Enum SessionUpdateEvent 

pub enum SessionUpdateEvent {
    AgentMessageChunk {
        session_id: String,
        text: String,
    },
    AgentThoughtChunk {
        session_id: String,
        text: String,
    },
    ToolCall {
        session_id: String,
        tool_call_id: String,
        name: String,
        raw_input: Value,
    },
    ToolResult {
        session_id: String,
        tool_call_id: String,
        name: String,
        raw_output: String,
    },
    ApprovalRequest {
        session_id: String,
        request_id: String,
        tool_name: String,
        arguments_summary: String,
        timeout_secs: u64,
    },
    ContextUsage {
        session_id: String,
        input_tokens: Option<u64>,
        input_tokens_source: Option<TokenCountSource>,
        max_context_tokens: Option<u64>,
        model_context_window: Option<u64>,
    },
    Plan {
        session_id: String,
        entries: Vec<PlanEntry>,
    },
    TurnComplete {
        session_id: String,
        outcome: TurnCompletionOutcome,
        content: String,
        client_turn_generation: Option<u64>,
        message_count: Option<usize>,
        safeguard_fallback: Option<SafeguardFallbackWire>,
        usage: Option<Box<TurnUsageTotals>>,
    },
    BudgetWarning {
        session_id: String,
        current_usd: f64,
        limit_usd: f64,
        period: BudgetWarningPeriod,
    },
    HistoryTrimmed {
        session_id: String,
        dropped_messages: usize,
        dropped_turns: usize,
        kept_turns: usize,
        reason: String,
        token_budget: Option<u64>,
        tokens_before: Option<u64>,
        tokens_after: Option<u64>,
        tokens_before_source: Option<TokenCountSource>,
        tokens_after_source: Option<TokenCountSource>,
        unsatisfiable_floor: Option<bool>,
    },
}

Typed session update events pushed via session/update notifications. Replaces the hand-built notification_for_turn_event function.

Variants

AgentMessageChunk

Fields

session_id: String

text: String

AgentThoughtChunk

Fields

session_id: String

text: String

ToolCall

Fields

session_id: String

tool_call_id: String

name: String

raw_input: Value

ToolResult

Fields

session_id: String

tool_call_id: String

name: String

raw_output: String

ApprovalRequest

Fields

session_id: String

request_id: String

tool_name: String

arguments_summary: String

timeout_secs: u64

ContextUsage

Per-LLM-call token usage. input_tokens is the cumulative context size for this turn. max_context_tokens is the preemptive-trim budget (the resolved effective_context_budget), preserving its original meaning as the value the meter fills toward. model_context_window is the model’s full context window (provider context_window), exposed distinctly so a client can render capacity and budget separately. Any may be absent when the provider doesn’t report usage or the value can’t be resolved.

input_tokens_source states where input_tokens came from. It is provider for a provider-reported count and estimate when the provider omitted usage and the runtime substituted a local estimate so the meter still has a numerator. Clients should render an estimated value as approximate. It is absent when input_tokens is absent.

Fields

session_id: String

input_tokens: Option<u64>

input_tokens_source: Option<TokenCountSource>

max_context_tokens: Option<u64>

model_context_window: Option<u64>

Plan

Emitted when the TodoWrite tool produces a plan. The entries array carries the normalized PlanEntry values (content, status, priority, optional activeForm) so the client can render the live tracker.

Fields

session_id: String

entries: Vec<PlanEntry>

TurnComplete

Terminal event for a turn. Replaces the response of session/prompt. outcome distinguishes a clean finish from a user-initiated cancel.

Fields

session_id: String

outcome: TurnCompletionOutcome

content: String

Final assistant text (Completed) or partial accumulated text at cancel point (Cancelled).

client_turn_generation: Option<u64>

Optional client-local turn identity, echoed from session/prompt. Absent for legacy callers that do not send one.

message_count: Option<usize>

Authoritative projected conversation-entry count after this turn. Absent for legacy or missing-session terminal events.

safeguard_fallback: Option<SafeguardFallbackWire>

Structured safeguard-fallback attribution for a completed turn. content keeps its rendered footer for clients that predate this field. Only model names cross the wire, never the classifier category.

usage: Option<Box<TurnUsageTotals>>

Token, cost, and context totals for the turn. Absent when the turn never reached a model call (refusals, missing sessions).

BudgetWarning

Recorded spend reached cost.warn_at_percent of a shared daily or monthly limit. Non-blocking and at most once per turn; the turn continues. period is "day" or "month".

Fields

session_id: String

current_usd: f64

limit_usd: f64

period: BudgetWarningPeriod

HistoryTrimmed

Emitted whenever older whole turns were dropped from structured history to fit a token budget or message cap. Surfaces a user-visible “context was cut here” marker so trimming is never silent. dropped_messages is the count of conversation messages removed; dropped_turns and kept_turns describe the user-facing whole-turn accounting.

Fields

session_id: String

dropped_messages: usize

dropped_turns: usize

kept_turns: usize

reason: String

token_budget: Option<u64>

Configured context token budget in effect at trim time. None for message-limit trims, which carry no token accounting.

tokens_before: Option<u64>

Token count before trimming.

tokens_after: Option<u64>

Token count after trimming.

tokens_before_source: Option<TokenCountSource>

Provenance of tokens_before (“provider”, “estimate”, “calibrated”).

tokens_after_source: Option<TokenCountSource>

Provenance of tokens_after (“provider”, “estimate”, “calibrated”).

unsatisfiable_floor: Option<bool>

The retained provider-facing request cannot be brought under the configured budget (protected newest turn plus schemas). History MAY have been trimmed on the way to that floor, so this flag — not dropped_messages == 0 — is the authoritative “unsatisfiable” signal. Absent for ordinary trims.

Trait Implementations

impl Clone for SessionUpdateEvent

fn clone(&self) -> SessionUpdateEvent

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for SessionUpdateEvent

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for SessionUpdateEvent

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for SessionUpdateEvent

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

where __S: Serializer,

Serialize this value into the given Serde serializer.