#Class: EvalService

The one eval path the CLI, the programmatic API, and the workbench browser all use. It resolves every filesystem location itself: a caller names suites, cases, and a trial count, never a run directory, artifact copy, or command.

#Constructors

#Constructor

new EvalService(options): EvalService

#Parameters

#options

EvalServiceOptions

#Returns

EvalService

#Methods

#cancel()

cancel(runId): Promise<boolean>

#Parameters

#runId

string

#Returns

Promise<boolean>

---

#close()

close(): Promise<void>

#Returns

Promise<void>

---

#compare()

compare(baseRunId, candidateRunId): Promise<EvalComparison>

Alignment facets the run store does not persist yet stay unverified rather than being asserted as aligned, so a delta never rests on an assumption.

#Parameters

#baseRunId

string

#candidateRunId

string

#Returns

Promise<EvalComparison>

---

#events()

events(runId, afterSequence): Promise<EvalRunEventsReplay>

#Parameters

#runId

string

#afterSequence

number

#Returns

Promise<EvalRunEventsReplay>

---

#list()

list(): Promise<readonly EvalRunRecord[]>

#Returns

Promise<readonly EvalRunRecord[]>

---

#openArtifact()

openArtifact(runId, artifactRef): Promise<EvalArtifactReader>

#Parameters

#runId

string

#artifactRef

string

#Returns

Promise<EvalArtifactReader>

---

#read()

read(runId): Promise<EvalRunResult>

#Parameters

#runId

string

#Returns

Promise<EvalRunResult>

---

#run()

run(request): Promise<EvalRunResult>

#Parameters

#request

EvalRunRequest

#Returns

Promise<EvalRunResult>

---

#start()

start(request): Promise<EvalRunAdmission>

#Parameters

#request

EvalRunRequest

#Returns

Promise<EvalRunAdmission>

---

#subscribeEvents()

subscribeEvents(runId, afterSequence): Promise<EvalEventSubscription>

#Parameters

#runId

string

#afterSequence

number

#Returns

Promise<EvalEventSubscription>

---

#suites()

suites(): Promise<EvalSuiteListing>

#Returns

Promise<EvalSuiteListing>