Divergences from Timefold

satisfactory's public API is a reimplementation of the Timefold Platform's model API: the same vocabulary, paths, statuses and delivery rules. The Timefold Platform has run as a product, and its choices encode operational experience. Where satisfactory differs it is for one of the reasons in this table.

Timefold Platform	satisfactory	Why
Enterprise Solver: multi-threaded solving, nearby selection, a throttling consumer	Community Solver: maxThreadCount validated and capped at 1; its own report throttle	Licence. Enterprise features arrive only as a deliberate purchase
Score analysis with match counts and justifications	{ score, constraints: [{ name, weight, score }] }, each constraint's contribution computed by switching it off; includeJustifications answered with "unsupported"	SolutionManager.analyze is Enterprise-only in Timefold Solver 2.7
Four commercial models	Two Apache-2.0 quickstart models behind an interface	satisfactory is the model author, not the model vendor
A managed cloud, or a self-hosted Kubernetes install	Two ankka services on any ankka installation	It is what ankka is for
Server-sent events with no documented resumption; 410 once final	?after=<seq> resumes exactly; ?follow=lineage continues into a successor	Every event carries a journaled sequence number, so the receiver can always dedupe and the platform never has to drop
Dataset listing deprecated in favour of the UI	GET /{entity} kept, with status, tag and paging filters	An integrator needs to find datasets by API, and there is no UI
No idempotency on submit	Idempotency-Key honoured for 24 hours	A retried submit should not solve twice
from-patch in preview, behind a flag	Built in from the first version	It is the real-time planning story
No documented behaviour for deriving from a dataset that is still solving	Supersede: the parent completes with its best solution and names the child; the child warm-starts	Lineage preserved; a client should not have to terminate to change a problem
Webhooks configured in the UI; X-Timefold-* headers; a read timeout not retried	Subscriptions on the platform API; X-Satisfactory-* headers with the same signing scheme; every failure retried, with a stable event id to dedupe on	The ankka integration registers programmatically; at-least-once with dedupe loses nothing
Personal access tokens issued by the platform	Tokens from the identity provider ankka already uses	One identity, already verified
Webhook signing over the body or the path, the timestamp beside it	The timestamp is inside the signed string	A replayed request cannot carry a fresh timestamp
POST /v1/demo-data in the docs, GET in the spec	GET	The spec agrees

What is adopted unchanged

The dataset vocabulary and the ten statuses; typed integer weights per constraint; configuration profiles referenced by id or name; operation=NONE and POST /{id}; the stateless score analysis; metadata on the event stream and in webhooks, never solutions; terminate as DELETE /{id} returning the dataset, purge as storage deletion, and restore while retention lasts; gzip with the same limits; the delivery rules of five seconds, ten retries, ten seconds apart; and the advice to receivers to answer at once and process asynchronously.