server:
  port: "8080"
  host: "0.0.0.0"
  mode: "debug" # debug, release, test
  splitwise: true # true means pd mode, false means mixed mode

scheduler:
  policy: "power_of_two"
  prefill-policy: "cache_aware"
  decode-policy: "request_num"
  eviction-interval-secs: 60
  eviction-duration-mins: 30 # eviction duration for cache-aware radix tree nodes (minutes); default: 30
  balance-abs-threshold: 1
  balance-rel-threshold: 0.2
  hit-ratio-weight: 1.0
  load-balance-weight: 0.05
  cache-block-size: 4
  waiting-weight: 10
  stats-interval-secs: 5 # interval in seconds for periodic stats logging (running requests, cache hit rate)

manager:
  health-failure-threshold: 3
  health-success-threshold: 2
  health-check-timeout-secs: 5
  health-check-interval-secs: 5
  health-check-endpoint: /health
  register-path: "config/register.yaml"

log:
  level: "info"  # debug, info, warn, error
  output: "file" # stdout, file