Struct EvalClient 

pub struct EvalClient<'a> { /* private fields */ }

Eval API client.

Use this to submit, poll, and cancel evaluation jobs.

Implementations

impl<'a> EvalClient<'a>

pub async fn submit(&self, request: EvalJobRequest) -> Result<String, CoreError>

Submit an evaluation job.

Arguments

• request - The eval job configuration

Returns

The job ID on success.

Example

let job_id = client.eval().submit(EvalJobRequest {
    container_url: "http://localhost:8000".into(),
    env_name: "default".into(),
    seeds: vec![1, 2, 3, 4, 5],
    policy: PolicyConfig::default(),
    ..Default::default()
}).await?;

pub async fn submit_raw(&self, request: Value) -> Result<String, CoreError>

Submit a raw evaluation job from a JSON value.

pub async fn get_status(&self, job_id: &str) -> Result<EvalResult, CoreError>

Get the current status of an eval job.

Arguments

• job_id - The job ID to check

Returns

The current eval result including status, mean reward, etc.

pub async fn get_results(&self, job_id: &str) -> Result<Value, CoreError>

Get detailed eval results for a job.

This includes per-seed metrics, tokens, costs, and errors.

pub async fn download_traces(&self, job_id: &str) -> Result<Vec<u8>, CoreError>

Download traces for an eval job as a ZIP archive.

pub async fn poll_until_complete( &self, job_id: &str, timeout_secs: f64, interval_secs: f64, ) -> Result<EvalResult, CoreError>

Poll an eval job until it reaches a terminal state.

Arguments

• job_id - The job ID to poll
• timeout_secs - Maximum time to wait (in seconds)
• interval_secs - Initial polling interval (in seconds)

Returns

The final eval result.

Example

let result = client.eval().poll_until_complete(&job_id, 1800.0, 10.0).await?;
println!("Mean reward: {:?}", result.mean_reward);

pub async fn cancel( &self, job_id: &str, reason: Option<String>, ) -> Result<Value, CoreError>

Cancel a running eval job.

Arguments

• job_id - The job ID to cancel
• reason - Optional cancellation reason

pub async fn query_workflow_state( &self, job_id: &str, ) -> Result<Value, CoreError>

Query workflow state for an eval job.

pub async fn list( &self, limit: Option<i32>, status: Option<EvalJobStatus>, ) -> Result<Vec<EvalResult>, CoreError>

List recent eval jobs.

Arguments

• limit - Maximum number of jobs to return
• status - Optional status filter