Struct AgentContinuation 

pub struct AgentContinuation {
    pub thread_id: ThreadId,
    pub turn: usize,
    pub total_usage: TokenUsage,
    pub turn_usage: TokenUsage,
    pub pending_tool_calls: Vec<PendingToolCallInfo>,
    pub awaiting_index: usize,
    pub completed_results: Vec<(String, ToolResult)>,
    pub state: AgentState,
    pub response_id: Option<String>,
    pub stop_reason: Option<StopReason>,
    pub response_content: Vec<ContentBlock>,
}

Continuation state that allows resuming the agent loop.

This contains all the internal state needed to continue execution after receiving a confirmation decision. Pass this back when resuming.

Turn-summary fields

response_id and stop_reason capture the turn-closing LLM call that produced AgentContinuation::pending_tool_calls before the pause. They are carried across the pause boundary so the TurnSummary emitted on the resume path can report the same LLM metadata as the pre-pause summary for the same turn.

Both are Option and default to None for forward compatibility with continuations persisted before these fields existed.

Fields

thread_id: ThreadId

Thread ID (used for validation on resume)

turn: usize

Current turn number

total_usage: TokenUsage

Total token usage so far

turn_usage: TokenUsage

Token usage for this specific turn (from the LLM call that generated tool calls)

pending_tool_calls: Vec<PendingToolCallInfo>

All pending tool calls from this turn

awaiting_index: usize

Index of the tool call awaiting confirmation

completed_results: Vec<(String, ToolResult)>

Tool results already collected (for tools before the awaiting one)

state: AgentState

Agent state snapshot

response_id: Option<String>

Provider response ID from the LLM call that produced this turn’s pending tool calls.

None for continuations persisted before this field was added, or when the provider did not return an ID.

stop_reason: Option<StopReason>

Stop reason from the LLM call that produced this turn’s pending tool calls.

None for continuations persisted before this field was added.

response_content: Vec<ContentBlock>

Full content blocks from the LLM response that produced this turn’s pending tool calls (text, thinking, and tool-use blocks).

When the LLM emits text before tool calls (e.g. “I will run that.” followed by a tool_use block), those text blocks must be preserved so Phase 5 can reconstruct the complete assistant message in the conversation history.

Empty for continuations persisted before this field was added.

Trait Implementations

impl Clone for AgentContinuation

fn clone(&self) -> AgentContinuation

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for AgentContinuation

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for AgentContinuation

fn deserialize<__D>( __deserializer: __D, ) -> Result<AgentContinuation, <__D as Deserializer<'de>>::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for AgentContinuation

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

where __S: Serializer,

Serialize this value into the given Serde serializer.