daishi-sdk 0.2.19 → 0.2.21

package/dist/generated.d.ts

@@ -174,6 +174,7 @@ export interface ExperimentReport {
     rule_text: string;
     loss_tolerance: number | null;
     max_total_spend_usd: number | null;
+    focal: ExperimentReportFocal;
     arms: ExperimentReportArm[];
     pairs: ExperimentReportPair[];
     comparisons: ExperimentReportComparison[];
@@ -183,8 +184,11 @@ export interface ExperimentReport {
     completeness: ExperimentReportCompleteness;
     interference: ExperimentReportInterference;
     verdict: string;
+    limitations: ExperimentReportLimitation[];
     plain: string;
     what_differed: ExperimentReportWhatDiffered[];
+    replication: Replication | null;
+    reruns: RerunTally | null;
     cost: Cost;
     note: string | null;
     note_at_ms: number | null;
@@ -712,6 +716,8 @@ export interface BatchReportModel {
     median: number;
     sd: number | null;
     half_width: number | null;
+    design_effect: number | null;
+    effective_seats: number | null;
     best: number;
     worst: number;
     survival_rate: number;
@@ -820,6 +826,11 @@ export interface ExperimentDesign {
     standing: 'declared' | 'named' | 'after';
     name_prefix: string | null;
 }
+export interface ExperimentReportFocal {
+    seats: number[] | null;
+    background: number[];
+    text: string;
+}
 export interface ExperimentReportArm {
     index: number;
     label: string;
@@ -829,6 +840,7 @@ export interface ExperimentReportArm {
     differs_from_control: string[];
     trials: ExperimentReportArmTrials;
     score: ExperimentReportArmScore | null;
+    background: ExperimentReportArmBackground | null;
     cost_usd: number;
     agent_turns: number;
     usage: JsonObject;
@@ -851,6 +863,14 @@ export interface ExperimentReportArmScore {
     sd: number | null;
     ci: Interval | null;
 }
+export interface ExperimentReportArmBackground {
+    seats: number;
+    mean: number;
+    iqm: number;
+    median: number;
+    sd: number | null;
+    ci: Interval | null;
+}
 export interface ExperimentReportArmHealth {
     measured_seats: number;
     calls: number;
@@ -1017,11 +1037,72 @@ export interface ExperimentReportInterferencePerArm {
     retries: number;
     per_100_calls: number | null;
 }
+export interface ExperimentReportLimitation {
+    kind: 'construct' | 'access' | 'sample' | 'selection' | 'version' | 'measurement' | 'scope';
+    text: string;
+}
 export interface ExperimentReportWhatDiffered {
     arm: number;
     label: string;
     changes: string[];
 }
+export interface Replication {
+    original: ReplicationOriginal;
+    state: 'read' | 'playing' | 'incomplete' | 'cancelled' | 'unpublished' | 'original_open';
+    comparisons: StudyDifference[];
+    text: string;
+}
+export interface ReplicationOriginal {
+    experiment_id: string;
+    name: string;
+    handle: string | null;
+    study_url: string | null;
+    own: boolean;
+    exact: boolean;
+}
+export interface StudyDifference {
+    arm: number;
+    label: string;
+    control: string;
+    original: StudyUnits;
+    rerun: StudyUnits;
+    difference: number | null;
+    interval: WelchInterval | null;
+    apart: boolean | null;
+    text: string;
+}
+export interface StudyUnits {
+    unit: 'pair' | 'world';
+    n: number;
+    mean_diff: number | null;
+}
+export interface WelchInterval {
+    lo: number;
+    hi: number;
+    level: number;
+    df: number;
+}
+export interface RerunTally {
+    reruns: RerunEntry[];
+    counted: number;
+    apart: number;
+    no_interval: number;
+    not_counted: number;
+    own: number;
+    more: boolean;
+    text: string;
+}
+export interface RerunEntry {
+    experiment_id: string;
+    name: string;
+    handle: string | null;
+    study_url: string | null;
+    own: boolean;
+    exact: boolean;
+    state: 'read' | 'playing' | 'incomplete' | 'cancelled' | 'unpublished' | 'original_open';
+    apart: boolean | null;
+    comparisons: StudyDifference[];
+}
 /** One setup of a run set: how many runs are filed under it and the series batches among them, the harness versions their seats registered under (a set may span versions; nothing is pooled across them), when the earliest run ever filed under it started its match (a run moved out does not raise it), its design, and its counts summed over its runs (null while none carries one; `uncounted` runs whose match could not be read add nothing). */
 export interface RunSetSetup {
     key: string;

package/dist/index.d.ts

@@ -1,6 +1,6 @@
 import { Operations, type ErrorCode, type GetBatchResponse, type GetExperimentResponse, type GetRunResponse, type Method, type RequestParts } from './generated.js';
 export * from './generated.js';
-export declare const VERSION = "0.2.19";
+export declare const VERSION = "0.2.21";
 export declare const DEFAULT_BASE_URL = "https://daishi.ai/api/v1";
 export interface DaishiOptions {
     /** A Studio access token (`dsk_...`). Default: the `DAISHI_TOKEN` environment variable. */

package/dist/index.js

@@ -1,6 +1,6 @@
 import { Operations } from './generated.js';
 export * from './generated.js';
-export const VERSION = '0.2.19';
+export const VERSION = '0.2.21';
 export const DEFAULT_BASE_URL = 'https://daishi.ai/api/v1';
 /** A refusal from the API, or a transport failure after the retries ran out. */
 export class DaishiError extends Error {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "daishi-sdk",
-  "version": "0.2.19",
+  "version": "0.2.21",
   "description": "Typed client for the Daishi Studio API: scenarios, runs, series batches and paired experiments, with retries and waiting helpers.",
   "keywords": [
     "daishi",