FunctionType.TEXTEMBEDDING is incorrectly parsed as TextEmbedding, resulting in the error: java.lang.IllegalArgumentException: No enum constant io.milvus.common.clientenum.FunctionType.TextEmbedding

[FbyBruce]

When using FunctionType.TEXTEMBEDDING in the milvus-sdk-java 2.6.1 , the generated function type string is incorrectly set to "TextEmbedding" , as a result, the enumeration item TextEmbedding cannot be found in the io.milvus.common.clientenum.FunctionType package.
For example:
schema.addFunction(Function.builder() .name("siliconflow_embedding") .functionType(FunctionType.TEXTEMBEDDING) )
The above code internally sets the function type as "TextEmbedding", which is not recognized by Milvus.
This appears to be due to incorrect string mapping in the FunctionType enum—likely a missing explicit toString() override or improper annotation for serialization.
This issue prevents proper use of text embedding functions and may affect other functionality relying on correct function type names.
Could you please confirm if this is a known issue or consider fixing the enum string representation?
Thanks!

[yhmo]

No, it's not a known issue.

functionType is converted to a number and passed to the server. Both the SDK and milvus server define the TextEmbedding = 2.
Is your milvus-sdk-java 2.6.1 downloaded from the Maven repo, or built by yourself?

[FbyBruce]

from the Maven repo

[yhmo]

I didn't get the error.

import io.milvus.common.clientenum.FunctionType;


        CreateCollectionReq.CollectionSchema schema = CreateCollectionReq.CollectionSchema.builder()
                .build();
        schema.addField(AddFieldReq.builder()
                .fieldName(ID_FIELD)
                .dataType(DataType.Int64)
                .isPrimaryKey(true)
                .autoID(false)
                .build());
        schema.addField(AddFieldReq.builder()
                .fieldName("text")
                .dataType(DataType.VarChar)
                .maxLength(65535)
                .enableAnalyzer(true) // must enable this if you use Function
                .build());
        schema.addField(AddFieldReq.builder()
                .fieldName(VECTOR_FIELD)
                .dataType(DataType.FloatVector)
                .dimension(1536)
                .build());

        schema.addFunction(Function.builder()
                .functionType(FunctionType.TEXTEMBEDDING)
                .name("openai_embedding")
                .inputFieldNames(Collections.singletonList("text"))
                .outputFieldNames(Collections.singletonList(VECTOR_FIELD))
                .param("provider", "openai")
                .param("model_name", "text-embedding-3-small")
                .param("credential", "apikey1")
                .param("dim", "1536")
                .param("user", "user123")
                .build());

        List<IndexParam> indexes = new ArrayList<>();
        indexes.add(IndexParam.builder()
                .fieldName(VECTOR_FIELD)
                .indexType(IndexParam.IndexType.SPARSE_INVERTED_INDEX)
                .metricType(IndexParam.MetricType.BM25) // to use full text search, metric type must be "BM25"
                .build());

        CreateCollectionReq requestCreate = CreateCollectionReq.builder()
                .collectionName(COLLECTION_NAME)
                .collectionSchema(schema)
                .indexParams(indexes)
                .consistencyLevel(ConsistencyLevel.BOUNDED)
                .build();
        client.createCollection(requestCreate);

[yhmo]

Did you get the error when you called the client.describeCollection() ?

[FbyBruce]

If I comment out the following part：

schema.addFunction(Function.builder()
     .name("siliconflow_embedding")
     .functionType(FunctionType.TEXTEMBEDDING)
     .inputFieldNames(Collections.singletonList("patentName"))
     .outputFieldNames(Collections.singletonList("embedding"))
     .param("provider", "siliconflow")
     .param("model_name", "BAAI/bge-large-en-v1.5")
     .build());

calling client.describeCollection() outputs normal
But once I add this part，calling client.describeCollection() will report the following error:

java.lang.IllegalArgumentException: No enum constant io.milvus.common.clientenum.FunctionType.TextEmbedding
	at java.base/java.lang.Enum.valueOf(Enum.java:273) ~[na:na]
	at io.milvus.common.clientenum.FunctionType.valueOf(FunctionType.java:24) ~[milvus-sdk-java-2.6.1.jar:na]
	at io.milvus.v2.utils.SchemaUtils.convertFromGrpcFunction(SchemaUtils.java:218) ~[milvus-sdk-java-2.6.1.jar:na]
	at io.milvus.v2.utils.SchemaUtils.convertFromGrpcCollectionSchema(SchemaUtils.java:162) ~[milvus-sdk-java-2.6.1.jar:na]
	at io.milvus.v2.utils.ConvertUtils.convertDescCollectionResp(ConvertUtils.java:153) ~[milvus-sdk-java-2.6.1.jar:na]
	at io.milvus.v2.service.collection.CollectionService.describeCollection(CollectionService.java:318) ~[milvus-sdk-java-2.6.1.jar:na]
	at io.milvus.v2.client.MilvusClientV2.lambda$describeCollection$15(MilvusClientV2.java:384) ~[milvus-sdk-java-2.6.1.jar:na]
	at io.milvus.v2.utils.RpcUtils.retry(RpcUtils.java:100) ~[milvus-sdk-java-2.6.1.jar:na]

[FbyBruce]

And I tried the code you gave above,There is a type incompatibility issue in the index configuration of the vector database.I tried changing the indexType and metricType,The problem of No enum constant io.milvus.common.clientenum.FunctionType.TextEmbedding appeared again.

[yhmo]

Yes it is a bug. The name of the io.milvus.common.clientenum.FunctionType.TEXTEMBEDDING is not eqaul to io.milvus.grpc.FunctionType.TextEmbedding.
I will fix it by this pr #1536

[FbyBruce]

Thank you for confirming and fixing the issue! I appreciate your quick response and the PR #1536 to address it.

[yhmo]

I just published a new minor version v2.6.2, since this bug is a bit critical.
The jar is already published to the maven repo:
https://repo.maven.apache.org/maven2/io/milvus/milvus-sdk-java/