You feed Elastly your data in one batch: open it, stage each entity in dependency order, commit, and
poll until it drains. This works in every language. Set up your client first (see the
overview).
The connector framework does this whole loop for you,
with paging and checkpoints handled. Read this page when you want to drive each step yourself.
The records you send
Every SDK sends the same canonical shapes. Money is always integer cents. These are the fields; the
stage records endpoint has the full schema and a live playground.
Echo the pricingDecisionId on every quote line. It is the id Elastly returned when it priced
the line. Sending it back is the whole learning loop: it ties what Elastly recommended to what your
rep actually charged. A sync of quote lines with none is flagged loudly, and the model learns
nothing from them.
Open, stage, commit, poll
Stage each entity in dependency order (products, then customers, then quotes and orders, so a line
always resolves to a customer and product already sent). A batch stages up to 1000 records per
call, so page large catalogs across several stage calls before you commit.
Watch two things on the drained status:
skippedOverCap counts products not persisted because your workspace hit its SKU cap. Nothing is
dropped silently; the number tells you to raise the cap or trim.
- A
failed status carries an error. The
batch status endpoint documents every field.
Next: apply the prices you approve.