humanish 0.108.0 → 0.109.1

package/dist/index.d.ts

@@ -1,12 +1,12 @@
-export { runPackageLab as runStudy } from "./run-lab.js";
-export type { LabOutcome as StudyOutcome, LabResult as StudyResult, RunLabOptions as RunStudyOptions, } from "./run-lab.js";
-export type { LabEvent as StudyEvent } from "./lab/run-lab-events.js";
-export type { ProviderContext } from "./lab/run-lab-homes.js";
-export { routeOf } from "./lab/plan.js";
-export type { LabRoute as StudyRoute } from "./lab/plan.js";
-export { parseLabConfig as parseStudy } from "./lab/config.js";
-export { STUDY_SCHEMA } from "./lab/types.js";
-export type { LabConfig as StudyConfig } from "./lab/types.js";
+export { runStudy } from "./run-study.js";
+export type { StudyOutcome, StudyResult, RunStudyOptions } from "./run-study.js";
+export type { StudyEvent } from "./study/run-study-events.js";
+export type { ProviderContext } from "./study/run-study-homes.js";
+export { routeOf } from "./study/plan.js";
+export type { StudyRoute } from "./study/plan.js";
+export { parseStudy } from "./study/config.js";
+export { STUDY_SCHEMA } from "./study/types.js";
+export type { StudyConfig } from "./study/types.js";
 export { verifyRun } from "./verify/verify.js";
 export type { VerifyResult } from "./verify/verify.js";
 export { renderObserver } from "./observer/render.js";
@@ -18,14 +18,12 @@ export type { CuaAction as ComputerUseAction, CuaExecutor as ComputerUseExecutor
 export type { ActorCapabilities } from "./actors/contract.js";
 export { createOpenAiResponsesProvider } from "./actors/computer-use/openai-provider.js";
 export type { OpenAiResponsesProviderOptions } from "./actors/computer-use/openai-provider.js";
-export { CuaAdmissionLimitError as ComputerUseAdmissionLimitError } from "./actors/computer-use/admission-limit.js";
+export { ComputerUseAdmissionLimitError } from "./actors/computer-use/admission-limit.js";
 export { defaultRedactionHooks } from "./evidence/redaction.js";
 export type { RedactionHooks } from "./evidence/redaction.js";
 export type { E2BDesktopSandbox } from "./substrates/e2b/sdk.js";
-export { browserScorer, terminalScorer } from "./lab/adapter-scorer-loader.js";
-export type { AdapterScorerModule, AdapterScoringContext } from "./lab/adapter-scorer-loader.js";
-export type { BrowserLabScoringContext as BrowserScoringContext } from "./lab/adapter-extension.js";
+export { browserScorer, terminalScorer } from "./study/adapter-scorer-loader.js";
+export type { AdapterScorerModule, AdapterScoringContext } from "./study/adapter-scorer-loader.js";
+export type { BrowserScoringContext } from "./study/adapter-extension.js";
 export type { TerminalProductScoringContext } from "./routes/terminal/types.js";
 export type { RunAdapterArtifact, RunAdapterScore } from "./run/bundle.js";
-export { CuaAdmissionLimitError, LAB_CONFIG_SCHEMA, parseLabConfig, runLab, } from "./library-aliases.js";
-export type { BrowserLabScoringContext, CuaAction, CuaExecutor, CuaLoopOptions, CuaLoopResult, CuaObservation, CuaProvider, CuaSafetyCheck, CuaTurn, CuaTurnRequest, LabConfig, LabEvent, LabOutcome, LabResult, LabRoute, RunLabOptions, } from "./library-aliases.js";

package/dist/index.js

@@ -2,19 +2,17 @@
 // score a run. Everything else is internal and reached through the `humanish` bin. See
 // docs/contracts/schemas.md, "Library options", and the site's library page.
 // Run a study.
-export { runPackageLab as runStudy } from "./run-lab.js";
-export { routeOf } from "./lab/plan.js";
-export { parseLabConfig as parseStudy } from "./lab/config.js";
-export { STUDY_SCHEMA } from "./lab/types.js";
+export { runStudy } from "./run-study.js";
+export { routeOf } from "./study/plan.js";
+export { parseStudy } from "./study/config.js";
+export { STUDY_SCHEMA } from "./study/types.js";
 // Read a run.
 export { verifyRun } from "./verify/verify.js";
 export { renderObserver } from "./observer/render.js";
 // Bring a participant.
 export { runComputerUseLoop, stableProgressKey } from "./actors/computer-use/loop.js";
 export { createOpenAiResponsesProvider } from "./actors/computer-use/openai-provider.js";
-export { CuaAdmissionLimitError as ComputerUseAdmissionLimitError } from "./actors/computer-use/admission-limit.js";
+export { ComputerUseAdmissionLimitError } from "./actors/computer-use/admission-limit.js";
 export { defaultRedactionHooks } from "./evidence/redaction.js";
 // Score a run.
-export { browserScorer, terminalScorer } from "./lab/adapter-scorer-loader.js";
-// The 0.107 names, deprecated until 0.109 removes them. Each is its new name's value or type.
-export { CuaAdmissionLimitError, LAB_CONFIG_SCHEMA, parseLabConfig, runLab, } from "./library-aliases.js";
+export { browserScorer, terminalScorer } from "./study/adapter-scorer-loader.js";

package/dist/observer/artifact.d.ts

@@ -2,7 +2,7 @@ import type { LoadedAnalysis } from "../analysis/types.js";
 import { type ObserverData } from "./data.js";
 /**
  * Resolve (and in a repo checkout, build) the Observer artifact before any run or
- * lab work starts. A missing artifact must cost seconds at startup, never a completed
+ * study work starts. A missing artifact must cost seconds at startup, never a completed
  * session (a live participant spends real money before the render step would have noticed).
  * Unconditional on purpose: the artifact is the only renderer now, so the fail-before-
  * spend property this preflight bought under the flag matters more, not less.

package/dist/observer/artifact.js

@@ -12,7 +12,7 @@ const OBSERVER_DATA_SLOT = `<script id="observer-data" type="application/json">$
 let cachedObserverArtifact = null;
 /**
  * Resolve (and in a repo checkout, build) the Observer artifact before any run or
- * lab work starts. A missing artifact must cost seconds at startup, never a completed
+ * study work starts. A missing artifact must cost seconds at startup, never a completed
  * session (a live participant spends real money before the render step would have noticed).
  * Unconditional on purpose: the artifact is the only renderer now, so the fail-before-
  * spend property this preflight bought under the flag matters more, not less.

package/dist/observer/data.d.ts

@@ -42,7 +42,7 @@ export interface ObserverData {
         /** The same thing as one readable line, so a renderer cannot accidentally show a number
          *  without its denominator. */
         participantsLine?: string;
-        /** The study's per-task completion rates, when the lab declared a protocol. */
+        /** The study's per-task completion rates, when the study declared a protocol. */
         tasks?: RunBundle["review"]["tasks"];
         /** Pre-formatted like participantsLine, denominator on every number. */
         tasksLine?: string;

package/dist/observer/library.d.ts

@@ -13,6 +13,8 @@ export interface LibraryHistory {
         estimatedCostUsd: number | null;
         costRatesAsOf: string | null;
         costPlaceholder: boolean;
+        /** The row's cost text: participants and desktops plus the run's analyses (src/run/run-cost.ts). */
+        costLabel: string | null;
     }>;
 }
 export interface LibraryRenderOptions {

package/dist/observer/library.js

@@ -127,7 +127,10 @@ function libraryClientJs() {
       var metaBits = [run.mode || "unknown", run.streamCount + " participants", run.createdAt || ""];
       // Labeled cost token: ALWAYS "~$X est." (never a bare "$X"), so the library never implies an
       // authoritative charge. Null = omitted (advisory, fail-open on display).
-      if (run.estimatedCostUsd != null) {
+      if (run.costLabel) {
+        metaBits.push(run.costLabel + (run.costPlaceholder ? " (placeholder)" : ""));
+        if (run.costRatesAsOf) meta.title = "estimated, rates as of " + run.costRatesAsOf;
+      } else if (run.estimatedCostUsd != null) {
         metaBits.push("~$" + Number(run.estimatedCostUsd).toFixed(2) + " est." + (run.costPlaceholder ? " (placeholder)" : ""));
         if (run.costRatesAsOf) meta.title = "estimated, rates as of " + run.costRatesAsOf;
       }

package/dist/observer/render.js

@@ -1,5 +1,7 @@
 import { spawn } from "node:child_process";
 import { runNotFoundMessage } from "../run/run-not-found.js";
+import { readAnalysisAccounting } from "../run/costs.js";
+import { analysisCostOf } from "../run/run-cost.js";
 import { accessSync, constants } from "node:fs";
 import { realpath } from "node:fs/promises";
 import { createServer } from "node:http";
@@ -92,7 +94,11 @@ export async function renderObserver(cwdInput, runInput, options = {}) {
         throw new Error("Observer output directory does not match the selected run.");
     }
     const observerPath = path.join(preparedRunPaths.physicalRunRoot, "observer", "index.html");
-    const analysis = projectShareCheckedAnalysis(await loadAnalysis(preparedRunPaths));
+    const spend = analysisCostOf(await readAnalysisAccounting(preparedRunPaths));
+    const analysis = {
+        ...projectShareCheckedAnalysis(await loadAnalysis(preparedRunPaths)),
+        ...(spend === null ? {} : { spend }),
+    };
     const observerData = buildObserverData(loaded.bundle);
     observerData.publicSafety.share = {
         status: verified.shareSafety.status,

package/dist/observer/run-routes.d.ts

@@ -27,6 +27,7 @@ export declare function buildHistoryIndex(proofRoot: PinnedDirectory, admitRun?:
         estimatedCostUsd: number | null;
         costRatesAsOf: string | null;
         costPlaceholder: boolean;
+        costLabel: string | null;
     }>;
 }>;
 /** internal: consumed by src/observer/serve.ts */

package/dist/observer/run-routes.js

@@ -5,34 +5,55 @@ import path from "node:path";
 import { isAnalysisRecordPath, projectShareCheckedAnalysis } from "../analysis/sharing.js";
 import { loadAnalysis } from "../analysis/load.js";
 import { listRuns } from "../run/stored-runs.js";
-import { bindExistingRunArtifactPaths, isPathInside, isSafeRunIdSegment } from "../run/paths.js";
+import { readAnalysisAccounting } from "../run/costs.js";
+import { bindExistingRunArtifactPaths, isPathInside, isSafeRunIdSegment, } from "../run/paths.js";
+import { analysisCostOf, runCost, runCostLabel } from "../run/run-cost.js";
 import { RUN_STATUS_FILE, RUN_STATUS_STALE_MS, isRunStatusRecord } from "../run/status.js";
 import { renderObserverHtml } from "./artifact.js";
 import { buildObserverData, recordedStreamEmbed, withObserverEndings, } from "./data.js";
 import { buildArtifactSecurityHeaders } from "./http.js";
 import { assertPinnedDirectory, openContainedFile, pinDirectChildDirectory, readContainedFile, relativeToRoot, sha256OfOpenedFile, } from "./pinned-files.js";
+/** The run's artifact paths, only when they are still the pinned directory being served. */
+async function servedRunPaths(runRoot) {
+    const runId = path.basename(runRoot.physicalPath);
+    const cwd = path.dirname(path.dirname(path.dirname(runRoot.physicalPath)));
+    const prepared = await bindExistingRunArtifactPaths(cwd, runId);
+    if (prepared.physicalRunRoot !== runRoot.physicalPath ||
+        prepared.runRootIdentity.birthtimeNs !== runRoot.birthtimeNs ||
+        prepared.runRootIdentity.dev !== runRoot.dev ||
+        prepared.runRootIdentity.ino !== runRoot.ino) {
+        throw new Error("ANALYSIS_STORAGE_CHANGED");
+    }
+    return prepared;
+}
 /** Analysis cannot grant filesystem authority or make an otherwise readable recording disappear. */
 async function readObserverAnalysis(runRoot) {
+    let loaded;
     try {
-        const runId = path.basename(runRoot.physicalPath);
-        const cwd = path.dirname(path.dirname(path.dirname(runRoot.physicalPath)));
-        const prepared = await bindExistingRunArtifactPaths(cwd, runId);
-        if (prepared.physicalRunRoot !== runRoot.physicalPath ||
-            prepared.runRootIdentity.birthtimeNs !== runRoot.birthtimeNs ||
-            prepared.runRootIdentity.dev !== runRoot.dev ||
-            prepared.runRootIdentity.ino !== runRoot.ino) {
-            throw new Error("ANALYSIS_STORAGE_CHANGED");
-        }
-        return projectShareCheckedAnalysis(await loadAnalysis(prepared));
+        loaded = projectShareCheckedAnalysis(await loadAnalysis(await servedRunPaths(runRoot)));
     }
     catch {
-        return {
+        loaded = {
             state: "invalid",
             analysis: null,
             corrections: [],
             warnings: ["Analysis could not be validated against this recording."],
         };
     }
+    const spend = await readServedAnalysisSpend(runRoot);
+    return spend === undefined ? loaded : { ...loaded, spend };
+}
+/**
+ * The run's analysis spend, read by the same reader as stats, so a request whose report could not
+ * be validated still counts. Undefined when the run sent none or its history cannot be bound.
+ */
+async function readServedAnalysisSpend(runRoot) {
+    try {
+        return analysisCostOf(await readAnalysisAccounting(await servedRunPaths(runRoot))) ?? undefined;
+    }
+    catch {
+        return undefined;
+    }
 }
 /** internal: consumed by src/observer/serve.ts */
 export async function serveRunPath(runRoot, relativePath, response, runtimeStreamUrls = [], request) {
@@ -219,6 +240,7 @@ admitRun) {
         if (admitRun && !root)
             return null;
         const data = root ? await readObserverData(root) : null;
+        const spend = root && data ? await readServedAnalysisSpend(root) : undefined;
         // listRuns read run.json before admission, so an admitted run's fields come from its
         // guarded read instead.
         return {
@@ -233,6 +255,9 @@ admitRun) {
             estimatedCostUsd: data?.cost?.estimatedTotalUsd ?? null,
             costRatesAsOf: data?.cost?.ratesAsOf ?? null,
             costPlaceholder: data?.cost?.placeholder ?? false,
+            // What the row shows: participants and desktops plus the run's analyses, as every
+            // surface that shows one run's cost reads it.
+            costLabel: data ? (runCostLabel(runCost(data.cost, spend)) ?? null) : null,
         };
     }));
     await assertPinnedDirectory(proofRoot);

package/dist/observer/serve.d.ts

@@ -41,7 +41,7 @@ export interface ServeResult {
 }
 interface ServeControlPlane {
     startRun?(request: {
-        labId: string;
+        studyId: string;
         dryRun: boolean;
     }): Promise<{
         accepted: boolean;