AI SRE persistent memories

AI SRE can save and search persistent memories across chat sessions. Memories are short, self-contained pieces of knowledge that are created during or after conversations, and retrieved via semantic search in future sessions.

AI SRE automatically creates and updates memories when relevant, but you can also explicitly ask it to remember something. Memories are stored on a per-user basis.

Below are some use cases where memories can be helpful:

Use case	Description
Recalling and refining an investigation	Let's say you investigated high CPU on prod-server last week. This week, you can ask "are we still seeing prod-server CPU issues?" AI SRE finds a memory with the prior root cause and skips the exploratory phase. Memories also capture dead ends, such as "ruled out Kafka partition imbalance — issue was compute-bound on encoders", so that AI SRE doesn't re-explore paths already eliminated.
Remembering user-specific context	Across sessions, the agent accumulates context such as "User is on the payments team. When they say 'the gateway' they mean payment-gateway." Future sessions skip generic clarifying questions.
Accumulating domain knowledge	AI SRE learns patterns like "payment-gateway 5xx errors from /health/stripe-sandbox are a known false alarm" and distinguishes known patterns from genuinely new issues.

By default, AI SRE memories are disabled.

Perform the following steps to enable persistent memories for AI SRE:

Hover on your user name in the left navigation, then select Manage account.
Select AI settings.
Slide the toggle in the Memories section to the on position.