@datalayer/core 1.1.66 → 1.2.12

package/lib/api/evals/client.d.ts

@@ -0,0 +1,402 @@
+/**
+ * The evals API, one function per endpoint.
+ *
+ * Paths mirror `tests/test_evals_api_contract.py` in the AI Agents service
+ * and `evalsApiContract.spec.ts` in the landings UI. Downloads (export and
+ * report) are URLs rather than requests: the browser opens them and the
+ * service answers with an attachment.
+ *
+ * @module api/evals/client
+ */
+import { type EvalsClientOptions } from './request';
+import type { CaseListResponse, CreateReportRequest, MoveReportRequest, RegenerateReportResponse, ReportListResponse, ReportResponse, ReportsQuery, DecisionListResponse, DecisionRequest, DecisionResponse, DecisionSubject, SharedWithMeResponse, ContinueReportImportResponse, ReportImportListResponse, ReportImportRequest, ReportImportResponse, EvalsPermissionsResponse, EvalsSharingResponse, EvalsSharingUpdate, SharedEvalsRecord, EvalEvaluatorRef, EvalsetVersionListResponse, EvalsetVersionResponse, InvestigationListResponse, InvestigationResponse, InvestigationScope, InvestigationStatus, LexicalReportResponse, ReportDocumentResponse, ResumeSandboxRequest, ResumeSandboxResponse, TaskInvestigationResponse, UpdateInvestigationRequest, EvalsetCategory, ImportEvalsetRequest, ImportEvalsetResponse, CaseResultListResponse, CaseResultResponse, RunNetworkResponse, RunNetworkMessageResponse, LaunchNetworkResponse, CreateLaunchRequest, LaunchCancelResponse, LaunchListResponse, ClaimTrialResponse, LaunchPlanResponse, LaunchResponse, LiveAlertListResponse, SubjectsResponse, ReviewCaseRequest, RunCancelResponse, CaseRequest, CaseResponse, CreateEvalsetRequest, CreateExperimentRequest, CreateLiveEventRequest, CreateRunRequest, EvalKind, EvalRunEnvironment, EvalsetDeleteResponse, EvalsetListResponse, EvalsetResponse, EvalsetPackagePreviewResponse, EvalsetComparisonResponse, SubjectUsageResponse, SandboxProvenanceResponse, BenchmarkComputeResponse, EvalsSearchResponse, EvalsetLeaderboardResponse, EvalsetPublicationResponse, EvalsetPublicationListResponse, ExperimentDeleteResponse, ExperimentListResponse, ExperimentResponse, LiveEventListResponse, LiveEventResponse, LiveTargetDeleteResponse, LiveTargetListResponse, PublicEvalsetDetailsResponse, RunListResponse, RunResponse, SuccessResponse, UpdateEvalsetRequest, UpdateExperimentRequest } from './types';
+export type ListEvalsetsQuery = {
+    kind?: EvalKind;
+    run_environment?: EvalRunEnvironment;
+    category?: EvalsetCategory;
+    q?: string;
+    /** Only the benchmarks reading this dataset (B5-12, section 15.3). */
+    dataset?: string;
+    /** …and pinned to this revision of it: one revision is not evidence about another. */
+    revision?: string;
+    limit?: number;
+    offset?: number;
+};
+export declare const listEvalsets: (options: EvalsClientOptions, query?: ListEvalsetsQuery) => Promise<EvalsetListResponse>;
+export declare const getEvalset: (options: EvalsClientOptions, evalsetId: string) => Promise<EvalsetResponse>;
+export declare const createEvalset: (options: EvalsClientOptions, body: CreateEvalsetRequest) => Promise<EvalsetResponse>;
+export declare const updateEvalset: (options: EvalsClientOptions, evalsetId: string, body: UpdateEvalsetRequest) => Promise<EvalsetResponse>;
+export declare const setEvalsetPublic: (options: EvalsClientOptions, evalsetId: string, isPublic: boolean) => Promise<EvalsetResponse>;
+/**
+ * Take a benchmark into the account the options name (B5-03): a private copy
+ * of its definition that says what it was taken from. A published benchmark
+ * is anybody's to take; an unpublished one, a Viewer's. `derivation` says why:
+ * `fork` to change it, `compare` to run it against one's own agent (B5-05).
+ */
+export declare const cloneEvalset: (options: EvalsClientOptions, evalsetId: string, body?: {
+    name?: string;
+    derivation?: "fork" | "compare";
+}) => Promise<EvalsetResponse>;
+/**
+ * Propose the next version of a benchmark's definition, from an investigation
+ * (B5-07): a task added or corrected, a corrected evaluator, a threshold
+ * moved. The version records the investigation that asked for it.
+ */
+export declare const reviseEvalset: (options: EvalsClientOptions, evalsetId: string, body: {
+    investigation_id: string;
+    note?: string;
+    case?: Record<string, unknown>;
+    evalset_evaluators?: EvalEvaluatorRef[];
+    report_evaluators?: EvalEvaluatorRef[];
+}) => Promise<EvalsetResponse>;
+/**
+ * What publishing this benchmark would put in the library (B5-02): the
+ * contents of the package, the enumeration the publication review shows, what
+ * is left out and why, and everything that would refuse it — before anybody
+ * presses the button, because a publication cannot be taken back from whoever
+ * already read it.
+ */
+export declare const previewEvalsetPublication: (options: EvalsClientOptions, evalsetId: string, query?: {
+    /** Launch ids, comma separated; the benchmark's own where empty. */
+    launches?: string;
+    report?: string;
+    /** Comment uids selected for publication; nothing is published unnamed. */
+    comments?: string;
+    decisions?: string;
+    version?: number;
+}) => Promise<EvalsetPackagePreviewResponse>;
+/**
+ * Publish the package: the immutable snapshot of the definition, the data, the
+ * subjects, the environment, the results, the report, the evidence and the
+ * evaluators (section 14.3). Refused with `detail.problems` where section 14.5
+ * says it cannot be published.
+ */
+export declare const publishEvalsetPackage: (options: EvalsClientOptions, evalsetId: string, body?: {
+    launch_ids?: string[];
+    report_id?: string;
+    comment_uids?: string[];
+    decision_uids?: string[];
+    note?: string;
+    version?: number;
+}) => Promise<EvalsetPublicationResponse>;
+/** Every package published of this benchmark, withdrawn ones included. */
+export declare const listEvalsetPublications: (options: EvalsClientOptions, evalsetId: string, query?: {
+    limit?: number;
+    offset?: number;
+}) => Promise<EvalsetPublicationListResponse>;
+/**
+ * Take a package out of the library. What it holds stays as it was written:
+ * withdrawing a snapshot does not make it editable.
+ */
+export declare const withdrawEvalsetPublication: (options: EvalsClientOptions, evalsetId: string, publicationId: string) => Promise<EvalsetPublicationResponse>;
+/** The package a published benchmark stands for, to anybody (B5-02). */
+export declare const getPublicEvalsetPublication: (options: EvalsClientOptions, evalsetId: string) => Promise<EvalsetPublicationResponse>;
+/**
+ * Launches of one benchmark side by side (B5-08): what each scored, what each
+ * subject scored in each, and which tasks went from passing to failing or
+ * back. Launch ids are compared in the order given, oldest first; the two
+ * newest launches where none is named.
+ */
+export declare const compareEvalsetLaunches: (options: EvalsClientOptions, evalsetId: string, query?: {
+    launches?: string;
+    /** The published report that sent somebody here, when one did (B5-14). */
+    report?: string;
+}) => Promise<EvalsetComparisonResponse>;
+/**
+ * The subjects that ran a published benchmark, best first. Only runs a package
+ * published are counted (B5-02), and each row names the package its best run
+ * came from.
+ */
+export declare const getPublicEvalsetLeaderboard: (options: EvalsClientOptions, evalsetId: string) => Promise<EvalsetLeaderboardResponse>;
+/**
+ * What an agent or a model has actually run (B5-12): the benchmarks that ran
+ * it, what its runs scored, what they cost, how long they took, and the
+ * launches still going. Empty where it has never been run.
+ *
+ * The ref keeps its slashes — an agentspec id has one — so each segment is
+ * encoded rather than the whole string.
+ */
+export declare const getSubjectUsage: (options: EvalsClientOptions, subjectRef: string) => Promise<SubjectUsageResponse>;
+/**
+ * Whether anything will run the account's benchmarks (B5-12, section 15.2):
+ * the executor, whether a worker polls the benchmark queue, and the account's
+ * own queued and running launches.
+ */
+export declare const getBenchmarkCompute: (options: EvalsClientOptions) => Promise<BenchmarkComputeResponse>;
+/**
+ * What benchmark work a sandbox is doing (B5-12, section 15.4).
+ *
+ * The ref is a pool slot's name (`benchmark-<run>-slot-<n>`) or the snapshot
+ * a sandbox was restored from. A sandbox of nobody's benchmark answers
+ * `found: false`, so a page asks without knowing in advance.
+ */
+export declare const getSandboxProvenance: (options: EvalsClientOptions, sandboxRef: string) => Promise<SandboxProvenanceResponse>;
+/**
+ * Find things across the caller's benchmarks (B5-11, section 19): benchmarks,
+ * launches, tasks, reports and investigations, each row keeping its kind and
+ * what it belongs to.
+ *
+ * Nobody types a field name, so the service reads the words first: a status
+ * word becomes a status, `run 128` becomes a launch number, and what is left
+ * is the text. An empty query finds nothing rather than everything.
+ */
+export declare const searchEvals: (options: EvalsClientOptions, query?: {
+    q?: string;
+    kinds?: string;
+    limit?: number;
+}) => Promise<EvalsSearchResponse>;
+export declare const renameEvalset: (options: EvalsClientOptions, evalsetId: string, name: string) => Promise<EvalsetResponse>;
+export declare const deleteEvalset: (options: EvalsClientOptions, evalsetId: string) => Promise<EvalsetDeleteResponse>;
+export declare const listCases: (options: EvalsClientOptions, evalsetId: string) => Promise<CaseListResponse>;
+export declare const createCase: (options: EvalsClientOptions, evalsetId: string, body: CaseRequest) => Promise<CaseResponse>;
+export declare const updateCase: (options: EvalsClientOptions, evalsetId: string, caseId: string, body: Partial<CaseRequest>) => Promise<CaseResponse>;
+export declare const deleteCase: (options: EvalsClientOptions, evalsetId: string, caseId: string) => Promise<SuccessResponse>;
+export type EvalsetExportFormat = 'json' | 'pydantic-evals';
+export type EvalsetReportFormat = 'markdown' | 'csv';
+/** The URL the browser opens to download an evalset definition. */
+export declare const evalsetExportUrl: (options: EvalsClientOptions, evalsetId: string, format?: EvalsetExportFormat) => string;
+/** The URL the browser opens to download a generated report. */
+export declare const evalsetReportUrl: (options: EvalsClientOptions, evalsetId: string, format?: EvalsetReportFormat, runLimit?: number) => string;
+export declare const getPublicEvalset: (options: EvalsClientOptions, evalsetId: string) => Promise<EvalsetResponse>;
+export declare const getPublicEvalsetDetails: (options: EvalsClientOptions, evalsetId: string, runsLimit?: number) => Promise<PublicEvalsetDetailsResponse>;
+export declare const listPublicRuns: (options: EvalsClientOptions, evalsetId: string, experimentId: string, query?: {
+    limit?: number;
+    offset?: number;
+}) => Promise<RunListResponse>;
+export type ListExperimentsQuery = {
+    evalset_id?: string;
+    status?: string;
+    limit?: number;
+    offset?: number;
+};
+export declare const createExperiment: (options: EvalsClientOptions, body: CreateExperimentRequest) => Promise<ExperimentResponse>;
+export declare const listExperiments: (options: EvalsClientOptions, query?: ListExperimentsQuery) => Promise<ExperimentListResponse>;
+export declare const getExperiment: (options: EvalsClientOptions, experimentId: string) => Promise<ExperimentResponse>;
+export declare const updateExperiment: (options: EvalsClientOptions, experimentId: string, body: UpdateExperimentRequest) => Promise<ExperimentResponse>;
+export declare const deleteExperiment: (options: EvalsClientOptions, experimentId: string) => Promise<ExperimentDeleteResponse>;
+export declare const createRun: (options: EvalsClientOptions, experimentId: string, body: CreateRunRequest) => Promise<RunResponse>;
+export declare const listRuns: (options: EvalsClientOptions, experimentId: string, query?: {
+    limit?: number;
+    offset?: number;
+}) => Promise<RunListResponse>;
+export declare const getRun: (options: EvalsClientOptions, runId: string) => Promise<RunResponse>;
+export declare const deleteRun: (options: EvalsClientOptions, runId: string) => Promise<SuccessResponse>;
+export declare const compareRuns: (options: EvalsClientOptions, runIds: string[]) => Promise<RunListResponse>;
+export type ListLiveEventsQuery = {
+    target_id: string;
+    target_type?: string;
+    window?: string;
+    evaluator_name?: string;
+    limit?: number;
+    offset?: number;
+};
+export declare const createLiveEvent: (options: EvalsClientOptions, body: CreateLiveEventRequest) => Promise<LiveEventResponse>;
+export declare const listLiveTargets: (options: EvalsClientOptions, query?: {
+    window?: string;
+    limit?: number;
+}) => Promise<LiveTargetListResponse>;
+export declare const listLiveEvents: (options: EvalsClientOptions, query: ListLiveEventsQuery) => Promise<LiveEventListResponse>;
+export type ListLiveAlertsQuery = {
+    target_id?: string;
+    target_type?: string;
+    experiment_id?: string;
+    launch_id?: string;
+    window?: string;
+    limit?: number;
+};
+/** The alerts the rolling windows raised: failure spikes and drifts (B2-13). */
+export declare const listLiveAlerts: (options: EvalsClientOptions, query?: ListLiveAlertsQuery) => Promise<LiveAlertListResponse>;
+export declare const deleteLiveTarget: (options: EvalsClientOptions, targetId: string, targetType?: string) => Promise<LiveTargetDeleteResponse>;
+export type ListRunsQuery = {
+    evalset_id?: string;
+    launch_id?: string;
+    experiment_id?: string;
+    status?: string;
+    limit?: number;
+    offset?: number;
+};
+export declare const listRunsAcross: (options: EvalsClientOptions, query?: ListRunsQuery) => Promise<RunListResponse>;
+export declare const cancelRun: (options: EvalsClientOptions, runId: string) => Promise<RunCancelResponse>;
+export type ListCaseResultsQuery = {
+    status?: string;
+    category?: string;
+    difficulty?: string;
+    failure_mode?: string;
+    min_score?: number;
+    max_score?: number;
+    q?: string;
+    sort?: string;
+    limit?: number;
+    offset?: number;
+};
+/** The task grid of a run, filtered and paged by the service. */
+export declare const listCaseResults: (options: EvalsClientOptions, runId: string, query?: ListCaseResultsQuery) => Promise<CaseResultListResponse>;
+export declare const getCaseResult: (options: EvalsClientOptions, runId: string, caseId: string) => Promise<CaseResultResponse>;
+/**
+ * The run as the network of agents that worked on it. `budget` is how many
+ * tasks' messages to answer: every task that did not pass, then passed ones
+ * while there is room.
+ */
+export declare const getRunNetwork: (options: EvalsClientOptions, runId: string, query?: {
+    budget?: number;
+}) => Promise<RunNetworkResponse>;
+/**
+ * One message of a run's network, in full. The id is the service's own and
+ * holds colons, which the route reads as a path: each segment is encoded, the
+ * separators are kept.
+ */
+export declare const getRunNetworkMessage: (options: EvalsClientOptions, runId: string, messageId: string) => Promise<RunNetworkMessageResponse>;
+/** A launch's runs, each as one line of its own network. */
+export declare const getLaunchNetwork: (options: EvalsClientOptions, launchId: string) => Promise<LaunchNetworkResponse>;
+export declare const reviewCaseResult: (options: EvalsClientOptions, runId: string, caseId: string, body: ReviewCaseRequest) => Promise<CaseResultResponse>;
+/**
+ * Give an anonymous trial's work to the person who just signed in (B2-14):
+ * the benchmark, the launch, its runs and everything under them, moved
+ * rather than copied so a link the visitor kept still opens.
+ *
+ * Takes two credentials: the person's, in the client's options, and the
+ * trial's own key, which travels in its header.
+ */
+export declare const claimTrial: (options: EvalsClientOptions, trialUid: string, trialToken: string) => Promise<ClaimTrialResponse>;
+/** The subjects an experiment can have, and the models offered (B2-11). */
+export declare const listSubjects: (options: EvalsClientOptions) => Promise<SubjectsResponse>;
+/**
+ * What a page may report about what somebody did (B2-25, B3-12).
+ *
+ * The six lines of the funnel no record answers: nothing is written when
+ * somebody opens the Run Benchmark wizard, reads a task, asks the result agent
+ * a question, runs a follow-up cell, or pins evidence into a report. A closed
+ * set, because a free-form label from a browser is a cardinality bomb with an
+ * authenticated route in front of it.
+ */
+export type ProductEvent = 'wizard.started' | 'wizard.completed' | 'task.opened' | 'question.asked' | 'cell.executed' | 'evidence.added' | 'network.viewed' | 'network.node_selected' | 'network.message_opened' | 'network.investigate_clicked';
+/**
+ * Count one of those.
+ *
+ * Nothing is stored, and the answer says only that it was counted. Never worth
+ * failing a page for: a measure that breaks what somebody was doing is worse
+ * than a measure nobody has.
+ */
+export declare const recordProductEvent: (options: EvalsClientOptions, event: ProductEvent) => Promise<{
+    success: boolean;
+}>;
+/**
+ * The plan of a launch before it is made (B2-07): estimated duration and
+ * cost, the compute, and the problems that would stop it. Same body as the
+ * launch; nothing is created.
+ */
+export declare const validateLaunch: (options: EvalsClientOptions, evalsetId: string, body: CreateLaunchRequest) => Promise<LaunchPlanResponse>;
+/** One submission of a benchmark across its experiments. */
+export declare const createLaunch: (options: EvalsClientOptions, evalsetId: string, body: CreateLaunchRequest) => Promise<LaunchResponse>;
+export type ListLaunchesQuery = {
+    evalset_id?: string;
+    status?: string;
+    include_archived?: boolean;
+    limit?: number;
+    offset?: number;
+};
+export declare const listLaunches: (options: EvalsClientOptions, query?: ListLaunchesQuery) => Promise<LaunchListResponse>;
+export declare const getLaunch: (options: EvalsClientOptions, launchId: string) => Promise<LaunchResponse>;
+export declare const cancelLaunch: (options: EvalsClientOptions, launchId: string) => Promise<LaunchCancelResponse>;
+export declare const archiveLaunch: (options: EvalsClientOptions, launchId: string) => Promise<SuccessResponse>;
+/** An evalset from a spec file, the same body the CLI and the action send. */
+export declare const importEvalset: (options: EvalsClientOptions, body: ImportEvalsetRequest) => Promise<ImportEvalsetResponse>;
+export declare const listEvalsetVersions: (options: EvalsClientOptions, evalsetId: string, query?: {
+    limit?: number;
+    offset?: number;
+}) => Promise<EvalsetVersionListResponse>;
+export declare const getEvalsetVersion: (options: EvalsClientOptions, evalsetId: string, version: number) => Promise<EvalsetVersionResponse>;
+/** The benchmark's whole report (every launch) as a serialized Lexical editor state. */
+export declare const getEvalsetReportDocument: (options: EvalsClientOptions, evalsetId: string, query?: {
+    run_limit?: number;
+}) => Promise<LexicalReportResponse>;
+/** The launch's report as a serialized Lexical editor state. */
+export declare const getLaunchReportDocument: (options: EvalsClientOptions, launchId: string, query?: {
+    run_limit?: number;
+}) => Promise<LexicalReportResponse>;
+/** Where a launch's Markdown or CSV report is downloaded from. */
+export declare const launchReportUrl: (options: EvalsClientOptions, launchId: string, format: "markdown" | "csv") => string;
+/** One run's report as a serialized Lexical editor state, with its task documents. */
+export declare const getRunReportDocument: (options: EvalsClientOptions, runId: string) => Promise<LexicalReportResponse>;
+export declare const runReportUrl: (options: EvalsClientOptions, runId: string, format: "markdown" | "csv") => string;
+/** Write the launch's report into the account's benchmarks space. */
+export declare const writeLaunchReportDocument: (options: EvalsClientOptions, launchId: string, body?: {
+    name?: string;
+}) => Promise<ReportDocumentResponse>;
+/** Open the investigation of one task, or get the one already open. */
+export declare const openTaskInvestigation: (options: EvalsClientOptions, runId: string, caseId: string) => Promise<TaskInvestigationResponse>;
+export declare const openLaunchInvestigation: (options: EvalsClientOptions, launchId: string) => Promise<InvestigationResponse>;
+export type InvestigationsQuery = {
+    evalset_id?: string;
+    launch_id?: string;
+    run_id?: string;
+    scope?: InvestigationScope;
+    status?: InvestigationStatus;
+    limit?: number;
+    offset?: number;
+};
+export declare const listInvestigations: (options: EvalsClientOptions, query?: InvestigationsQuery) => Promise<InvestigationListResponse>;
+export declare const getInvestigation: (options: EvalsClientOptions, investigationId: string) => Promise<InvestigationResponse>;
+export declare const updateInvestigation: (options: EvalsClientOptions, investigationId: string, body: UpdateInvestigationRequest) => Promise<InvestigationResponse>;
+/**
+ * Publish an investigation to the library, or take it back: its owner's alone.
+ * The document and the notebooks it is written in go public and private with
+ * it.
+ */
+export declare const setInvestigationPublic: (options: EvalsClientOptions, investigationId: string, isPublic: boolean) => Promise<InvestigationResponse>;
+/**
+ * Keep the investigation's page as it is shown, under a name everybody on
+ * the investigation sees; a view of that name is replaced (B4-11).
+ */
+export declare const saveInvestigationView: (options: EvalsClientOptions, investigationId: string, body: {
+    name: string;
+    query: string;
+}) => Promise<InvestigationResponse>;
+/** Forget a saved view of the investigation's page. */
+export declare const forgetInvestigationView: (options: EvalsClientOptions, investigationId: string, name: string) => Promise<InvestigationResponse>;
+/** Bring the task's sandbox back from its snapshot, bound to its investigation. */
+export declare const resumeTaskSandbox: (options: EvalsClientOptions, runId: string, caseId: string, body?: ResumeSandboxRequest) => Promise<ResumeSandboxResponse>;
+/** Who a benchmark, one of its runs or an investigation is shared with; its owner's to read. */
+export declare const getEvalsSharing: (options: EvalsClientOptions, record: SharedEvalsRecord, uid: string) => Promise<EvalsSharingResponse>;
+/** Replace the grants at the levels named; the others are kept. Its owner's to do. */
+export declare const updateEvalsSharing: (options: EvalsClientOptions, record: SharedEvalsRecord, uid: string, access: EvalsSharingUpdate) => Promise<EvalsSharingResponse>;
+/** The caller's role on the record, and what it allows. */
+export declare const getEvalsPermissions: (options: EvalsClientOptions, record: SharedEvalsRecord, uid: string) => Promise<EvalsPermissionsResponse>;
+/** A report over runs of a benchmark, written as a draft. */
+export declare const createReport: (options: EvalsClientOptions, body: CreateReportRequest) => Promise<ReportResponse>;
+export declare const listReports: (options: EvalsClientOptions, query?: ReportsQuery) => Promise<ReportListResponse>;
+export declare const getReport: (options: EvalsClientOptions, reportId: string) => Promise<ReportResponse>;
+/** Send a report for review, send it back, approve it or publish it. */
+export declare const moveReport: (options: EvalsClientOptions, reportId: string, body: MoveReportRequest) => Promise<ReportResponse>;
+/** A new version over newer runs, or the same ones; the old report is superseded by it. */
+export declare const regenerateReport: (options: EvalsClientOptions, reportId: string, launchIds?: string[]) => Promise<RegenerateReportResponse>;
+/** What a reviewer decided about part of a result, recorded on a report or in an investigation. */
+export declare const recordDecision: (options: EvalsClientOptions, subject: DecisionSubject, uid: string, body: DecisionRequest) => Promise<DecisionResponse>;
+/** The decisions made on a report or in an investigation, in the order they were made. */
+export declare const listDecisions: (options: EvalsClientOptions, subject: DecisionSubject, uid: string) => Promise<DecisionListResponse>;
+/**
+ * What was decided about a benchmark, or about one of its launches (B6-04).
+ *
+ * `listDecisions` answers what was decided *in* one report or investigation;
+ * this answers what was decided about the benchmark, which is what something
+ * outside the product has to ask — CI knows the benchmark it ran and nothing
+ * else. Naming neither answers nothing rather than everything.
+ */
+export declare const listDecisionsAbout: (options: EvalsClientOptions, query?: {
+    evalset_id?: string;
+    launch_id?: string;
+    limit?: number;
+}) => Promise<DecisionListResponse>;
+export type ReportExportFormat = 'markdown' | 'csv';
+/** Where a report is downloaded as it was written, with its decisions (B4-08). */
+export declare const reportExportUrl: (options: EvalsClientOptions, reportId: string, format?: ReportExportFormat) => string;
+/** The reports and investigations other people shared with the caller. */
+export declare const listSharedWithMe: (options: EvalsClientOptions) => Promise<SharedWithMeResponse>;
+/** Keep a report file CI produced on its benchmark, as its text. */
+export declare const importReport: (options: EvalsClientOptions, evalsetId: string, body: ReportImportRequest) => Promise<ReportImportResponse>;
+/** The reports imported on a benchmark, newest first, without their text. */
+export declare const listReportImports: (options: EvalsClientOptions, evalsetId: string) => Promise<ReportImportListResponse>;
+/** One imported report, with its text as it was. */
+export declare const getReportImport: (options: EvalsClientOptions, importId: string) => Promise<ReportImportResponse>;
+/** Continue investigation: a live report and an investigation over its runs. */
+export declare const continueReportImport: (options: EvalsClientOptions, importId: string) => Promise<ContinueReportImportResponse>;