es如何先distinct某些字段再group其中的部分字段？

问题来源：有一个AWS的redshift数据仓库的数据想搬回自己搭建的Elasticsearch内，之前基于postgresql语法写的一些sql需要转换成DSL
SQL简化如下，省略where条件等如下：

with cte as (

  SELECT distinct user_id,countymd,countdate,price,action_type,createdate FROM xxx_table

)

select countymd, action_type, count(distinct user_id) as uu, sum(price) as total from cte group by countymd, action_type

本想借着distinct A, B 就等于group by A, B 的思路先 group by 那些distinct的字段，然后在结果集中再group by 
countymd, action_type，但是报错：
[composite] aggregation cannot be used with a parent aggregation
DSL如下：

{

    "size": 0,

    "query": {

        "bool": {

            "filter": {

                "range":{"countdate":{"gt":"2020-08-06 11:00:00", "lt":"2031-08-06 11:59:59"}}

            }

        }

    },

    "_source": false,

    "aggs": {

        "distinct": {

            "composite": {

                "sources": [

                    {"user_id": {"terms": {"field": "user_id"}}},

                    {"countymd": {"terms": {"field": "countymd"}}},

                    {"countdate": {"terms": {"field": "countdate"}}},

                    {"price": {"terms": {"field": "price"}}},

                    {"action_type": {"terms": {"field": "action_type"}}},

                    {"createdate": {"terms": {"field": "createdate"}}}

                ]

            },

            "aggs": {

                "groupby": {

                    "composite": {

                        "sources": [

                            {"countymd": {"terms": {"field": "countymd"}}},

                            {"action_type": {"terms": {"field": "action_type"}}}

                        ]

                    }

                },

                "aggs": {

                    "uu": {

                        "cardinality": {"field": "user_id"}

                    },

                    "total": {

                        "stats": {"field": "price"}

                    }

                }

            }

        }

    }

}

请问该如何解决这种先 distinct 再 group by 的查询？