Version 0.1.0-beta.5

TL;DR: Data hub arrives in Loomi Connect: browse item collections and what's inside them, manage collections and their configurations, and track catalogs and jobs. Marketing gains email preview links, live customer attribute values, event retention settings, and customer filter creation. Saved reports, trends, and funnels now return the definitions behind them, and scenario writes finally explain what went wrong.

Manage Data hub item collections

Browse the item collections in a workspace, the product fields observed in each one, and the items and records they hold.
Create collections, update their metadata, and delete them.
Guardrails apply to every write. Blank workspace IDs and invalid collection names are rejected before the request leaves, and you are asked to confirm the target first.
Updates and deletes are recorded with a snapshot of the previous state. There is no automatic rollback, so confirm the target before applying a change.
Data hub writes are throttled more tightly than reads. See performance and rate limits.

Inspect and change item collection configurations

Fetch a collection's current configuration, or an earlier version of it by configuration ID.
Compare two configurations to see exactly what differs between them.
Validate an attribute against a collection before you commit a configuration change.

Track Data hub catalogs, jobs, and system settings

Browse the unified catalogs in a workspace.
List jobs for a workspace or for a project, filter them by resource, and poll a single job by ID until it finishes.
Read a workspace's system configuration.

More Marketing capabilities

Generate a shareable browser preview link for an email campaign, optionally personalized against a real customer, in a specific language, or for one A/B variant. The link is publicly accessible, so treat it accordingly.
Evaluate a customer's aggregate and expression values live. When some attributes can't be evaluated you still get the ones that could, along with which failed.
Read a project's event retention settings, including the project-wide default and any per-event-type overrides. Use it to explain analytics metrics that dropped because event data expired. See project settings and access.
Create persisted customer filters.

See the definitions behind saved analyses

Saved reports, trends, and funnels now return their metrics, filters, and parameter details, so you can inspect an analysis without running it. All three previously reported an empty definition.
Parametrized reports can be calculated by supplying values at run time, without saving those values in the Marketing interface first.
A failed calculation returns the backend's own message, so a missing parameter names the parameter instead of failing opaquely.

Author scenarios with the real schema

Scenario detail now includes the connections between nodes, so you can see how an existing scenario is wired before copying the pattern. See automation scenarios.
Branch connector numbering is now documented correctly: a condition's match output is 1 and its no-match output is 2, split outputs start at 1, and a destination is normally 0.
Recurring and date-attribute triggers now document the schema the backend actually uses, including that weekly days run 1 to 7 and that times are hour and minute pairs rather than strings.

Recurring triggers are also stricter than before. A payload using the outdated field names previously reported success and then silently ran on the wrong cadence, because the backend dropped the unrecognized fields. Those payloads are now rejected before they are sent. If a recurring scenario was appearing to save while running on an unexpected schedule, that is why, and it will now report an error instead.

Find customers by what they did not do

Ad-hoc analytics queries now cover how to express absence, for cases like customers who bought something but never gave consent. This includes absence within a time window and absence combined with positive conditions. It also spells out where absence works directly and where you need a zero-count check instead, which is a common source of silently wrong results. See ad-hoc analytics queries.

Bug fixes

Customer event history ignored the date range you passed and now respects it.
Recurring scenarios could not be saved at all. They now save, and validation errors name the exact fields at fault instead of returning a bare bad-request error.
A failed campaign write returns the backend's validation message, and no longer clears cached data when the write did not succeed.
Formula variables in expression details now include the ID of the aggregate or expression they are bound to, so you can follow a formula through to what feeds it.