Released: November 19, 2026
Overview
Version 1.23 of the API Engine Library helps integrations stay within the limits of the platforms they connect to, and makes it easier to work with large amounts of data. It introduces a rate limit framework that lets the API Engine pause and hold calls when a platform asks it to slow down, and a Run in Background action that runs an automation group without tying up your screen. Large CSV responses of up to 150 MB now load (over 10 MB, with Error Processing Method None or Default), and field mappings can be made conditional so that a value is only assigned when the response, or the state of the record, calls for it.
The release also includes improvements to the GraphQL Variables Builder and the API Sets page, along with fixes for value totals, failed calls, and stale response data.
Added
Rate Limit Framework for API Calls:
- Each API Function now has a Rate Limit Method setting, so the API Engine can respect a platform's requests to slow down instead of sending calls that will be refused.
- The default is None, which leaves current behavior unchanged. Methods other than None are supplied by Suite Engine integration apps and custom extensions, so existing functions are not affected until an app provides one.
- When a platform asks the engine to slow down, calls that share the same platform allowance (for example, one store) are paused. The pause is shared by all users, background sessions, and companies in the environment, and is capped at 15 minutes.
- Before each call is sent, the engine checks for an active pause (only for functions with a Rate Limit Method other than None). A short pause is waited out, for up to 60 seconds in an automation step running in a background session (such as the job queue) and up to 5 seconds everywhere else, such as the web client or Run once. If the pause is longer than the wait allowed, the call is not sent.
- A call that is held back, or that the platform refuses, is marked on the API Message with a new Rate Limited Until field. For a held call, the reason is written in the response text (for example, "Not sent: rate limit pause until ..."). For a refused call, the platform's own status code and response are kept. Either way it counts as unsuccessful, so the calling app can retry it, and a paged retrieval stops at that message. The engine never sends a held call later on its own.
- A rate limited attempt does not load or process a response and does not move the function's Last Execution Timestamp forward, so the next retrieval does not skip any records.
Run in Background for Automation Groups:
- A new Run in Background action on the Automation Groups page runs the selected group once in the background, so you can keep working while it runs.
- It works for groups with or without their own scheduled job queue entry, and the results appear in the run history.
- A notification confirms the run has started and links directly to the group's run history.
- The action will not start a second run while an earlier background run of the same group is still queued or running, or while a scheduled job queue entry covering the group is running. The message explains why and points you to Job Queue Entries.
- Users who do not have permission to schedule background jobs are offered the option to run the group on screen instead. Disabled groups are refused with an error.
Large CSV Response Support:
- CSV responses are now read and loaded into the Data Buffer one row at a time, instead of being built as several complete copies (text, parsed rows, JSON) in memory, so memory use no longer multiplies with the size of the file.
- The limits are raised for every CSV function, from 100,000 cells to 10,000,000 cells, and from 10 MB to 150 MB per response.
- Rows can have up to 16,384 columns and up to 10,485,760 characters.
- Responses over 10 MB load only when the function's Error Processing Method is None or Default. Smaller responses work with every method, as before.
- Zero bytes inside a response are skipped, and the remaining rows still load.
Conditional Field Mappings (Apply When):
- Non-key field mappings of type Element and Record Node Value can now have one or more Apply When conditions. All conditions must match for the mapping to assign its value. If they do not, the mapping is skipped and the field keeps its existing value.
- Conditions compare a value from the response using Equals, Does Not Equal, In (a list of allowed values), Greater Than, or Less Than. Value Type can be Text, Boolean (entered as Yes/No), Number, Date, or Date and Time. Greater Than and Less Than require Number, Date, or Date and Time.
- A condition can read a value from the same object as the mapping or from one of its parent objects, so a mapping for an order line can depend on a value in the order header. A value in a separate repeated list cannot be used. If the source cannot be found in the response, the mapping is skipped. The source must be a single value given as a full response path starting with a slash. If the path matches several values, or points to an object or list, processing stops with an error instead of skipping.
- The Mapping Card has a new Apply When section with Add Condition, All Conditions, Remove Condition, and Allowed Values actions, and a Response Source lookup that lists paths from the most recent saved response for the function and from the function's mapping sources. A confirmation is shown before a condition is removed, or before a change of comparison or value type clears the values already entered.
- The Mappings list has a new When column that summarizes each mapping, for example “Always” or “2 conditions”.
- Conditions stay with their mapping when a mapping, function, or API Set is renamed, when a function is copied or deleted, and in both configuration package and JSON export and import.
- When the processing log level is 5 or higher, each skipped mapping is logged with the target field, the response source, and the value that did not match.
- The Data Buffer now records whether a JSON value is a simple value or a container, so an empty text value can be told apart from an empty object. Responses loaded by an earlier version may need to be reloaded before a condition can read an empty value.
Apply During for Field Mappings:
- A new Apply During setting on field mappings controls whether a value is assigned when the API Engine inserts a new record, when it modifies an existing record, or both. The options are Insert and Modify (the default), Insert Only, and Modify Only.
- Existing mappings keep Insert and Modify, so current behavior does not change.
- Only records that the API Engine inserts from a Record Node count as new. The primary record of a function always counts as modified, so Insert Only never applies to it.
- Apply During works together with Apply When. Both must pass for a mapping to assign its value.
- Apply During is supported on the same mappings as Apply When. A different value on any other mapping is rejected.
- Mappings skipped because of Apply During are logged with the reason at log level 5 or higher.
Updated
GraphQL Variables Builder:
- Problems with the function, the JSON, or the GraphQL text are now reported before any confirmation question is asked.
- The question about deleting existing variables now appears only when the function has Body variables that would be deleted.
- Leaving the GraphQL text box now checks the text and shows either “GraphQL structure looks valid.” or a warning, matching the check already done for the JSON box.
- The Variables Generated count now opens the generated Body variables when selected.
- The FastTabs are renamed General and Query or Mutation.
- The page no longer appears in search. It opens from the API Functions list.
API Sets Page Ribbon:
- All actions on the API Sets page (Functions, Credentials, Shared Session, Load Data, Export, Import, and Delete) are now together in a single ribbon instead of being spread across separate menus.
- Action names are shortened, for example Functions, Credentials, Load Data, and Delete.
- The previous Setup group is hidden and marked obsolete.
Credentials Page Export:
- Export is now on the promoted action bar of the Credentials page, so it is available without opening a menu.
Data Load Timestamp Handling:
- The Last Execution Timestamp seed supplied when API function data is loaded is now read as ISO 8601 only (yyyy-MM-ddTHH:mm:ss, or with a trailing Z for UTC), so the stored value is the same in every region. A blank seed still means the function has never been executed.
- A value that is not ISO 8601, such as a regional date like 01/01/21 12:00 AM, is no longer applied and no longer stops the data load. A new function starts at the beginning of today, an existing function keeps its current timestamp, and a warning is logged naming the API Set, the function, and the rejected value.
- Apps and custom extensions that load API data should supply a blank or ISO 8601 seed.
Clearer Wording for Automations With an Unknown Outcome:
- Messages for an automation that was handled by another extension and did not report its outcome now say “handled by another extension” instead of “run by another Suite Engine app”, so it is clear the API Engine did not call a separate app.
- The run history result for these automations now tells the extension developer to call CompleteAutomationRun or FailAutomationRun to record the outcome.
Page Caption Consistency:
- Four page captions now start with “API” for consistency. They are API Application Executions, API Automation Run History, API GraphQL Variables Builder, and API Processing Log Entries.
- The Processing Log Entries page no longer shows the internal SENAPI prefix in its caption.
Spelling Corrections:
- Spelling errors are corrected in captions, tooltips, help text, and messages across the API Engine, including Export API Configuration, “formatted”, “received”, “parameters”, and “successful”.
Fixed
Sum Value Processing Totaled Only One Node:
-
When the path to total was entered in Source ID, the Sum value processing type totaled only the one node handed to it, so an order total could come back too low. Sum now totals every node on the path within the record being built, such as the whole order, and Mapping Source Filters still apply. If the right level cannot be found, Sum returns a blank value instead of totaling rows outside the record. Mappings with the path in Static Value are unchanged.
Failed Calls Without an Error Root Mapping Caused an Error:
-
A failed call, such as an HTTP 429 or 400 response, on a Single Record Update or Multiple Records Update function with no error root mapping caused an error during response processing. The engine now recognizes that there is nothing to process, and the message keeps its HTTP status and response body so the calling app can read the result and act on it. This also stops a reused processing instance from picking up data from the previous message.
Mappings Without a Field Could Reuse the Previous Mapping's Data:
-
A mapping with no target field could inherit the source element found for the mapping before it. Each mapping is now evaluated only against its own source element. As a result, a mapping with no target field that uses Event Subscription value processing no longer raises its event unless Set Value on Missing Source ID is turned on.
Stale Data Buffer Rows After a Failed Reload:
-
When a response was reloaded into the Data Buffer and the load failed, rows from the earlier load could remain or come back. For RAW, JSON, XML, and CSV responses, the old rows are now removed and the removal is saved before the response is read, so a failed load leaves no rows behind. A response that is not valid JSON is no longer marked as Loaded, and a warning is logged instead. The Load Response action on the API Message page now tells you when the response was not loaded.