Struct RunConfig 

pub struct RunConfig {

    pub max_turns: u32,
    pub max_tool_calls: u32,
    pub max_attempts: u32,
    pub best_of: u32,
    pub allow: Vec<Vec<String>>,
    pub allow_prefix: Vec<Vec<String>>,
    pub approval_budget: u32,
    pub commit_per_task: bool,
    pub acceptance: bool,
    pub held_out: bool,
    pub strength: StrengthConfig,
    pub max_converge_rounds: u32,
    pub monitor: bool,
    pub review_required: bool,
    pub oracle_samples: u8,
    pub frozen_additions: bool,
    pub format: Option<Vec<String>>,
    pub approval_wait: Duration,
    pub attest: AttestConfig,
    pub instructions: Vec<String>,
    pub reproduce_tries: u32,
    pub bug_test: Option<Vec<String>>,
}

Limits and command allowances for an attempt.

Fields

max_turns: u32

Model turns per attempt.

max_tool_calls: u32

Tool calls per attempt.

max_attempts: u32

Attempts per task run; failed verdicts are retried up to this.

best_of: u32

Independent samples per task, the best by evidence kept.

allow: Vec<Vec<String>>

Exact argvs run_command runs without asking ([commands] allow).

allow_prefix: Vec<Vec<String>>

Argv prefixes (two elements or more) it runs without asking ([commands] allow_prefix), saved only from a coverage preview.

approval_budget: u32

Prompts a run shows a person before further calls wait together in Needs You ([approvals] budget, 0–100, default 12).

commit_per_task: bool

Commit each passed task on the speccoder branch ([git] commit = "per-task").

acceptance: bool

Derive acceptance tests from the spec before attempts ([acceptance] generate).

held_out: bool

Also author held-out acceptance tests the implementer never sees ([acceptance] held_out, default on; nothing without generate).

strength: StrengthConfig

Acceptance strength.

max_converge_rounds: u32

Convergence phases a feature may get ([converge] max_rounds).

monitor: bool

Ask the gaming monitor about every passing task ([monitor] enabled).

review_required: bool

A run waits until every requirement is approved on its current wording ([review] required).

oracle_samples: u8

Independent author calls Write Tests makes at review ([review] oracle_samples, 1 or 2, default 2): two are compared for ambiguity, one turns the comparison off.

frozen_additions: bool

Test files that existed before an attempt may gain lines without approval ([tests] frozen = "additions"; default "ask").

format: Option<Vec<String>>

The formatter speccoder runs after each attempt, before verifying ([format] argv): formatting is not the model’s job to remember.

approval_wait: Duration

How long a call waits for a person before it is denied as expired ([approvals] wait, default 7 days).

attest: AttestConfig

[attest].

instructions: Vec<String>

[context] instructions: the file names read per directory as the repository’s instructions for agents; empty: none.

reproduce_tries: u32

[bug] reproduce_tries: the reproduction author’s tries (1–5).

bug_test: Option<Vec<String>>

[bug] test: how one named test runs ({file}, {name} filled), when the toolchain’s own rule does not fit.

Trait Implementations

impl Clone for RunConfig

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for RunConfig

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

Formats the value using the given formatter.

impl Default for RunConfig

fn default() -> Self

Returns the “default value” for a type.

impl Eq for RunConfig

impl PartialEq for RunConfig

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

Equality operator ==.

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

Inequality operator !=.

impl StructuralPartialEq for RunConfig