Skip to main content

AI Memory

What is AI Memory​

AI Memory is OpsWorker's persistent context layer. It captures facts about your environment and preferences for how the AI should behave, then loads them automatically during investigations and AI Chat sessions.

Without memory, the AI starts from zero with every alert and every question. With memory, OpsWorker builds up understanding of your specific environment over time (ownership, deployment rules, known quirks, escalation policies) and applies it to every interaction.

Layered Scope​

Memory entries are scoped to one of three levels:

ScopeVisible ToUse For
Personal (My Memory)Only youYour preferences, personal notes about a cluster
ClusterEveryone with cluster accessCluster-specific facts (ownership, quirks, rules)
OrganizationEveryone in the orgCompany-wide policies, escalation rules, standards

Facts can be promoted upward, from personal to cluster and from cluster to organization. Promotion copies the entry (the original is kept) and is gated by role: personal to cluster requires the Admin or ReadOnly role, and cluster to organization requires the Admin role. Preferences are personal only.

Entry Types​

TypeWhat It IsExample
FactSomething objectively true"Team B owns all pay-* services"
PreferenceHow the AI should behave for you"Always include the exact kubectl command"
SummaryAuto-generated session summaryCreated automatically from chat sessions

You create facts and preferences directly. Summaries are generated automatically.

How It's Used​

OpsWorker loads relevant memory at the start of every investigation and chat session. Investigations load shared cluster facts and organization facts. AI Chat additionally loads your personal notes, and can both recall and store memory mid-conversation (for example, when you ask it to remember a fact or a preference).

The AI factors these entries into its reasoning the same way it factors in live cluster data, as context for interpreting what it sees.

Next Steps​