Skip to content

Classify knowledge base chunks

POST
/knowledge-bases/{knowledgeBaseId}/classify

Classify chunks using custom schemas beyond binary relevance. Supports tri-state classification (supporting/contradicting/neutral), built-in presets, and custom user-defined schemas.

knowledgeBaseId
required

ID of the knowledge base

string

ID of the knowledge base

object
chunks

Chunks to classify directly (alternative to query)

Array<object>
object
chunkId
required

Unique identifier for the chunk

string
content
required

Text content of the chunk

string
query

Query to fetch chunks (alternative to providing chunks)

string
queryOptions

Options for fetching chunks via query

object
limit

Maximum number of chunks to fetch via query

integer
default: 10
model

Embedding model to use for query (defaults to primary)

string
Allowed values: BAAI/bge-small-en-v1.5 BAAI/bge-base-en-v1.5 BAAI/bge-large-en-v1.5 sentence-transformers/all-MiniLM-L6-v2 sentence-transformers/all-mpnet-base-v2 nomic-ai/nomic-embed-text-v1.5
preset

Use a built-in classification preset

string
Allowed values: tri-state evidence-quality relevance-scored
labels

Simple classification labels (e.g., [‘supporting’, ‘contradicting’, ‘neutral’])

Array<string>
>= 2 items
labelDefinitions

Detailed label definitions with natural language guidance

Array<object>
>= 2 items
object
name
required
string
>= 1 characters
when
required
string
>= 1 characters
jsonSchema

Custom JSON Schema for advanced classification (power users)

object
key
additional properties
nullable
options

Classification behavior options

object
multiSelect
boolean
includeConfidence
boolean
includeReasoning
boolean
context

Additional context to help the LLM classify chunks accurately

string
batchSize

Number of chunks to classify per LLM call (default: 5)

integer
default: 5

Classification complete

object
results
required

Classification results for each chunk

Array<object>
object
chunkId
required
string
content
required
string
label
string
labels
Array<string>
classification
object
key
additional properties
nullable
confidence
number
<= 1
reasoning
string
fileId
required
string
nullable
datasetTableId
required
string
nullable
charStart
required
number
charEnd
required
number
metadata
required

Metadata about the classification operation

object
mode
required
string
Allowed values: preset labels labelDefinitions jsonSchema
preset
string
chunksClassified
required
number
batchesProcessed
required
number

Bad Request - Validation error or invalid input

object
error
required
string
code
string
details
nullable
retryable
boolean

Unauthorized - Authentication required or invalid token

object
error
required
string
code
string
details
nullable
retryable
boolean

Forbidden - Insufficient permissions

object
error
required
string
code
string
details
nullable
retryable
boolean

Not Found - Resource does not exist

object
error
required
string
code
string
details
nullable
retryable
boolean