Back to the changelog

• Improvement

  Rules wait for the phase's tasks

  When a case enters a phase, rules are evaluated only after the phase's tasks have been created. Previously a re-evaluation arriving at that moment could move the case on before a blocking task was created, leaving the task orphaned on a phase already left. Now that evaluation is postponed by a few seconds, and a late re-evaluation of the previous phase can no longer decide for the new one; an event received while the case was entering the phase is evaluated when the entry completes. If entering a phase is interrupted by a failure, automatic recovery completes it: tasks still open are not duplicated, while automatic tasks already completed before the failure may run again. A few edge cases (a manual move exactly while the tasks are being created, entry from the WIP queue, the phase timeout) are described in the changelog.

• Improvement

  A phase's tasks see freshly written data

  Every task reads the case data as it is when the task starts: what a script or an AI or Jev task has just written is immediately available to the next task in the same phase, with or without “Run tasks sequentially”, and for these tasks there is no longer any need to split dependent steps across two phases. The result of an integration, a wait or a manual task, which arrives later, is visible to the next task only with “Run tasks sequentially” and with that task blocking.

• Improvement

  Typed values passed to sub-flows

  A field passed to a sub-flow through a single reference keeps its type: 59.9 stays a number, a yes/no stays a yes/no, an object stays an object; text made of several parts stays text. The whole case data set can be passed as well: restricted fields stay restricted in the sub-flow too. If the sub-flow schema expects a number or a yes/no and the value cannot be converted, the task stops with a clear message instead of creating a child with wrong data.

• Improvement

  One numeric field, consistent with your language

  In Process Map nodes (Value €, Minutes, Runs per month), in simulations and in the Saving Challenge (fee, recurring costs, goals), the numeric field no longer rewrites what you type and no longer misreads the Italian decimal comma: 1.500 is rejected with an explanation (it could mean 1.5 or 1500), while 100,07 is accepted correctly. The value is only saved when it is a valid number; pasting an ambiguous number never saves anything silently. Explicit actions (Create, Save, Simulate, Next…) stay blocked until the field in error returns to a valid value. Score, attempts, percentages and costs that used to be silently corrected when out of range now show the same rejection, keeping the previous value. What-If and ROI simulations now update when you leave the field or press Enter, not on every keystroke; pressing Enter on the API key expiry field no longer generates the key by itself.

• Improvement

  Esc closes one window at a time

  Pressing Esc on a confirmation window no longer also closes the panel underneath it (AI Form, Properties, Flow settings) together with an unsaved change: each press closes one window at a time, from the most recently opened to the oldest — if other windows remain open, it takes as many Esc presses to close them. Also fixed inline Task renaming (Esc now only cancels the rename) and, in the process map Designer, What-If mode: it now only exits when nothing is open above it (a menu, AI Analysis, Automation, the map rename field, a confirmation). In the Flow Agent panel too, Esc only collapses the panel, without also closing whatever is underneath it.

• Improvement

  Jev scores as levels from 1 to N

  A Jev Score question now writes the number of the most likely level, from 1 to N as in the Designer, instead of a decimal value counted from 0: with five levels, an "urgency >= 4" rule matches the two highest levels at any confidence above the threshold. The confidence stays in its own property and Try shows the answer as "Level 4 of 5 · High". Items evaluated earlier keep the value written back then: review the rules that use it.

• Improvement

  Scheduled cases on a suspended flow

  A case scheduled on a flow that is later disabled or archived no longer stays stuck forever: if the flow has a published version, the case starts on it at start time, from its initial phase, and the history says so ("Started on v3"). If there is no published version the case waits unchanged: its detail shows "Waiting for an operational flow" with the way to unblock it, the case list has a dedicated filter, and starting it early returns a clear message instead of skipping the initial phase. Publish and restore messages for a disabled version now point to the right path: "Reactivate", then publish the new version.

• Improvement

  Early start with the real state

  When you start a scheduled case early and, on entering its initial phase, it moves on by itself (for example because the awaited event had already arrived), the move you chose is no longer applied from a different phase: the case stays started where it arrived, the message tells you which phase it is in, and the page refreshes so you can continue from there.

• Security

  IPv6 addresses hiding an internal address

  Address checks for webhooks, connectors and providers now also recognise IPv6 addresses that contain an internal address, however they are written, and reject them just like the internal address itself: when saving a flow, on import, on publish and before every call. Public addresses work as before. A mail server given as a numeric address written in an ambiguous form (for example with leading zeros) is rejected too.