Struct CompactionConfig 

pub struct CompactionConfig {
    pub threshold_tokens: usize,
    pub retain_recent: usize,
    pub min_messages_for_compaction: usize,
    pub auto_compact: bool,
    pub max_retained_tail_tokens: usize,
    pub summary_max_tokens: usize,
}

Configuration for context compaction.

Controls when and how context compaction occurs.

Example

use agent_sdk::context::CompactionConfig;

let config = CompactionConfig::default()
    .with_threshold_tokens(100_000)
    .with_retain_recent(20);

Fields

threshold_tokens: usize

Token threshold to trigger compaction. When estimated tokens exceed this, compaction is triggered. Default: 80,000 (conservative for 128K context models)

retain_recent: usize

Number of recent messages to keep intact (not summarized). These messages remain in full to preserve immediate context.

The retained tail is additionally bounded by max_retained_tail_tokens: if keeping retain_recent messages would exceed that token budget, the oldest of those messages are summarized instead so the tail stays within the cap. Default: 10

min_messages_for_compaction: usize

Minimum messages before compaction is considered. Prevents compaction when conversation is still short. Default: 20

auto_compact: bool

Whether to automatically compact when threshold is reached. If false, compaction only occurs on explicit request. Default: true

max_retained_tail_tokens: usize

Soft cap on the estimated tokens kept in the retained tail.

Bounds retain_recent: the most recent messages are kept only until this token budget is reached, after which older messages are folded into the summary instead. The cap is soft because an indivisible tool_use/tool_result pair may push the tail slightly over. Raise it on large-context models, or lower it for more aggressive compaction. Default: 20,000

summary_max_tokens: usize

max_tokens budget for the LLM summarization call.

If the summarizer hits this ceiling (a MaxTokens stop reason), the compactor logs a warning and retries once with a larger budget before marking the summary as truncated. Default: 4,096

Implementations

impl CompactionConfig

pub fn new() -> Self

Create a new configuration with default values.

pub const fn with_threshold_tokens(self, threshold: usize) -> Self

Set the token threshold for compaction.

pub const fn with_retain_recent(self, count: usize) -> Self

Set the number of recent messages to retain.

pub const fn with_min_messages(self, count: usize) -> Self

Set the minimum messages for compaction.

pub const fn with_auto_compact(self, auto: bool) -> Self

Set whether to auto-compact.

pub const fn with_max_retained_tail_tokens(self, tokens: usize) -> Self

Set the soft cap on tokens kept in the retained tail.

Bounds retain_recent. See the field docs for the precise semantics.

pub const fn with_summary_max_tokens(self, tokens: usize) -> Self

Set the max_tokens budget for the summarization call.

Trait Implementations

impl Clone for CompactionConfig

fn clone(&self) -> CompactionConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for CompactionConfig

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

Formats the value using the given formatter.

impl Default for CompactionConfig

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for CompactionConfig

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for CompactionConfig

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.