Batch document ingestion

POST

/data/v1/ingest/batch

• Production server

Ingest multiple documents in a single request. Each document is chunked, embedded, and stored in Milvus. Duplicates are skipped.

Idempotent: each document is independently checked for duplicates.

Authorizations

• BearerAuth

Request Body ^required

object

documents

required

Array of documents to ingest

Array<object>

object

content

required

The full document text to ingest

string

metadata

required

object

source

required

Source identifier (e.g. ‘manual’, ‘wikipedia’, ‘eve_sde’)

string

manual

source_id

Unique ID for deduplication (same source_id + source = duplicate)

string

title

required

Document title

string

url

Source URL (optional)

string

content_type

Content category (e.g. ‘article’, ‘game_item’, ‘docs’)

string

tags

Categorization tags

Array<string>

custom

Arbitrary custom metadata

object

key

additional properties

any

Responses

200

Successful response with per-document results

object

total_documents

Number of documents submitted

integer

total_chunks

Total chunks created across all documents

integer

successful

Number of documents successfully ingested

integer

failed

Number of documents that failed

integer

results

Per-document result details

Array<object>

object

document_id

ID of the first chunk (empty if no content)

string

chunks_created

Number of chunks created (0 if duplicate)

integer

status

Ingestion result status

string

Allowed values: success duplicate empty failed

message

Human-readable result message

string

400

Invalid request

object

error

object

message

string

type

string

code

string

param

string

401

Invalid or missing API key

object

error

object

message

string

type

string

code

string

param

string