Due to the fact that Serverless doesn’t support deletion of record with metadata filtering, the doc here explains how to use the list
operation to fetch the IDs of the records and then delete the records by ID.
The doc suggests that “If there are more than 100 records to return, the list
response contains a pagination key”:
# Response:
# {
# "vectors":
# [
# { "id": "doc1#chunk1" },
# { "id": "doc1#chunk2" },
# { "id": "doc1#chunk3" },
# { "id": "doc1#chunk4" },
...
# ],
# "pagination": {
# "next": "c2Vjb25kY2FsbA=="
# }
# }
However in reality I am getting a pagination key with much less than 100 vectors in the response:
pagination: {
next: 'eyJza2lwX3Bhc3QiOiI2MGZmODg5ZS1hYWY1LTRiNTktOWUwNS1mNmRmNTVhOWM5Yjc6Y2h1bmtfMyIsInByZWZpeCI6IjYwZmY4ODllLWFhZjUtNGI1OS05ZTA1LWY2ZGY1NWE5YzliNyJ9'
}
Decoding the pagination value reveals the following in my case:
{
"skip_past": "60ff889e-aaf5-4b59-9e05-f6df55a9c9b7:chunk_3",
"prefix": "60ff889e-aaf5-4b59-9e05-f6df55a9c9b7"
}
which is OK but not documented. What should be the decoded value of the pagination key if there are more than 100 records? skip_next, skip_forward?