Struct ToolRound 

struct ToolRound {
    pending_items: Vec<PendingRoundItem>,
    next_round_input: Vec<APIInputMessage>,
    response_id: String,
    suspended: bool,
}

What one tool-call round accumulates while it streams: the items it acts on once Azure has sent them all, the input the next round is sent, and whether the turn ends suspended instead of asking again.

Fields

pending_items: Vec<PendingRoundItem>

next_round_input: Vec<APIInputMessage>

response_id: String

The response every item of this round belongs to.

suspended: bool

Implementations

impl ToolRound

fn new(response_id: String) -> Self

fn queue(&mut self, item: OutputItem) -> ChatbotResult<()>

Queues an item for the finalize pass.

Only an item Azure has sent whole may be queued: it sends each one twice, and the earlier copy has neither a call’s arguments nor a reasoning item’s payload, both of which the next request has to carry. Errors on the item kinds the round handles before it reaches here: a Message is either streamed or, as a refusal, stored as it arrives, and a FunctionCallOutput never arrives from Azure at all.

fn has_function_calls(&self) -> bool