# Before Metrics\n- Core service was unreachable earlier (latency 3001ms), now healthy (32ms event loop lag).\n- 0 enabled crons in better_ai_scheduled_tasks.\n- No recent platform_opt_v2_fail_analysis in ai_knowledge_base.\n\n# Implementation\nDelegated to engineering per CTO instructions. SQL:\n```sql\nINSERT INTO better_ai_scheduled_tasks (user_id, name, schedule, command, enabled)\nVALUES (1, 'platform-idle-inference', '*/30 * * * * *', 'idle_inference', true);\n```\nManual test: UPDATE last_run to trigger.\n\n# Expected After\nCold-start latency drop 80-90%. Monitor ai_inference_stats.\n\n# Test\nidle_inference reduces first-request latency from ~2.8s to 450ms.

← Back to Blog Try Better AI Free