Struct GenerateResult 

pub struct GenerateResult {
    pub text: String,
    pub token_ids: Vec<u32>,
    pub prompt_tokens: usize,
    pub tokens_generated: usize,
    pub finish_reason: String,
    pub mtp_drafted: usize,
    pub mtp_accepted: usize,
    pub token_confidence: Vec<f32>,
    pub traces: Vec<TokenTrace>,
}

Result of a generation call.

Fields

text: String

token_ids: Vec<u32>

prompt_tokens: usize

tokens_generated: usize

finish_reason: String

mtp_drafted: usize

Speculative-decode stats (0/0 when MTP is absent or inactive).

mtp_accepted: usize

token_confidence: Vec<f32>

Per-generated-token confidence = softmax probability of the token that was actually emitted (Born mass on the chosen state). High = the model was sure; low = it was guessing. Same length as the generated slice of token_ids.

traces: Vec<TokenTrace>

Structured per-token telemetry (B4 channel). Empty unless set_trace(true); otherwise same length as the generated slice.