Pagination & Auto-Iteration

The SDK auto-paginates every list endpoint so you never hand-roll an offset= loop. Three pagination bugs routinely show up in customer code — the iter_* helpers make all three unreachable.

Available from SDK 1.2.0. The original list_jobs(offset=..., limit=...) style still works for small queries but returns at most 200 rows; use iter_jobs() whenever the result set could grow beyond that.

#The bugs we prevent

Off-by-one overflow. A hand-rolled loop that stops on items == [] issues one extra HTTP call after the server has already returned every row.
Infinite loop on partial pages. Filters can make the server return fewer than limit rows even when more rows exist behind later offsets. Callers who check len(page) == 0 loop forever.
Silent filter drop. Hand-written second-page calls often forget to re-include status= / kind=query params, so page 2 returns unfiltered rows. The customer's UI then shows seemingly-contradictory results.

#Streaming iteration

from radmah_sdk import RadMahClient

client = RadMahClient(api_key="sl_live_…")

# Yields every matching row without ever holding more than one page in RAM.
for job in client.iter_jobs(status="succeeded", kind="synthesize"):
    process_job(job)

# Cap the iteration at 500 rows — short-circuits before the next HTTP call
for job in client.iter_jobs(status="running", max_items=500):
    reassign(job)

#Eager collection

When you genuinely need every row materialised (e.g. building a cross-job comparison table), list_all_* drains the iterator and returns a list:

all_datasets = client.list_all_datasets()
all_seals = client.list_all_seals(page_size=200)
all_connectors = client.list_all_connectors()
all_synthetic = client.list_all_synthetic_datasets()

page_size is clamped to the server cap of 200. Passmax_items= if you have an upper bound — the iterator stops issuing HTTP calls as soon as the cap is reached.

#Async iteration

from radmah_sdk import AsyncRadMah AI

async with AsyncRadMah(api_key="sl_live_…") as client:
    async for job in client.iter_jobs(status="succeeded"):
        await process_job(job)

    # Or eager:
    jobs = await client.list_all_jobs(kind="virtual_scada")

#Termination guarantees

The iterator stops on any of three conditions, in order:

Server-reported total reached — the envelope carries { items, total, offset, limit } so the SDK knows when it has seen every row.
Short trailing page — the server returned fewer rows than the requested limit, so no further page can exist.
Empty page — defensive: if total was missing and a later page comes back empty, stop.