enh(notion) change how we process dbs

[fontanierh]

Description

• Remove all the inline database logic (which was broken anyway...). This may save quite a bit of requests to notion 🤔. We continue to add child databases we don't know about to "resources to check"
• Insert inline placeholders (Child Database ${DB_TITLE}) for databases that are children of a block
• upsert databases to semantic store as CSV files (using header as prefix)
• upsert pages that don't have a body if one of their properies has over 256 characters

Risk

Notion sync could get stuck, but easy to rollback

Deploy Plan

Requires a queue bump + restart of workflows

[spolu]

Looks great but the prefix logic seems not adequate? We want the headler here and not in the csvForDocument

[spolu]

csvHeader?

[fontanierh]

So this is the header right ? What you don't want is DB title ?

happy to remove csv header from the body though

[fontanierh]

Ok I see it ! Fixed (removed DB name from prefix also, did we want it ?)

[fontanierh]

used the renderPrefixSection with the defaults -- should I override to have more than 64 tokens for the header ?

[fontanierh]

Ok ended up doubling it.

[spolu]

LGTM

[spolu]

I thiiiink we want the DB name in the prefix no?
Also if csvHeader is > than half chunk size the upsert will fail.

You can likely use the functions that have been built by @philipperolet that are meant to do that but maybe the numbers are a bit tight for this use case.

In any case we want to tokenize and split at half chunk size aka 256 tokens.

[fontanierh]

This is outdated right?