Skip to main content
Alerts and notices

Q-Notice categories, templates and audit log

Manage Q-Notice categories and templates, read the audit log of everything that happened to a notice, and understand the AI provider health notices QuantumOps raises.

Written By Chris Scaminaci

Last updated About 2 hours ago

Three parts of the Q-Notice Hub are for administrators: the categories and templates that shape the notices people issue, the audit log that records what happened to each notice, and the notices QuantumOps raises about the health of your AI providers.

Before you start: you need the Administrator or Q-Notice Admin role, and the Q-Notice Engine module must be on for your organisation. Someone without the role who selects the Categories & Templates or Audit Log tab sees a message that says they need the Q-Notice Admin role.

Set up categories

A category groups notices of one type and gives them defaults, so people issue them faster and more consistently. On the Categories & Templates tab, the Categories list shows each category as a card, grouped by notice type and ordered by its sort order. A System label marks a built-in category.

To add a category:

  1. Open the Categories & Templates tab and select Categories.
  2. Select Create Category at the top right of the Hub.
  3. Fill in the form:
    • Name (required) and Description.
    • Notice Type: the type of notice the category belongs to. The wizard offers a category only for its own type.
    • Default Severity: filled in when someone picks the category.
    • Icon CSS Class and Color (hex): how the category looks on its card. The form shows a preview.
    • Default Expiration (minutes): filled in when someone picks the category. Leave it empty for none.
    • Sort Order: lower numbers come first.
    • Category is active: untick it to stop offering the category. It stays in the list, marked Inactive.
    • Default Custom Instructions: text for the AI, used when notices of this category are included in AI processing.
  4. Select Create.

Select Edit on a card to change a category. Select Delete to remove one; the confirmation warns that notices in the category lose their category. A System category cannot be deleted.

When someone issues a notice, the category's severity and expiry are filled in for them and they can change them. If they include the notice in AI processing and leave Inherit category instructions on, the category's AI instructions go to the AI along with the notice's own. See Issuing a Q-Notice.

Review and edit templates

A template is a starting point for a notice you write often. The tab has no button for adding a template; you review and edit the templates in the list. When someone selects it under Load Template in the wizard, it fills in the type, severity, message, AI instructions and expiry, and sets whether the notice is included in AI processing.

The Templates list shows each active template with its name, type, default severity, an LLM badge when it is included in AI processing by default, and a shortened default message. Select Edit to change one:

  • Name (required) and Description.
  • Notice Type and Default Severity.
  • Default Expiration (minutes).
  • Include in LLM by default: new notices from the template are passed to AI analysis unless the person turns it off.
  • Template is active: untick it to stop offering the template.
  • Default Message: the text the wizard fills in.
  • Default Custom Instructions: the AI instructions the wizard fills in.

Select Save Changes. Delete removes a template from the list and from the wizard. A template with the System label is built in and cannot be deleted.

Read the audit log

The Audit Log tab records what happened to every notice, including notices that have expired. It shows the 500 most recent entries that match your filters, 20 to a page, newest first. Select a column heading to sort.

Filter with:

  • Notice Type: All Types, or one type.
  • Action: All Actions, or one of Created, Modified, Expired, Dismissed, Delivered, Delivery failed and Included in LLM.
  • From and To: a range of dates.
  • Clear resets the filters.

Each entry has a Timestamp, the Notice Title and Type, the Action, Performed By, System? and Details.

Performed By shows the account identifier of the person who acted in QuantumOps, or the chat e-mail or display name of the person who clicked a card button. Entries made by rules and by the platform read System, and the System? box is ticked. Entries for the buttons on Slack and Teams cards, such as Acknowledged and Snoozed, name the person who clicked and the channel they clicked in; they appear in the list even though they are not in the Action filter.

The Action column shows the stored name rather than the filter wording, for example DeliveryFailed where the filter reads Delivery failed. An Expired entry's Details say why, for example Manually expired, TimeExpiry, EventTriggered or ConditionMet.

Check a failed delivery

Every time a notice is sent to a channel, the log records Delivered or DeliveryFailed. Filter on Delivery failed to see the failures. The Details column names the event, the channel and what went wrong:

What it saysWhat it means
Channel not configured or send failedSlack or Teams has no incoming webhook set, or the post was refused. Check the Messaging tab; see Q-Director: Slack, Teams and agent mapping.
No recipients configuredE-mail was chosen but no address was available.
Email send failedThe e-mail could not be sent.
Interactive card unavailable; the sentiment webhook card was already sent, so the webhook fallback is skippedThe sentiment alert's own message already went to the channel, and no second message was sent. There is nothing to fix.
Unsupported channelThe notice names a channel that Q-Notices cannot deliver to. This can happen to a rule imported from the legacy Alert Engine.

A notice that reaches only the Hub and the indicators has no delivery entries. See How Q-Notices are delivered for what each channel needs.

AI provider health notices

QuantumOps watches the AI providers it uses for your organisation. When something that only you can fix goes wrong, it raises a Q-Sentinel notice for the System, named AI provider health. It sends the notice once to Slack and Teams where those are set up for your organisation, and by e-mail to your organisation's primary account address. It also shows in the Hub. These notices are for people to read; they are not passed to ticket analysis.

Each one is raised once when the condition is found, not once per failed call, so one notice can stand for hundreds of failures. You do not have to dismiss them. Each one expires on its own, and a new one appears only if the condition lasts or comes back.

NoticeWhat it meansSeverityStays visible
A provider refuses your accountA provider keeps rejecting calls for a reason only you can fix. The title starts with the provider's name and continues is refusing this tenant's account, then gives the cause: no credit left, no API key configured, API key not accepted, or the selected model cannot be used (it is missing, unreachable, or answers only in streaming mode). The message also says whether another provider is absorbing the failures or nothing is being served.Alert26 hours. It repeats at most once a day while the condition lasts.
Configured models are replaced by a fallbackA sustained share of calls did not run on the model you configured and went to a fallback instead. The title starts AI fallback at, then the share of calls, then configured models are being silently replaced. Work continues, but at the fallback's cost and quality. The message lists the most frequent substitutions.Alert12 hours
AI is unavailableThe title reads AI is completely unavailable: every provider tried is failing. No call has succeeded on any provider and nothing is left to take over, so AI-dependent processing is producing nothing.Critical2 hours

The fix is never on the notice itself. Top up or fix the account in the provider's own billing, or change the API key or the model in your organisation's tenant settings. See AI providers and models and Tenant settings. QuantumOps does not pause processing for you while a provider is failing; to pause it yourself, see Pausing and resuming AI processing.