HighLevelRestClient7.17.5 date_histogram offset is unexpected

Elastic Stack Elasticsearch
1 
DateHistogramAggregationBuilder dateHistogramAggregationBuilder = AggregationBuilders
        .dateHistogram(AggValueEnum.TIME_INTERVAL.getAggName()).field(AggValueEnum.TIME_INTERVAL.getField()).minDocCount(0).extendedBounds(longBounds);
dateHistogramAggregationBuilder.fixedInterval(timeIntervalAggRequest.getDimension());
dateHistogramAggregationBuilder.offset("7d");

The final HTTP request sent was

"aggregations": {
    "time_interval": {
      "date_histogram": {
        "field": "ddtime",
        "fixed_interval": "7d",
        "offset": 604800000,
        "order": {
          "_key": "asc"
        },
        "keyed": false,
        "min_doc_count": 0,
        "extended_bounds": {
          "min": 1761696000021,
          "max": 1769472000000
        }
      },
......

offset is parsed as a long type.

"offset": 604800000,

But ES expects a string. ("offset": "6d",)

"aggregations": {
    "time_interval": {
      "date_histogram": {
        "field": "ddtime",
        "fixed_interval": "7d",
        "offset": "6d",
        "order": {
          "_key": "asc"
        },
        "keyed": false,
        "min_doc_count": 0,
        "extended_bounds": {
          "min": 1761696000021,
          "max": 1769472000000
        }
      },

client code:

public static long parseStringOffset(String offset) {
        if (offset.charAt(0) == '-') {
            return -TimeValue.parseTimeValue(
                offset.substring(1),
                null,
                DateHistogramAggregationBuilder.class.getSimpleName() + ".parseOffset"
            ).millis();
        }
        int beginIndex = offset.charAt(0) == '+' ? 1 : 0;
        return TimeValue.parseTimeValue(
            offset.substring(beginIndex),
            null,
            DateHistogramAggregationBuilder.class.getSimpleName() + ".parseOffset"
        ).millis();
    }

Is this a bug, how can I solve it?