Expand description
Success criteria from the epic.
Constantsยง
- COLD_
START_ MAX_ MS - Cold start must be under 2 seconds.
- MEMORY_
MAX_ MB - Memory usage must be under 300MB per model.
- QUALITY_
MIN_ RATIO - Quality must be at least 80% of baseline (MiniLM).
- WARM_
P99_ MAX_ MS - Warm p99 latency must be under 250ms.