How to configure KafkaAvroDeserializer to use RecordNameStrategy?

[chiang8]

Hi,

I am having trouble browsing the topic messages via kafka-ui.

A bit of the background first. I am running Kafka cluster, schema registry and kafka-ui in a local Docker container. Messages, which are serialized in Avro format, are sent to a topic successfully with a Kafka client. I was trying to browse the messages and the message values are in a scrambled format.
I checked the backend log and it's having trouble getting the right schema from schema registry as it's using a wrong subject(ie. ${topic}-value). Since this topic contains messages with different schemas, the right SubjectNameStrategy should have been io.confluent.kafka.serializers.subject.RecordNameStrategy.
I dig a little bit deeper. I can see SchemaRegistrySerde picks the right MessageFormatter(ie. AvroMessageFormatter) to deserialize the message but the internal KafkaAvroDeserializer is configured to use io.confluent.kafka.serializers.subject.TopicNameStrategy. As a result of the lookup failure, kafka-ui falls back to the StringSerde, which explains the scrambled message.

I am wondering if there is a way to override this behaviour. Thx.

[germanosin]

@chiang8 By default, SchemaRegistrySerde uses the first 4 bytes to determine the schema version during deserialization. The TopicNameStrategy is applied only when serializing.

[chiang8]

This is the request captured from the schema registry log. It does extract the schema id from the message. Unfortunately subject name(ie. xxx-value) is also part of the query. It would have worked if subject name is omitted.

[2025-09-29 03:41:41,354] INFO 172.253.63.141 - - [29/Sep/2025:03:41:41 +0000] "GET /contexts/.dev/schemas/ids/6?fetchMaxId=false&subject=xxx-value HTTP/1.1" 404 6512 "-" "Java/21.0.6" 14 (io.confluent.rest-utils.requests)

Please correct me if I am wrong. I guess it comes down to the underlying deserializer on how to deserialize the message.

    @Override
    public String format(String topic, byte[] value) {
      Object deserialized = avroDeserializer.deserialize(topic, value);
      var schema = AvroSchemaUtils.getSchema(deserialized);
      return JsonAvroConversion.convertAvroToJson(deserialized, schema).toString();
    }

ref: https://github.com/kafbat/kafka-ui/blob/v1.3.0/api/src/main/java/io/kafbat/ui/serdes/builtin/sr/MessageFormatter.java#L46-L50

'KafkaAvroDeserializer' will include both the subject and schema id upon looking up for the schema from schema registry. It would have omitted the subject name from the query if subject is null though. With TopicNameStrategy in place, it returns ${topic}-value. With RecordNameStrategy instead, it would have returned null.