CloudWatch Logs Insights query:

fields @message |
filter @message like 'request body to the server' |
parse @message "'sender': '*', 'message'" as sender |
stats count(*) by sender

Query results:

-------------------------------------------------
|               sender               | count(*) |
|------------------------------------|----------|
| 65ddd20eac244AAe619383e4d8cb558834 |     4    |
| 55ddd20eac244AAe619383e4d8cb558834 |     3    |
-------------------------------------------------

Screenshot:

you can use filter.

fields @timestamp, @message
| filter @message like "65ddd20eac244AAe619383e4d8cb558834"
| sort @timestamp desc
| limit 20

it will filter all the messages limit to 20 that send by 65ddd20eac244AAe619383e4d8cb558834.

update:

suppose the JSON log formate is this

{
    "sender": "65ddd20eac244AAe619383e4d8cb558835",
    "message": "Hi"
}

Now I want to count number of messages from 65ddd20eac244AAe619383e4d8cb558835

how many messages are coming from each user?

so simple you can run the query

stats count(sender) by sender |
# To filter only message the contain sender, to avoid lambda default logs
filter @message like "sender"

if you want to see messages as well then modify the query a bit

stats count(*) by sender, message |

filter @message like "sender"

Here @message refers to whole to index where message refer to the JSON object message.

count_distinct

Returns the number of unique values for the field. If the field has very high cardinality (contains many unique values), the value returned by count_distinct is just an approximation.

how many distinct users in the selected interval?

It will list distinct users in 3hr of interval

 stats count_distinct(sender) as distinct_sender by bin(3hr) as interval