Struct ResponseCache 

pub struct ResponseCache { /* private fields */ }

Two-tier response cache: in-memory LRU (hot) + SQLite (warm).

The hot cache avoids SQLite round-trips for frequently repeated prompts. On miss from hot cache, falls through to SQLite. On hit from SQLite, the entry is promoted to the hot cache.

Implementations

impl ResponseCache

pub fn new( workspace_dir: &Path, ttl_minutes: u32, max_entries: usize, ) -> Result<Self>

Open (or create) the response cache database.

pub fn with_hot_cache( workspace_dir: &Path, ttl_minutes: u32, max_entries: usize, hot_max_entries: usize, ) -> Result<Self>

Open (or create) the response cache database with a custom hot cache size.

pub fn cache_key( model: &str, system_prompt: Option<&str>, user_prompt: &str, ) -> String

Build a deterministic cache key from model + system prompt + user prompt.

pub fn get(&self, key: &str) -> Result<Option<String>>

Look up a cached response. Returns None on miss or expired entry.

Two-tier lookup: checks the in-memory hot cache first, then falls through to SQLite. On a SQLite hit the entry is promoted to hot cache.

pub fn put( &self, key: &str, model: &str, response: &str, token_count: u32, ) -> Result<()>

Store a response in the cache (both hot and warm tiers).

pub fn stats(&self) -> Result<(usize, u64, u64)>

Return cache statistics: (total_entries, total_hits, total_tokens_saved).

pub fn clear(&self) -> Result<usize>

Wipe the entire cache (useful for construct cache clear).