surface↓classification↓bucket
Classificationdimension · returns varchar

BUCKET

Classify text into user-specified buckets via zero-shot NLI

Per-row classifier — stable across GROUP BY.

classificationnlispecialist-zootext

Arguments

nametypedescription
textVARCHAR—
bucketsVARCHARComma-separated list of bucket names
strictBOOLEAN—

About

General-purpose closed-label classifier — buckets text into labels the user provides as a comma-separated string. The most flexible dimension operator: you bring the label set, it buckets the rows. Backend: specialist zoo zero-shot NLI via /classify_batch. User- provided labels are passed straight through to the cross-encoder hypothesis templating. When `strict=false` (default), an "Other" bucket is added for rows that don't fit any user label well — the cell adds it by detecting when the top entailment score is low across all provided labels. For LLM-style bucketing with nuanced label semantics, use BUCKET_LLM — see bucket_dimension_llm.cascade.yaml.

Examples

Value bucketed

SELECT
  bucket (75, 'numeric,text,date')

Nearby rabbit holes

same domain
dimension

AUDIENCE

Identify target audience via zero-shot NLI

classificationnlispecialist-zootext+1
dimension

AUDIENCE_LLM

LLM-backed audience identification (escape hatch for AUDIENCE)

classificationllmllm-escape-hatchtext
dimension

AUTHENTICITY

Assess authenticity of content via zero-shot NLI

classificationnlispecialist-zootext
dimension

AUTHENTICITY_LLM

LLM-backed authenticity assessment (escape hatch for AUTHENTICITY)

classificationllmllm-escape-hatchtext
dimension

BUCKET_LLM

LLM-backed bucketing (escape hatch for BUCKET)

classificationllmllm-escape-hatchtext
scalar

CASE

Multi-branch semantic classification via specialist zoo NLI

classificationnlispecialist-zootext
Climb back to The Looking Glass