Struct StepRecord 

pub struct StepRecord {
    pub step: u32,
    pub decision: String,
    pub result: String,
    pub prompt: String,
    pub tool_call: String,
    pub tokens: u64,
}

One recorded loop step — the full trace entry, as written and read back.

use io_harness::{StepRecord, Store};

let store = Store::memory()?;
let run_id = store.start_run("add a hello function", "src/hello.rs")?;

// A transient provider failure, and then the step that worked. Both are step
// 1: a retry writes its own row under the step number it retried.
store.record(run_id, &StepRecord::new(1, "retry 1 after a 503", ""))?;

// `new` alone records a decision and its result; `with_trace` adds the audit
// half — the exact prompt sent, the call the model made, and what it cost.
// Without it the trace says what happened and not why.
store.record(
    run_id,
    &StepRecord::new(1, "wrote src/hello.rs", "ok").with_trace(
        "<the assembled prompt>",
        r#"{"name":"write_file","arguments":{"path":"src/hello.rs"}}"#,
        1_280,
    ),
)?;

// Read back for audit — and the reason `RunSummary::steps` is `MAX(step)`
// rather than a row count: two rows here, one agent step.
let steps = store.steps(run_id)?;
assert_eq!(steps.len(), 2);
assert_eq!(store.last_step(run_id)?, 1);
assert!(steps[1].tool_call.contains("write_file"));
assert_eq!(steps.iter().map(|s| s.tokens).sum::<u64>(), 1_280);

Fields

step: u32

1-based step number within the run.

decision: String

What the agent decided this step (e.g. “wrote file”, “retry 1 after error”).

result: String

Intermediate result / model text for the step.

prompt: String

The prompt sent to the model this step.

tool_call: String

The tool call the model made, as JSON, or “” if none.

tokens: u64

Total tokens used this step, 0 if the provider reported none.

Implementations

impl StepRecord

pub fn new( step: u32, decision: impl Into<String>, result: impl Into<String>, ) -> Self

A trace entry with the audit fields empty — for callers that only record a decision and result.

pub fn with_trace( self, prompt: impl Into<String>, tool_call: impl Into<String>, tokens: u64, ) -> Self

Attach the prompt, tool call, and token count for the full trace.

Trait Implementations

impl Clone for StepRecord

fn clone(&self) -> StepRecord

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for StepRecord

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

Formats the value using the given formatter.

impl PartialEq for StepRecord

fn eq(&self, other: &StepRecord) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl StructuralPartialEq for StepRecord