Struct SummarizationStrategy 

pub struct SummarizationStrategy<P: Provider> { /* private fields */ }

Summarizes old messages using an LLM provider, preserving recent messages verbatim.

When compaction is triggered, messages older than preserve_recent are sent to the provider with a summarization prompt. The response replaces the old messages with a single User message containing the summary, followed by the preserved recent messages.

Example

use neuron_context::SummarizationStrategy;

let strategy = SummarizationStrategy::new(provider, 5, 100_000);

Implementations

impl<P: Provider> SummarizationStrategy<P>

pub fn new(provider: P, preserve_recent: usize, max_tokens: usize) -> Self

Creates a new SummarizationStrategy.

Arguments

• provider — the LLM provider used for summarization
• preserve_recent — number of most-recent messages to keep verbatim
• max_tokens — token threshold above which compaction is triggered

pub fn with_counter( provider: P, preserve_recent: usize, max_tokens: usize, counter: TokenCounter, ) -> Self

Creates a new SummarizationStrategy with a custom TokenCounter.

Trait Implementations

impl<P: Provider> ContextStrategy for SummarizationStrategy<P>

fn should_compact(&self, messages: &[Message], token_count: usize) -> bool

Whether compaction should be triggered given the current messages and token count.

fn compact( &self, messages: Vec<Message>, ) -> impl Future<Output = Result<Vec<Message>, ContextError>> + WasmCompatSend

Compact the message list to reduce token usage.

fn token_estimate(&self, messages: &[Message]) -> usize

Estimate the token count for a list of messages.