Struct ConversationSummarizer 

pub struct ConversationSummarizer { /* private fields */ }

Generates and caches conversation summaries for episodic memory.

Uses LLM to create concise summaries of conversation history. Implements incremental summarization - only new messages since the last summary are sent to the LLM, merged with the previous summary.

Implementations

impl ConversationSummarizer

pub fn new() -> Self

Create a new summarizer with default settings.

pub fn with_threshold(self, n: usize) -> Self

Set the regeneration threshold (number of new messages before re-summarizing).

pub fn with_max_length(self, n: usize) -> Self

Set the maximum summary length in characters.

pub fn with_exclude_last(self, n: usize) -> Self

Set how many recent messages to exclude from summarization.

pub fn needs_regeneration(&self, current_message_count: usize) -> bool

Check if summary needs regeneration based on message count delta.

pub fn get_cached_summary(&self) -> Option<&str>

Get cached summary if available.

pub fn generate_summary<F>( &mut self, messages: &[Value], llm_caller: F, ) -> String

where F: Fn(&[Value]) -> Option<String>,

Generate summary of conversation history using incremental approach.

Only sends new messages since the last summary to the LLM, along with the previous summary for context merging.

The llm_caller receives a slice of message values (system + user prompt) and should return Some(summary_text) on success.

pub fn format_conversation(messages: &[Value]) -> String

Format messages into readable conversation text.

pub fn clear_cache(&mut self)

Clear the cached summary.

pub fn to_json(&self) -> Option<Value>

Serialize cache for session persistence.

pub fn from_json(&mut self, data: Option<&Value>)

Load cache from session persistence.

Trait Implementations

impl Default for ConversationSummarizer

fn default() -> Self

Returns the “default value” for a type.