Struct PolicyContext 

pub struct PolicyContext<'a> {
    pub turn_index: usize,
    pub accumulated_usage: &'a Usage,
    pub accumulated_cost: &'a Cost,
    pub message_count: usize,
    pub overflow_signal: bool,
    pub new_messages: &'a [AgentMessage],
    pub state: &'a SessionState,
}

Shared read-only context available to every policy evaluation.

Fields

turn_index: usize

Zero-based index of the current/completed turn.

accumulated_usage: &'a Usage

Accumulated token usage across all turns.

accumulated_cost: &'a Cost

Accumulated cost across all turns.

message_count: usize

Number of messages in context.

overflow_signal: bool

Whether context overflow was signaled.

new_messages: &'a [AgentMessage]

Messages added since the last policy evaluation for this slot.

• PreTurn: user/pending messages appended since the previous turn.
• PostTurn / PostLoop: empty — current-turn data is in TurnPolicyContext.

Policies should only scan this slice, never the full session history, to avoid redundant work on messages that have already been evaluated.

state: &'a SessionState

Read-only access to the session state.

Trait Implementations

impl<'a> Debug for PolicyContext<'a>

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

Formats the value using the given formatter.