chunk_size: 16 # the "golden spot" for high throughput transfers for redis is ~4 MB (any higher or lower will cause performance degradation, for a model like meta-llama/Llama-3.1-8B-Instruct, this is around 16 tokens)
local_cpu: false
max_local_cpu_size: 20.0
remote_url: "resp://localhost:6379"
async_loading: false # LMCache RESP also support async loading
remote_serde: "naive"
blocking_timeout_secs: 120
extra_config:
  resp_num_threads: 8 # default is 8 if not specified
  save_chunk_meta: False # make sure we have fixed size payloads
  # Authentication: prefer environment variables to keep secrets out of logs.
  #   export LMCACHE_RESP_USERNAME="default"
  #   export LMCACHE_RESP_PASSWORD="YOUR_REDIS_PASSWORD"
  # Alternatively, set them here (they will appear in logged config):
  # username: "default"
  # password: "YOUR_REDIS_PASSWORD"