Common issues

Most problems in SPARK come from one of a few predictable causes: you are in the wrong workspace or organization, an agent was not named, a workflow has not been activated, something is waiting on a human, an extension is off, or a limit has been reached. This page walks through the situations users hit most, with the symptom, the reason behind it, and the fix.

I can't see my data, or a colleague's data

Each workspace is an isolated tenant. Data, tasks, workflows, threads and agents live inside one workspace and are never visible from another. If a table or record you expect is missing, you are almost always looking at the wrong workspace.

Click the workspace name at the top of the sidebar to open the workspace menu, and confirm the name matches where the data was created. The organization you are signed in to is named at the top of that menu; if you belong to several, the data may be in another one, which Switch organisation reaches. If a colleague can see something you cannot, you may not be a member of that workspace yet: ask an admin to add you (Members).

I can't find a page that used to be in the sidebar

The sidebar has three tabs: Threads, Ecosystem and Settings. What was Home is now Ecosystem (⌘⇧E), and a workspace opens on the New Thread page; chats are now threads, on the Threads tab (⌘⇧C). On the Ecosystem tab, Workflows, Data and Files sit under More, and agent configuration is under Studio. In Settings, Managed SFTP moved to Extensions, as SFTP, beside the Mailbox; the workspace's pages are under Workspace, and Workspaces is under Organisation. Old links to a moved page still open it. When in doubt, press ⌘F (Ctrl+F) and type the page's name. See Navigating the workspace and Search.

An agent doesn't answer

An agent answers only a message that names it. In a channel or a thread, type @ and the agent's name; on the New Thread page, a message that names no agent goes to Spark. In your own direct message with an agent, it answers every message without being named. See Start a thread.

If the agent you want is not offered at all, an administrator may have switched it off: the built-in agents are switched on and off on Agent Studio's Agents page. See The Spark agent.

My old chats are gone, or a thread says I don't have access

Chats became threads. Each chat moved into the direct message between the person who started it and the agent it was with, so it is still there, under Direct Messages on the Threads tab. People who had only written in someone else's chat no longer see it.

A thread you cannot open reads You don't have access to this thread: It's private to the people in it, or it was deleted. Ask the person who started it to move it to a channel you're in. Its starter can move it with Move thread; see Move a thread.

A button says Requires admin role

Some actions are an administrator's. Creating a channel (New channel, New private channel), renaming, archiving or managing the members of a channel, and switching a built-in agent on or off are greyed for anyone else, and hovering them reads Requires admin role. Ask an administrator of your workspace, or ask for a role that allows it. See Channels and Teams & roles.

The Mailbox or Managed SFTP is off

The Mailbox and Managed SFTP are extensions: each workspace switches them on in Settings, under Extensions.

• The Mailbox is off for this workspace: click Activate to give the workspace its address. See Mailbox.
• The Mailbox isn't available in this environment. Contact support to enable it.: the SPARK installation you use has no email set up for the Mailbox, which only the SPARK team can do. See Getting support.
• Managed SFTP is off for this workspace: click Activate; its users and their SSH keys come back as they were. See Managed SFTP.

My workflow isn't running

A newly created workflow is saved as a draft. Drafts do not execute, and their triggers stay dormant until you turn them on.

Open the workflow and activate it from the editor's toolbar. After activating, confirm the trigger matches what you expect (for example a schedule or a file event). Only active workflows respond to triggers or run on a schedule. A schedule that keeps failing waits longer between attempts and then pauses itself: on the Schedulers tab its Status reads Paused, and hovering it gives the reason. Fix the cause, then turn the scheduler back on from its ⋯. See Scheduling. If the workflow's runs were refused, your organization may have used its workflow runs for the month or the hour; see I hit a limit.

My workflow is stuck partway through

If a run started but never finished, it may be paused on a Human Review step, waiting for someone to approve or reject before it can continue.

On the Ecosystem tab, open Workflows and the Interventions tab to see runs awaiting a decision, then approve or reject the pending step. See Human in the loop for how these pauses work, and Runs and debugging if the run failed for another reason.

The agent doesn't know about my data

An agent answers from the context of the workspace it runs in. If it cannot find something, that data usually does not exist in this workspace, or you are working in a different one.

Confirm the data object and records exist in the current workspace, then start a new thread. You can also add the item to the thread's context so the agent works from it directly, or start the thread from the item's page with the new thread button in the page header. See Start a thread.

The agent stopped partway through an answer

An agent's turn ends when your organization has used its NGT tokens for the hour or the month, with a message saying when to try again. A turn can also end because it ran out of steps, or because someone pressed Stop. Check the message at the end of the reply, then see I hit a limit if it names a limit.

I hit a limit

When something would pass one of your organization's limits, SPARK refuses it with a Limit reached message that says what ran out and when to try again, with a Contact support button. The messages read, for example:

• Your organization has used all of its NGT tokens for this hour. Try again at …, or contact support for a higher limit. (or for this month)
• Your organization has started all of its workflow runs for this hour. Try again at …, or contact support for a higher limit. (or for this month)
• Your organization has used all of its document extractions for this hour. Try again at …, or contact support for a higher limit. (the same for extractions with OCR)
• Your organization is using all of its seats. Contact support for a higher limit.
• Your organization's storage is full. Contact support for a higher limit.

Hourly limits reset at the top of the hour and monthly limits on the 1st; seats and storage free up when you remove people or data. Open Settings, then Limits & Quotas under Organisation, to see every limit, how much is used and when it resets. To ask for more, click Request increase. See Limits & Quotas.

I can't invite someone

If the invitation is refused with a Limit reached message, every seat in your organization is taken; remove someone, or revoke an invitation nobody answered, to free one. Also check that you picked a Team: the Invite button stays disabled until both an email and a team are set. See Members.

A CSV import mapped columns wrong

During import, SPARK infers an attribute type for each column and matches it to a header. A misread date or number usually means the source was inconsistent.

Review the column mapping and attribute types before confirming the import. If it already landed, correct the attribute types on the data object, then re-import a clean file with one row per record and consistent formats.

I can't create a data object

Data objects are created from the Data page (on the Ecosystem tab, under More) with New, and their structure is managed in Settings, under Features, on the Data Objects page. If New is missing, it is most likely a permission your role doesn't grant.

Where to go next

• Key concepts glossary
• Getting support