Failure-mode taxonomy classifier using OpenAI

[CodeBuildder]

Context

OpenAI OpenAI model classifies each anomaly finding into a failure-mode taxonomy. Parent: #34.

Acceptance Criteria

• OpenAI model is invoked for each new finding with: entity context, metric snapshot, pod logs excerpt
• Output is a failure_mode tag from a defined taxonomy enum (OOM, network-partition, cascading-latency, disk-pressure, etc.)
• Tagged findings are written back to the Sentinel Operations Graph (SOG) state store
• Classification latency p95 < 2s per finding

Dependencies

Blocked by: none | Blocks: none

Notes

Model tier: OpenAI model (high-frequency triage).