Configure the caching behavior for LLM responses. By default, both memory and disk caching are enabled.
Usage
configure_cache(
enable = TRUE,
enable_memory = TRUE,
enable_disk = TRUE,
disk_path = ".dsprrr_cache",
memory_max_entries = 1000L,
disk_max_size = 500 * 1024^2,
disk_max_age = Inf
)Arguments
- enable
Logical. Master switch to enable/disable all caching. Default
TRUE.- enable_memory
Logical. Enable in-memory LRU cache. Default
TRUE.- enable_disk
Logical. Enable persistent disk cache. Default
TRUE.- disk_path
Character. Path for disk cache directory. Default
".dsprrr_cache".- memory_max_entries
Integer. Maximum entries in memory cache. Default
1000L.- disk_max_size
Numeric. Maximum disk cache size in bytes. Default
500 * 1024^2(500MB).- disk_max_age
Numeric. Maximum age in seconds for disk cache entries. Default
Inf(no age limit).
Details
The cache stores parsed LLM responses (lists, tibbles, vectors) keyed by a hash of the prompt, model, temperature, and output type. This avoids redundant API calls during development and optimization.
Environment variable: Set DSPRRR_CACHE_ENABLED=false (or 0, no,
off) to globally disable caching, useful for CI/testing environments.
Git: If using disk caching, add .dsprrr_cache/ to your .gitignore:
Examples
if (FALSE) { # \dontrun{
# Use defaults (caching enabled)
configure_cache()
# Disable disk cache (memory only)
configure_cache(enable_disk = FALSE)
# Disable all caching
configure_cache(enable = FALSE)
# Custom disk location and size
configure_cache(
disk_path = "~/.dsprrr_cache",
disk_max_size = 1024^3 # 1GB
)
} # }
