Classification - Vector Retrieval Service for Milvus - Alibaba Cloud Documentation Center

AI_CLASSIFY selects the best matching label for text, images, or videos from a set of candidate labels. Use it for ticket routing, content moderation, or product category tagging. When configured as a Collection Function, AI_CLASSIFY automatically populates the classification field during data writes.

Command format

Call the REST API to classify content directly, or configure a Collection Function to automatically classify data during writes. The REST API supports two endpoints:

/v2/vectordb/ai/text_transform — Include "task": "ai_classify" in the request body.
/v2/vectordb/ai/classify — The task field is not required.

REST API

Send the following request to classify input content:

POST /v2/vectordb/ai/text_transform
Content-Type: application/json

{
  "model_name": "<model name>",
  "task": "ai_classify",
  "texts": ["<text or media URL>"],
  "params": {"labels": ["<label1>", "<label2>"]}
}

Python

Define a Collection Function to classify content automatically during data writes:

schema = MilvusClient.create_schema(auto_id=True, enable_dynamic_field=False)
schema.add_field("id", DataType.INT64, is_primary=True)
schema.add_field("content", DataType.VARCHAR, max_length=4096)
schema.add_field("category", DataType.VARCHAR, max_length=64)
schema.add_field("dummy_vector", DataType.FLOAT_VECTOR, dim=2)
schema.add_function(
    Function(
        name="classify_content",
        function_type=texttransform_function_type(),
        input_field_names=["content"],
        output_field_names=["category"],
        params={
            "provider": "aliyun_milvus",
            "model_name": "<model name>",
            "task": "ai_classify",
            "labels": "accounts,inquiries,faults,billing",
            "prompt": "Classify by the topic of the customer's question.",
            "temperature": "0",
            "max_concurrency": "2",
        },
    )
)

Parameters

Parameter	Description
`model_name`	Required. The model name. For images and videos, select a configured multimodal model (for example, `qwen3.7-plus`).
`texts`	Required for REST API. The content to classify, or an image or video URL.
`labels`	Required. Candidate labels, supplied as a JSON string array or a comma-separated string. 2 to 20 labels are supported.
`prompt`	Optional. Supplementary routing rules, up to 5,000 characters. `${...}` is not supported.
`media_type`	Optional. Set to `image` or `video` to indicate that the input is a media URL.
`temperature` / `max_concurrency` / `timeout_sec`	Optional. Controls model stability, concurrency, and timeout. Video classification is time-consuming. Set `timeout_sec` to 300 or higher to avoid timeouts.
`provider` / `task`	Required only for Collection Function. Fixed values: `aliyun_milvus` and `ai_classify`.

Return values

data.output.outputs returns label text in the same order as the input, containing only the label itself. The schema writes the label to the target text field.

Example 1: Auto-routing customer tickets (text)

A customer service system needs to route fault and inquiry tickets to different queues. When calling the unified endpoint /v2/vectordb/ai/text_transform, add task at the top level of the request.

REST API

#!/usr/bin/env bash
set -euo pipefail

MILVUS_REST_BASE_URL="http://c-xxxx.milvus.aliyuncs.com:19530"
MILVUS_AUTH_TOKEN="<yourUsername>:<yourPassword>"

post_json() {
  local path="$1"
  local body="$2"
  curl -X POST \
    "$MILVUS_REST_BASE_URL$path" \
    -H "Authorization: Bearer $MILVUS_AUTH_TOKEN" \
    -H "Content-Type: application/json" \
    -d "$body"
}

BODY=$(cat <<'JSON'
{
  "model_name": "qwen3.7-max",
  "task": "ai_classify",
  "texts": [
    "Cannot log in to the Milvus console; verification code error keeps appearing.",
    "How do I enable public network access for a Serverless Milvus instance?"
  ],
  "params": {
    "labels": ["accounts", "inquiries", "faults", "billing"],
    "prompt": "Classify by the topic of the customer's question.",
    "temperature": 0,
    "max_concurrency": 2
  }
}
JSON
)
RESPONSE_BODY="$(post_json "/v2/vectordb/ai/text_transform" "$BODY")"
if command -v jq >/dev/null 2>&1; then
  echo "$RESPONSE_BODY" | jq .
  [ "$(echo "$RESPONSE_BODY" | jq -r '.code // -1')" = "0" ] || exit 1
else
  echo "$RESPONSE_BODY"
fi

Python

from __future__ import annotations

from typing import Any

from pymilvus import DataType, Function, FunctionType, MilvusClient

MILVUS_URI = "http://c-xxxx.milvus.aliyuncs.com:19530"
MILVUS_TOKEN = "<yourUsername>:<yourPassword>"

DUMMY_VECTOR_DIM = 2
TEXTTRANSFORM_FUNCTION_TYPE = 9

def texttransform_function_type() -> Any:
    for type_name in ("TEXTTRANSFORM", "TEXT_TRANSFORM", "TextTransform"):
        function_type = getattr(FunctionType, type_name, None)
        if function_type is not None:
            return function_type
    # Alibaba Cloud Milvus provides TEXTTRANSFORM as a managed extension (function type value 9);
    # some pymilvus versions do not include this enum member, while Function(...) validates through FunctionType(...).
    existing = getattr(FunctionType, "_value2member_map_", {}).get(TEXTTRANSFORM_FUNCTION_TYPE)
    if existing is not None:
        return existing
    extension = int.__new__(FunctionType, TEXTTRANSFORM_FUNCTION_TYPE)
    extension._name_ = "TEXTTRANSFORM"
    extension._value_ = TEXTTRANSFORM_FUNCTION_TYPE
    FunctionType._value2member_map_[TEXTTRANSFORM_FUNCTION_TYPE] = extension
    FunctionType._member_map_["TEXTTRANSFORM"] = extension
    return extension

def add_id(schema: Any) -> None:
    schema.add_field("id", DataType.INT64, is_primary=True)

def add_dummy_vector(schema: Any) -> None:
    schema.add_field("dummy_vector", DataType.FLOAT_VECTOR, dim=DUMMY_VECTOR_DIM)

def run_texttransform_example(*, client, collection_name, input_fields, output_field, function_name, function_params, rows) -> None:
    if client.has_collection(collection_name):
        client.drop_collection(collection_name)
    schema = MilvusClient.create_schema(auto_id=True, enable_dynamic_field=False)
    add_id(schema)
    for name, data_type, max_length in input_fields:
        field_params = {"max_length": max_length} if max_length is not None else {}
        schema.add_field(name, data_type, **field_params)
    output_name, output_data_type, output_max_length = output_field
    output_params = {"max_length": output_max_length} if output_max_length is not None else {}
    schema.add_field(output_name, output_data_type, **output_params)
    add_dummy_vector(schema)
    schema.add_function(
        Function(
            name=function_name,
            function_type=texttransform_function_type(),
            input_field_names=[name for name, _, _ in input_fields],
            output_field_names=[output_name],
            params=function_params,
        )
    )
    index_params = client.prepare_index_params()
    index_params.add_index(field_name="dummy_vector", index_type="AUTOINDEX", metric_type="COSINE")
    client.create_collection(collection_name=collection_name, schema=schema, index_params=index_params)
    client.insert(collection_name, rows)
    client.flush(collection_name)
    fields = [name for name, _, _ in input_fields] + [output_name]
    for row in client.query(collection_name, filter="", output_fields=fields, limit=len(rows)):
        print(row)

client = MilvusClient(uri=MILVUS_URI, token=MILVUS_TOKEN)

run_texttransform_example(
    client=client,
    collection_name="simple_ai_classify",
    input_fields=[("content", DataType.VARCHAR, 4096)],
    output_field=("category", DataType.VARCHAR, 64),
    function_name="simple_ai_classify_fn",
    function_params={"provider": "aliyun_milvus", "model_name": "qwen3.7-max", "task": "ai_classify", "labels": "accounts,inquiries,faults,billing", "prompt": "Classify by the topic of the customer's question.", "temperature": "0", "max_concurrency": "2"},
    rows=[{"content": "Cannot log in to the Milvus console; verification code error keeps appearing.", "dummy_vector": [0.1, 0.2]}, {"content": "How do I enable public network access for a Serverless Milvus instance?", "dummy_vector": [0.1, 0.2]}],
)

Expected result: Each REST input returns one candidate label (in testing: accounts and inquiries). The category field in the Python records holds the corresponding candidate label. Login and verification-code issues are routed to accounts or faults depending on the model's judgment.

Example 2: Classifying product images (image)

A content moderation system needs to automatically assign images to the clothing, food, landscape, or animal categories.

REST API

#!/usr/bin/env bash
set -euo pipefail

MILVUS_REST_BASE_URL="http://c-xxxx.milvus.aliyuncs.com:19530"
MILVUS_AUTH_TOKEN="<yourUsername>:<yourPassword>"

post_json() {
  local path="$1"
  local body="$2"
  curl -X POST \
    "$MILVUS_REST_BASE_URL$path" \
    -H "Authorization: Bearer $MILVUS_AUTH_TOKEN" \
    -H "Content-Type: application/json" \
    -d "$body"
}

BODY=$(cat <<JSON
{"model_name":"qwen3.7-plus","texts":["https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20260415/hynnff/wan-video-edit-clothes.webp"],"params":{"media_type":"image","labels":["clothing","food","landscape","animal"],"temperature":0}}
JSON
)
RESPONSE_BODY="$(post_json "/v2/vectordb/ai/classify" "$BODY")"
if command -v jq >/dev/null 2>&1; then
  echo "$RESPONSE_BODY" | jq .
  [ "$(echo "$RESPONSE_BODY" | jq -r '.code // -1')" = "0" ] || exit 1
else
  echo "$RESPONSE_BODY"
fi

Python

from __future__ import annotations

from typing import Any

from pymilvus import DataType, Function, FunctionType, MilvusClient

MILVUS_URI = "http://c-xxxx.milvus.aliyuncs.com:19530"
MILVUS_TOKEN = "<yourUsername>:<yourPassword>"

DUMMY_VECTOR_DIM = 2
TEXTTRANSFORM_FUNCTION_TYPE = 9

def texttransform_function_type() -> Any:
    for type_name in ("TEXTTRANSFORM", "TEXT_TRANSFORM", "TextTransform"):
        function_type = getattr(FunctionType, type_name, None)
        if function_type is not None:
            return function_type
    # Alibaba Cloud Milvus provides TEXTTRANSFORM as a managed extension (function type value 9);
    # some pymilvus versions do not include this enum member, while Function(...) validates through FunctionType(...).
    existing = getattr(FunctionType, "_value2member_map_", {}).get(TEXTTRANSFORM_FUNCTION_TYPE)
    if existing is not None:
        return existing
    extension = int.__new__(FunctionType, TEXTTRANSFORM_FUNCTION_TYPE)
    extension._name_ = "TEXTTRANSFORM"
    extension._value_ = TEXTTRANSFORM_FUNCTION_TYPE
    FunctionType._value2member_map_[TEXTTRANSFORM_FUNCTION_TYPE] = extension
    FunctionType._member_map_["TEXTTRANSFORM"] = extension
    return extension

def add_id(schema: Any) -> None:
    schema.add_field("id", DataType.INT64, is_primary=True)

def add_dummy_vector(schema: Any) -> None:
    schema.add_field("dummy_vector", DataType.FLOAT_VECTOR, dim=DUMMY_VECTOR_DIM)

def run_texttransform_example(*, client, collection_name, input_fields, output_field, function_name, function_params, rows) -> None:
    if client.has_collection(collection_name):
        client.drop_collection(collection_name)
    schema = MilvusClient.create_schema(auto_id=True, enable_dynamic_field=False)
    add_id(schema)
    for name, data_type, max_length in input_fields:
        field_params = {"max_length": max_length} if max_length is not None else {}
        schema.add_field(name, data_type, **field_params)
    output_name, output_data_type, output_max_length = output_field
    output_params = {"max_length": output_max_length} if output_max_length is not None else {}
    schema.add_field(output_name, output_data_type, **output_params)
    add_dummy_vector(schema)
    schema.add_function(
        Function(
            name=function_name,
            function_type=texttransform_function_type(),
            input_field_names=[name for name, _, _ in input_fields],
            output_field_names=[output_name],
            params=function_params,
        )
    )
    index_params = client.prepare_index_params()
    index_params.add_index(field_name="dummy_vector", index_type="AUTOINDEX", metric_type="COSINE")
    client.create_collection(collection_name=collection_name, schema=schema, index_params=index_params)
    client.insert(collection_name, rows)
    client.flush(collection_name)
    fields = [name for name, _, _ in input_fields] + [output_name]
    for row in client.query(collection_name, filter="", output_fields=fields, limit=len(rows)):
        print(row)

client = MilvusClient(uri=MILVUS_URI, token=MILVUS_TOKEN)

run_texttransform_example(
    client=client,
    collection_name="simple_ai_classify_image",
    input_fields=[("image_url", DataType.VARCHAR, 4096)],
    output_field=("category", DataType.VARCHAR, 64),
    function_name="simple_ai_classify_image_fn",
    function_params={"provider": "aliyun_milvus", "model_name": "qwen3.7-plus", "task": "ai_classify", "media_type": "image", "labels": "clothing,food,landscape,animal", "temperature": "0"},
    rows=[{"image_url": "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20260415/hynnff/wan-video-edit-clothes.webp", "dummy_vector": [0.1, 0.2]}],
)

Expected result:category returns one of clothing, food, landscape, or animal (in testing: clothing).

Example 3: Classifying video materials (video)

A video operations team needs to filter materials by the fashion, sports, travel, and animation categories. Video inference is time-consuming. Set timeout_sec to 300 or higher.

REST API

#!/usr/bin/env bash
set -euo pipefail

MILVUS_REST_BASE_URL="http://c-xxxx.milvus.aliyuncs.com:19530"
MILVUS_AUTH_TOKEN="<yourUsername>:<yourPassword>"

post_json() {
  local path="$1"
  local body="$2"
  curl -X POST \
    "$MILVUS_REST_BASE_URL$path" \
    -H "Authorization: Bearer $MILVUS_AUTH_TOKEN" \
    -H "Content-Type: application/json" \
    -d "$body"
}

BODY=$(cat <<JSON
{"model_name":"qwen3.7-plus","texts":["https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20260409/dozxak/Wan_Video_Edit_33_1.mp4"],"params":{"media_type":"video","labels":["fashion","sports","travel","animation"],"temperature":0,"timeout_sec":300}}
JSON
)
RESPONSE_BODY="$(post_json "/v2/vectordb/ai/classify" "$BODY")"
if command -v jq >/dev/null 2>&1; then
  echo "$RESPONSE_BODY" | jq .
  [ "$(echo "$RESPONSE_BODY" | jq -r '.code // -1')" = "0" ] || exit 1
else
  echo "$RESPONSE_BODY"
fi

Python

from __future__ import annotations

from typing import Any

from pymilvus import DataType, Function, FunctionType, MilvusClient

MILVUS_URI = "http://c-xxxx.milvus.aliyuncs.com:19530"
MILVUS_TOKEN = "<yourUsername>:<yourPassword>"

DUMMY_VECTOR_DIM = 2
TEXTTRANSFORM_FUNCTION_TYPE = 9

def texttransform_function_type() -> Any:
    for type_name in ("TEXTTRANSFORM", "TEXT_TRANSFORM", "TextTransform"):
        function_type = getattr(FunctionType, type_name, None)
        if function_type is not None:
            return function_type
    # Alibaba Cloud Milvus provides TEXTTRANSFORM as a managed extension (function type value 9);
    # some pymilvus versions do not include this enum member, while Function(...) validates through FunctionType(...).
    existing = getattr(FunctionType, "_value2member_map_", {}).get(TEXTTRANSFORM_FUNCTION_TYPE)
    if existing is not None:
        return existing
    extension = int.__new__(FunctionType, TEXTTRANSFORM_FUNCTION_TYPE)
    extension._name_ = "TEXTTRANSFORM"
    extension._value_ = TEXTTRANSFORM_FUNCTION_TYPE
    FunctionType._value2member_map_[TEXTTRANSFORM_FUNCTION_TYPE] = extension
    FunctionType._member_map_["TEXTTRANSFORM"] = extension
    return extension

def add_id(schema: Any) -> None:
    schema.add_field("id", DataType.INT64, is_primary=True)

def add_dummy_vector(schema: Any) -> None:
    schema.add_field("dummy_vector", DataType.FLOAT_VECTOR, dim=DUMMY_VECTOR_DIM)

def run_texttransform_example(*, client, collection_name, input_fields, output_field, function_name, function_params, rows) -> None:
    if client.has_collection(collection_name):
        client.drop_collection(collection_name)
    schema = MilvusClient.create_schema(auto_id=True, enable_dynamic_field=False)
    add_id(schema)
    for name, data_type, max_length in input_fields:
        field_params = {"max_length": max_length} if max_length is not None else {}
        schema.add_field(name, data_type, **field_params)
    output_name, output_data_type, output_max_length = output_field
    output_params = {"max_length": output_max_length} if output_max_length is not None else {}
    schema.add_field(output_name, output_data_type, **output_params)
    add_dummy_vector(schema)
    schema.add_function(
        Function(
            name=function_name,
            function_type=texttransform_function_type(),
            input_field_names=[name for name, _, _ in input_fields],
            output_field_names=[output_name],
            params=function_params,
        )
    )
    index_params = client.prepare_index_params()
    index_params.add_index(field_name="dummy_vector", index_type="AUTOINDEX", metric_type="COSINE")
    client.create_collection(collection_name=collection_name, schema=schema, index_params=index_params)
    client.insert(collection_name, rows)
    client.flush(collection_name)
    fields = [name for name, _, _ in input_fields] + [output_name]
    for row in client.query(collection_name, filter="", output_fields=fields, limit=len(rows)):
        print(row)

client = MilvusClient(uri=MILVUS_URI, token=MILVUS_TOKEN)

run_texttransform_example(
    client=client,
    collection_name="simple_ai_classify_video",
    input_fields=[("video_url", DataType.VARCHAR, 4096)],
    output_field=("category", DataType.VARCHAR, 64),
    function_name="simple_ai_classify_video_fn",
    function_params={"provider": "aliyun_milvus", "model_name": "qwen3.7-plus", "task": "ai_classify", "media_type": "video", "labels": "fashion,sports,travel,animation", "temperature": "0", "timeout_sec": "300"},
    rows=[{"video_url": "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20260409/dozxak/Wan_Video_Edit_33_1.mp4", "dummy_vector": [0.1, 0.2]}],
)

Expected result:category returns one of fashion, sports, travel, or animation (in testing: animation). If {"code":10001,"message":"request timeout"} is returned, increase the value of timeout_sec.