Struct SmartRetrievalConfig 

pub struct SmartRetrievalConfig {
    pub max_retrieved_tokens: usize,
    pub importance_threshold: f32,
    pub chunk_contiguous_messages: bool,
    pub enabled: bool,
}

Configuration for smart retrieval

Fields

max_retrieved_tokens: usize

Maximum tokens for retrieved historical context (excludes current messages)

importance_threshold: f32

Minimum importance score to include a message (0.0-1.0)

chunk_contiguous_messages: bool

Group contiguous messages into chunks for better llama.cpp caching

enabled: bool

Enable smart retrieval (can be disabled to fall back to original behavior)

Implementations

impl SmartRetrievalConfig

pub fn from_ctx_size(ctx_size: u32) -> Self

Derive the historical retrieval budget from the model’s context window. 25% of CTX_SIZE is allocated to retrieved history (summaries + cold SQLite), ensuring the current conversation always gets the lion’s share.

Trait Implementations

impl Clone for SmartRetrievalConfig

fn clone(&self) -> SmartRetrievalConfig

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for SmartRetrievalConfig

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for SmartRetrievalConfig

fn default() -> Self

Returns the “default value” for a type.