Recipes — Valkey & Kvrocks

Both speak the Redis (RESP) protocol, so every Redis client and library works unchanged — these are doze’s cheap, open-source ways to run “Redis” locally. (Valkey is the Linux-Foundation fork that kept Redis open source after its 2024 relicense; Kvrocks is an Apache project with the same API on disk. The full story is in The engines.) The difference is where the data lives:

• Valkey — in-memory, like Redis. Fast, volatile. Use it as a cache.
• Kvrocks — RocksDB-backed (on disk). Redis API, durable storage, lower RAM. Use it as a persistent KV store that survives reaps and restarts.

A cache (Valkey)

valkey "cache" {
  version   = 9            # newest 9.x; or pin "9.1.0"
  maxmemory = "256mb"      # optional cap
  password  = "cache"      # optional
}

The URL is stable — redis://127.0.0.1:6379 (add :cache@ for the password), and connecting cold-boots the instance:

redis-cli -u redis://127.0.0.1:6379 ping              # -> PONG
redis-cli -u redis://127.0.0.1:6379 set greeting hello
redis-cli -u redis://127.0.0.1:6379 get greeting

A durable KV store (Kvrocks)

kvrocks "store" {
  version  = 2
  password = "store"       # optional
}

Identical client experience, but writes persist to disk — stop touching it, let it reap, reconnect later, and your keys are still there.

Connecting clients & GUIs

Find the address and point any redis tool at it (RedisInsight, TablePlus, redis-cli):

doze status
#   NAME    ENGINE   STATE   …   ENDPOINT
#   cache   valkey   idle        127.0.0.1:6433
redis-cli -h 127.0.0.1 -p 6433

If you set a password, pass it with -a (or it’s already in REDIS_URL):

redis-cli -h 127.0.0.1 -p 6433 -a cache

Cache + durable store together

valkey "cache" {
  version   = 9
  maxmemory = "128mb"
}
kvrocks "store" {
  version = 2
}

Each has its own stable URL — they differ only by the explicit port you declared:

redis-cli -u redis://127.0.0.1:6379 set session:42 active   # cache
redis-cli -u redis://127.0.0.1:6380 set user:42  "{...}"    # store

A process block that depends on both gets each one injected under its own env_var.

Common tasks

redis-cli -u redis://127.0.0.1:6379 flushall            # wipe everything
redis-cli -u redis://127.0.0.1:6379 info keyspace       # how many keys
redis-cli -u redis://127.0.0.1:6379 monitor             # watch commands live
doze sleep cache                                         # put it to sleep now

Tips

• Valkey is a drop-in Redis fork — your ioredis/redis-py/go-redis code needs no changes; just point it at the stable redis://127.0.0.1:<port>.
• Pick by durability: ephemeral cache → valkey; data you don’t want to lose on a reap → kvrocks.
• maxmemory (Valkey) caps memory; pair it with an eviction policy from your client if you want LRU behavior.
• Reaping is by connection count — a client that keeps a connection open keeps the instance awake.