Say you are on a team that has just signed off a flight-test campaign. The rules that decide whether your hardware passed are fixed in place: conditions on telemetry channels that mark a test run as failed or an asset as out of bounds. Now you are looking at an API that manages those rules, and the question on your mind is a simple one. Can someone make a rule disappear?
That is the question the APIs.io post about the Sift Stack Rule Service API answers. Sift is an observability platform for mission-critical hardware: rockets, satellites and aircraft. Rules are its alerting layer. The Rule Service API is one of 47 APIs that Sift Stack publishes on apis.io, and this one has 24 operations on https://api.siftstack.com, all under /api/v1/rules.
Here is the detail you would not guess from the outside. None of those 24 operations uses the HTTP DELETE verb.
So you read how the surface is laid out. The operations fall into four groups. The lifecycle group covers CreateRule, UpdateRule, GetRule, ListRules and SearchRules, plus batch versions of get and update. The removal group holds ArchiveRule, UnarchiveRule, DeleteRule and UndeleteRule, along with four batch variants of the same. Every one of those is a POST. The versions group has ListRuleVersions, GetRuleVersion and BatchGetRuleVersions, because each rule keeps its history. And the representations group lets you handle rules as JSON through ViewJsonRules, UpdateJsonRules and ValidateJsonRules, or in a readable form through ViewHumanFriendlyRules and UpdateHumanFriendlyRules.
The point you take from this is that removal is two-stage and reversible. Archive takes a rule out of use. Delete is also a POST, to /rules/delete, and /rules/undelete exists alongside it. Put that together with the version history and a rule's state at any point in time can be reconstructed. In a regulated test environment, the post says, that is the requirement, and the API surface shows it.
You would also want to know what it costs you, and the post is straightforward about that. The URL style is mixed. Some actions are sub-paths, like /rules/archive and /rules/search. Others use Google-style colon methods, like /rules:batchGet, /rules:validateJson and /rules:viewHumanFriendly. GetRule is a GET on the collection path /api/v1/rules rather than on /rules/{id}, while ListRuleVersions does use a {ruleId} path parameter. The surface is generated from gRPC, and the REST mapping shows that. If you are writing a client, expect to learn a few conventions instead of one.
There is one more thing to watch before you hard-code anything. Rule evaluation appears in two places. This API has EvaluateRules at /rules/evaluate. A separate Rule Evaluation Service API has two operations: EvaluateRules at /rules/evaluate-rules and a :preview variant. The post's advice is to check which one the current docs point to before you pick either.
Then comes the part where the post turns on its own catalog. The Sift Stack listing on apis.io has paired near-duplicate records, such as channelservice and channel-service, and jobservice and job-service. The post is blunt about whose problem that is: it is the catalog's cleanup to do, not Sift's. So if you land on a duplicate record while you are looking around, you know the mess is on the listing side and not a sign of anything wrong with Sift's API.
The contract and supporting material are published for you to check against. The OpenAPI contract, authentication, error catalog, rate-limit and conventions artifacts are all on artifacts.apis.io, and the human documentation is at docs.siftstack.com.
What you take away is a short checklist for any API that manages rules deciding whether hardware passed a test. Look for archive, undelete and versions before you look for create. This one has all three. Then confirm which evaluation endpoint the current docs point to, and budget a little time for the mixed URL style.
Read the original: https://apis.io/2026/11/17/sift-stacks-rule-service-api-never-deletes-a-rule-it-files-a-post-to-undo/