Module probe 

The socket’s own bound on asking the core for a fresh context reading, the same two numbers a view bounds its ask with: a probe is answered inline over the CLI’s whole transcript, so it is worth asking for only when the reading it would replace is old enough and the transcript small enough to pay for it.

Re-exports

use std::collections::HashMap; 🔒

use std::time::Duration; 🔒

use std::time::Instant; 🔒

use forge_primitives::SessionSlot; 🔒

use crate::surface::inspector::ContextUsage; 🔒

Structs

ContextProbe

The socket’s bound on asking, held for the life of the fold.

Enums

Declined

Why the socket did not ask a seat a page shows, so a reading that has stopped moving is answered from the log rather than from a repro.

Constants

MIN_INTERVAL 🔒

How long a seat’s reading stands before the socket asks for another, the same bound a view puts on its own ask.

TOKEN_GATE 🔒

Used tokens at or above which the socket does not ask at all, the same gate a view puts on its own ask: only a [1m]-class window reaches this size, where the CLI’s walk over the transcript can exceed the hook timeout.

Functions

used_tokens 🔒

The tokens reading implies have been used, None when the seat reports only one half of the sum - a seat that cannot be measured, which is asked rather than gated.