Go-elasticsearch Term Aggregations using Typed Client

Elastic Stack Elasticsearch
1

I've been struggling to create term aggregations using the typed client v8.13.1, and am wondering whether I've uncovered a bug.

To start, I've created three simple test documents like the following:

{
    "type": "A"
}

"type" is a text field with a keyword subfield. Using curl , I run the following aggregation perfectly fine:

curl --location 'http://localhost:9200/agg-test/_search' \
--header 'Content-Type: application/json' \
--data '{
	"size": 0,
    "aggregations": {
        "type_buckets": {
            "terms": {
                "field": "type.keyword"
            }
        }
    }
}'

which results in the following:

{
    "took": 184,
    "timed_out": false,
    "_shards": {
        "total": 1,
        "successful": 1,
        "skipped": 0,
        "failed": 0
    },
    "hits": {
        "total": {
            "value": 3,
            "relation": "eq"
        },
        "max_score": null,
        "hits": []
    },
    "aggregations": {
        "type_buckets": {
            "doc_count_error_upper_bound": 0,
            "sum_other_doc_count": 0,
            "buckets": [
                {
                    "key": "A",
                    "doc_count": 2
                },
                {
                    "key": "B",
                    "doc_count": 1
                }
            ]
        }
    }
}

In trying to reproducing this using the typed client, I've defined the aggregation like so in Go:

resp, err := es.Search().Index("agg-test").
		Request(&search.Request{
			Size: some.Int(0),
			Aggregations: map[string]types.Aggregations{
				"type_buckets": {
					Terms: &types.TermsAggregation{
						Field: some.String("type.keyword"),
					},
				},
			},
		}).
		Do(context.Background())

however, the response aggregation has 824637408744 buckets, all but the first of which has a doc count of 0. The bucket key is also nil, and I'm seeing the following docCountError:

 unreadable: protocol error E08 during memory read for packet $m2,8

I'm hoping someone can confirm whether this is indeed a defect in the typed client or a misuse of it.