AI Gateway Cache Ops

Admin Console /admin/cache writes enterprise/.runtime/admin/cache-config.json, then calls the gateway /internal/cache/reload after save.

AI Gateway Cache Ops

Feature flags

Env var	Default	Description
`GATEWAY_CACHE_L1`	`on`	L1 exact cache
`GATEWAY_CACHE_L2`	`off`	L2 semantic cache
`GATEWAY_CACHE_L1_TTL`	`5m`	L1 TTL
`GATEWAY_CACHE_SEMANTIC_THRESHOLD`	`0.92`	L2 similarity threshold
`GATEWAY_CACHE_REPLAY_MODE`	`burst`	Streaming replay mode
`REDIS_URL`	empty	When set, L1 uses Redis

Admin Console /admin/cache writes enterprise/.runtime/admin/cache-config.json, then calls the gateway /internal/cache/reload after save.

Eviction

bash

1curl -X POST "$GATEWAY/internal/cache/evict" \
2  -H "Authorization: Bearer $GATEWAY_INTERNAL_TOKEN" \
3  -H 'content-type: application/json' \
4  -d '{"prefix":"<key-hash-prefix>"}'

Fallback

Set GATEWAY_CACHE_L1=off and GATEWAY_CACHE_L2=off, then restart the gateway. Behavior matches cache disabled.

Made-with: Damon Li