Struct DecisionStatement 

pub struct DecisionStatement {

    pub type_: String,
    pub timestamp: String,
    pub actor: String,
    pub parent_id: Option<String>,
    pub model: Option<String>,
    pub model_version: Option<String>,
    pub provider: Option<String>,
    pub tokens_in: Option<u64>,
    pub tokens_out: Option<u64>,
    pub prompt_digest: Option<String>,
    pub summary: Option<String>,
    pub confidence: Option<f64>,
    pub alternatives: Option<Vec<String>>,
    pub meta: Option<Value>,
}

Records an agent’s reasoning and decision context.

This is the “why” layer – agents provide this explicitly to explain inference decisions, model usage, and confidence levels.

Fields

type_: String

Always TYPE_DECISION

timestamp: String

RFC 3339 timestamp, set at sign time.

actor: String

DID-style actor URI. e.g. “agent://analyst”

parent_id: Option<String>

Links this artifact to its parent in the chain.

model: Option<String>

Model used for inference. e.g. “claude-opus-4-7”, “kimi-k2”, “gpt-5”

model_version: Option<String>

Model version if known.

provider: Option<String>

Provider that hosts the model. e.g. “anthropic”, “moonshot”, “openai”, “google”, “meta”, “mistral”, “ollama”.

Distinct from model: a “surface” (the runtime that runs the agent loop – Claude Code, Cursor, Codex, OpenClaw, Hermes, Cline) can be paired with any provider/model. Kimi for example is model = "kimi-k2" with provider = "moonshot", runnable from any surface that speaks OpenAI-compatible APIs. Attributing both lets a downstream auditor reason about surface, model, and provider independently.

Defaulted on deserialization so pre-v0.10.2 artifacts that were signed without provider still parse cleanly.

tokens_in: Option<u64>

Number of input tokens consumed.

tokens_out: Option<u64>

Number of output tokens produced.

prompt_digest: Option<String>

SHA-256 digest of the full prompt (not the prompt itself).

summary: Option<String>

Human-readable summary of the decision.

confidence: Option<f64>

Confidence level 0.0-1.0 if the agent provides it.

alternatives: Option<Vec<String>>

Other options the agent considered.

meta: Option<Value>

Arbitrary additional metadata.

Implementations

impl DecisionStatement

pub fn new(actor: impl Into<String>) -> Self

Trait Implementations

impl Clone for DecisionStatement

fn clone(&self) -> DecisionStatement

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for DecisionStatement

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for DecisionStatement

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for DecisionStatement

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

where __S: Serializer,

Serialize this value into the given Serde serializer.