regex.

context_matches

string[]

Context keywords that boosted or triggered the match.

context_score

float

Combined context score.

validation_tests

string[]

Names of validation checks applied.

validation_passed

boolean

Whether all validation checks passed.

exclusion_reason

string

Reason this candidate was excluded, if rejected.

exclusion_details

string

Additional detail about the exclusion.

context_exclusion_id

integer

ID of the context exclusion rule that dropped this match, if any.

regex_exclusion_id

integer

ID of the regex exclusion rule that dropped this match, if any.

is_ml_classified

boolean

Whether the ML classifier contributed to this candidate.

ml_confidence

float

ML model confidence score, if applicable.

overlapped_by

object

Present when this candidate was dropped because a longer overlapping match won. Contains entity, start_byte, and end_byte of the winning match.

processing_stage

string

Pipeline stage at which this candidate was dropped, if rejected.

Fields marked as optional are omitted from the response when empty or zero.


Last updated

Was this helpful?