A daily batch sync from a lead CRM into a sales CRM, with a ledger

The scenario sets its variables, then makes three reads against the lead CRM's API for contacts and opportunities. Instead of writing each record as it arrives, an iterator turns every record into a JSON object and a text aggregator joins them into one batch body, sent to the sales CRM's batch upsert endpoint in a single request.

A second iterator, JSON builder and aggregator build the deal batch the same way, so deals are written after the contacts they belong to. A last iterator makes the per-record calls that remain once the two batches are in.

Both batches go to upsert endpoints, which update a record that already exists and create one that does not. A contact who changes on two different days is written twice to the same record rather than duplicated.

Each synced record is written to a data store with add-or-replace, so there is one row per record and a place to see what crossed and when without opening either CRM.

The batch runs once a day in the afternoon, when its cost is predictable and its batches are full. A second scenario takes the lead CRM's webhook and runs the same read, build and batch steps for a single event, for the cases where the next day is too late.