KnowledgeBase - EventBridge - Alibaba Cloud Documentation Center

Parameter	Type	Description	Example
	object	The detailed information about an EventHouse knowledge base, including the status, metadata schema, chunking configuration, and search configuration.
Catalog	string	The EventHouse data catalog to which the knowledge base belongs. This value cannot be modified after the knowledge base is created.	my_catalog
ChunkConfiguration	object	The default chunking strategy of the knowledge base. This configuration takes effect only for documents uploaded after the configuration is updated. Existing documents are not re-chunked.
HeadingLevel	integer	The heading level (1 to 6) used for splitting in the BY_HEADING strategy. Headings at or above this level serve as split boundaries. Deeper-level headings are retained in the chunk body.	2
MaxChunkSize	integer	The maximum character length of a single chunk. Starting from revision 22, this value is character-based. Valid values: 1 to 6000.	600
OverlapSize	integer	The overlap character length between adjacent chunks. This parameter takes effect only for the BY_LENGTH strategy. When the value is greater than 0, the beginning of the next chunk repeats the content from the end of the previous chunk within this window. The overlap does not cause a chunk to exceed MaxChunkSize. A value of 0 indicates no overlap.	40
PreprocessRules	object	The preprocessing rules that take effect during document parsing.
RemoveUrlsAndEmails	boolean	Specifies whether to remove URLs and email addresses during parsing.	false
ReplaceConsecutiveWhitespace	boolean	Specifies whether to replace consecutive whitespace characters (spaces, line breaks, and tab characters) with a single space.	true
Separator	string	The separator used in the BY_SEPARATOR strategy. The separator is matched as a literal string (not a regular expression). The maximum length is 32 characters.	\\n\\n
Strategy	string	The chunking strategy. Valid values: AUTO: intelligent splitting (heading-aware + paragraph packing). BY_LENGTH: sliding window splitting by length. You can specify OverlapSize. BY_SEPARATOR: splitting by separator. You must specify Separator. BY_HEADING: splitting by heading level. You must specify HeadingLevel.	BY_SEPARATOR
CreatedAt	string	The time when the knowledge base was created.	2026-08-24T10:00:00Z
Description	string	The description of the knowledge base.	Product documentation knowledge base
EmbeddingDimension	integer	The embedding vector dimension specified during creation or the default dimension of the model. This value cannot be modified after the knowledge base is created.	1024
EmbeddingModel	string	The embedding model specified during creation. This value cannot be modified after the knowledge base is created.	text-embedding-v4
FailureReason	string	The brief reason for the most recent creation or deletion failure. This parameter is returned only when the status is CREATE_FAILED or DELETE_FAILED.	OssException: BucketAlreadyExists ...
KnowledgeBaseName	string	The name of the knowledge base, which is unique within the namespace.	product-docs
MetadataSchema	array	The metadata fields declared when the knowledge base was created. These fields cannot be modified after the knowledge base is created.	[{"Name":"department","Type":"STRING"}]
	MetadataSchemaField	The definition of a single field in the metadata schema, including the field name, type, and value source.
Namespace	string	The EventHouse namespace to which the knowledge base belongs. This value cannot be modified after the knowledge base is created.	my_namespace
SearchConfiguration	object	The default search configuration at the knowledge base level. This configuration takes effect when the corresponding parameters are not specified in a search request. You can modify this configuration by calling the UpdateKnowledgeBase operation.
Mode	string	The retrieval mode. Valid values: KEYWORD: keyword retrieval. VECTOR: vector retrieval. HYBRID: hybrid retrieval.	HYBRID
RankAlgorithm	string	The fusion algorithm for hybrid search. This parameter takes effect only in hybrid search mode. Valid values: RRF: reciprocal rank fusion. WEIGHTED: weighted normalization fusion. Use this value together with VectorWeight. Default value: RRF.	RRF
RerankEnabled	boolean	Specifies whether reranking is enabled by default. This parameter takes effect for all search modes (KEYWORD, VECTOR, and HYBRID). This default value is used when the Rerank parameter is not specified in a search request.	false
RerankModel	string	The default reranking model used when the RerankModel parameter is not specified in a search request. Valid values: qwen3-rerank, gte-rerank-v2, and qwen3-vl-rerank. Default value: qwen3-rerank. Score distributions vary across models and cannot be compared. Use the same model consistently within a knowledge base.	qwen3-rerank
RrfK	integer	The k parameter of the RRF fusion algorithm. The value must be greater than 0. Default value: 60.	60
TopK	integer	The maximum number of results returned by default for a search request.	20
VectorWeight	number	The weight of the vector path in the WEIGHTED fusion algorithm. Valid values: 0 to 1. The keyword path weight equals 1 minus this value. Default value: 0.7.	0.7
Status	string	The current status of the knowledge base. Valid values: CREATING: The knowledge base is being created. ACTIVE: The knowledge base is available. CREATE_FAILED: The knowledge base failed to be created. DELETING: The knowledge base is being deleted. DELETE_FAILED: The knowledge base failed to be deleted.	ACTIVE
UpdatedAt	string	The time when the knowledge base was last updated.	2026-08-24T10:00:00Z

object

The detailed information about an EventHouse knowledge base, including the status, metadata schema, chunking configuration, and search configuration.

Catalog

string

The EventHouse data catalog to which the knowledge base belongs. This value cannot be modified after the knowledge base is created.

my_catalog

ChunkConfiguration

object

The default chunking strategy of the knowledge base. This configuration takes effect only for documents uploaded after the configuration is updated. Existing documents are not re-chunked.

HeadingLevel

integer

The heading level (1 to 6) used for splitting in the BY_HEADING strategy. Headings at or above this level serve as split boundaries. Deeper-level headings are retained in the chunk body.

2

MaxChunkSize

integer

The maximum character length of a single chunk. Starting from revision 22, this value is character-based. Valid values: 1 to 6000.

600

OverlapSize

integer

The overlap character length between adjacent chunks. This parameter takes effect only for the BY_LENGTH strategy. When the value is greater than 0, the beginning of the next chunk repeats the content from the end of the previous chunk within this window. The overlap does not cause a chunk to exceed MaxChunkSize. A value of 0 indicates no overlap.

40

PreprocessRules

object

The preprocessing rules that take effect during document parsing.

RemoveUrlsAndEmails

boolean

Specifies whether to remove URLs and email addresses during parsing.

false

ReplaceConsecutiveWhitespace

boolean

Specifies whether to replace consecutive whitespace characters (spaces, line breaks, and tab characters) with a single space.

true

Separator

string

The separator used in the BY_SEPARATOR strategy. The separator is matched as a literal string (not a regular expression). The maximum length is 32 characters.

\\n\\n

Strategy

string

The chunking strategy. Valid values:

AUTO: intelligent splitting (heading-aware + paragraph packing).
BY_LENGTH: sliding window splitting by length. You can specify OverlapSize.
BY_SEPARATOR: splitting by separator. You must specify Separator.
BY_HEADING: splitting by heading level. You must specify HeadingLevel.

BY_SEPARATOR

CreatedAt

string

The time when the knowledge base was created.

2026-08-24T10:00:00Z

Description

string

The description of the knowledge base.

Product documentation knowledge base

EmbeddingDimension

integer

The embedding vector dimension specified during creation or the default dimension of the model. This value cannot be modified after the knowledge base is created.

1024

EmbeddingModel

string

The embedding model specified during creation. This value cannot be modified after the knowledge base is created.

text-embedding-v4

FailureReason

string

The brief reason for the most recent creation or deletion failure. This parameter is returned only when the status is CREATE_FAILED or DELETE_FAILED.

OssException: BucketAlreadyExists ...

KnowledgeBaseName

string

The name of the knowledge base, which is unique within the namespace.

product-docs

MetadataSchema

array

The metadata fields declared when the knowledge base was created. These fields cannot be modified after the knowledge base is created.

[{"Name":"department","Type":"STRING"}]

MetadataSchemaField

The definition of a single field in the metadata schema, including the field name, type, and value source.

Namespace

string

The EventHouse namespace to which the knowledge base belongs. This value cannot be modified after the knowledge base is created.

my_namespace

SearchConfiguration

object

The default search configuration at the knowledge base level. This configuration takes effect when the corresponding parameters are not specified in a search request. You can modify this configuration by calling the UpdateKnowledgeBase operation.

Mode

string

The retrieval mode. Valid values:

KEYWORD: keyword retrieval.
VECTOR: vector retrieval.
HYBRID: hybrid retrieval.

HYBRID

RankAlgorithm

string

The fusion algorithm for hybrid search. This parameter takes effect only in hybrid search mode. Valid values:

RRF: reciprocal rank fusion.
WEIGHTED: weighted normalization fusion. Use this value together with VectorWeight.

Default value: RRF.

RRF

RerankEnabled

boolean

Specifies whether reranking is enabled by default. This parameter takes effect for all search modes (KEYWORD, VECTOR, and HYBRID). This default value is used when the Rerank parameter is not specified in a search request.

false

RerankModel

string

The default reranking model used when the RerankModel parameter is not specified in a search request. Valid values: qwen3-rerank, gte-rerank-v2, and qwen3-vl-rerank. Default value: qwen3-rerank. Score distributions vary across models and cannot be compared. Use the same model consistently within a knowledge base.

qwen3-rerank

RrfK

integer

The k parameter of the RRF fusion algorithm. The value must be greater than 0. Default value: 60.

60

TopK

integer

The maximum number of results returned by default for a search request.

20

VectorWeight

number

The weight of the vector path in the WEIGHTED fusion algorithm. Valid values: 0 to 1. The keyword path weight equals 1 minus this value. Default value: 0.7.

0.7

Status

string

The current status of the knowledge base. Valid values:

CREATING: The knowledge base is being created.
ACTIVE: The knowledge base is available.
CREATE_FAILED: The knowledge base failed to be created.
DELETING: The knowledge base is being deleted.
DELETE_FAILED: The knowledge base failed to be deleted.

ACTIVE

UpdatedAt

string

The time when the knowledge base was last updated.

2026-08-24T10:00:00Z