2026-02-04: Apdex SLO violation for inference_vertex in ai-gateway

Apdex SLO violation for inference_vertex in ai-gateway (Severity 4)

Problem: Request performance for non-streaming inferences to Vertex AI in ai-gateway dropped below the required threshold.

Impact: Non-streaming inference requests to Vertex AI through ai-gateway were slower than expected, resulting in an apdex score of 84.24%, which is below the target for acceptable performance.

Causes: LiteLLM v1.79.1, the library used to make Vertex AI requests, was not respecting configured timeout values.

Response strategy: A merge request has been drafted to upgrade LiteLLM to version 1.81.0, which should resolve the timeout handling issue.

This ticket was created to track INC-7107, by incident.io 🔥