Struct PrunRequest 

pub struct PrunRequest {
    pub tokens_to_remove: usize,
    pub memo: Option<String>,
}

A pending prun request the LLM submitted via prun or prun_with_memo.

Lifecycle:

1. PrunTool::execute() pushes one of these onto the shared pending queue.
2. The agent loop drains the queue between turns (see agent_loop/run.rs:424).
3. Each request is applied to AgentContext.inrun_context in submission order, producing a PrunRecord event that the session recorder captures.

tokens_to_remove is an upper bound — the loop walks the tail of inrun_context removing whole entries until at least this many tokens have been freed. User messages are never affected (they live in the separate user_context stream).

Fields

tokens_to_remove: usize

Lower bound on tokens to remove from the tail of inrun_context. The loop rounds up to the nearest whole entry so a single message is never split.

memo: Option<String>

Optional summary inserted in place of pruned content. Some for the prun_with_memo variant; None for the silent prun variant.

Trait Implementations

impl Clone for PrunRequest

fn clone(&self) -> PrunRequest

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for PrunRequest

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

Formats the value using the given formatter.