AI Enrichment & Chat

Enrichment

When a new incident is created, DataAgent automatically runs an enrichment pipeline using Claude (via AWS Bedrock). Enrichment typically completes within a few seconds of the incident appearing.

The AI enrichment includes:

• Root cause explanation — a plain-English description of what went wrong and why
• Pod metrics summary — restart count, OOM events, last error reason at time of detection
• Node conditions — memory, disk, and CPU state on the node hosting the affected workload
• Related incidents — other incidents involving the same resource or namespace

The enriched explanation appears in the incident detail view above the chat interface.

Chat

Every incident has a built-in chat interface. You can ask questions about the incident and the AI responds with context-aware answers drawn from:

• The incident’s detection data and metrics
• Historical incidents for the same resource
• The cluster’s current state

Example questions

• “What is the most likely root cause of this crash loop?”
• “Has this pod OOM-killed before? If so, how frequently?”
• “What would you check first to resolve this?”
• “Is this related to the node pressure incident in the same namespace?”

Conversation history

Chat history is stored per incident and visible to all team members. You can clear the conversation using the Clear chat button if needed.

Token budgets

AI enrichment and chat consume tokens from your organisation’s budget. When the budget is exhausted, enrichment and chat are paused until the next budget cycle resets.

See AI token budgets for details.

Limitations

• The AI has access to the data DataAgent has collected — it cannot query your cluster directly.
• The AI cannot make changes to your cluster resources.
• Enrichment quality depends on the metrics and detection data available at the time. Transient conditions that resolve quickly may produce limited context.