dsh-super-code 0.0.9

package/cordis.patch.yml

@@ -0,0 +1,21 @@
+# Register this package's preset directory in the host roster.
+# dsh-web-app already provides the agent-presets row; this patch replaces its
+# config so the package root is added without creating a second host registry.
+# DSH boots a profile from <profile>/cordis.yml and installs bundle packages in
+# that profile's node_modules (or the parent profiles fallback). Resolving from
+# that stable anchor works for both pnpm links and the packaged fallback; a bare
+# `presets/` path would incorrectly point beside the profile config instead.
+- id: agent-presets
+  name: '@deepseek-ai/dsh-agent-presets'
+  config:
+    default: super-code
+    roots:
+      - path: !!js "process.getBuiltinModule('node:url').fileURLToPath(new URL('node_modules/dsh-super-code/presets/', baseUrl))"
+        trust: system
+    includeShippedRoot: true
+    includeUserRoot: true
+
+# Register host projections and discover the browser contribution.
+- insert:
+    - id: super-code
+      name: dsh-super-code

package/lib/core/delegation.d.ts

@@ -0,0 +1,59 @@
+import type { TaskMemory } from './task-memory.js';
+export interface MemberAssignment {
+    owner: string;
+    attemptId: string;
+    objective: string;
+    ownedPaths: string[];
+    acceptance: string[];
+    maxTokens: number;
+}
+export interface MemberBrief extends MemberAssignment {
+    binding: string;
+    sessionId: string;
+    taskId: string;
+    createdAtSeq: number;
+    requirementsRevision: number;
+    delegationRevision: number;
+    workspace: string;
+    sourceVersion: string;
+    constraints: string[];
+    taskGoal: string;
+    taskAcceptance: string[];
+}
+export interface MemberResult {
+    binding: string;
+    owner: string;
+    attemptId: string;
+    outcome: 'completed' | 'failed' | 'cancelled' | 'stop_unknown';
+    summary: string;
+    artifacts: string[];
+    checks: string[];
+    unknowns: string[];
+    /** Optional self-report, never authenticated usage or budget enforcement. */
+    cost?: {
+        inputTokens: number;
+        outputTokens: number;
+        cachedTokens: number;
+        toolCalls: number;
+        complete: boolean;
+    };
+}
+export type MemberAccounting = {
+    kind: 'unavailable';
+    budget: 'unknown';
+} | {
+    kind: 'model-reported';
+    budget: 'unknown';
+    reportedBudget: 'within' | 'exceeded' | 'incomplete' | 'invalid';
+};
+export interface MemberReview {
+    reviewable: boolean;
+    reasons: string[];
+    accounting: MemberAccounting;
+    result: MemberResult;
+}
+/** Compact, version-bound inputs for the existing host subagent provider. */
+export declare function createMemberBrief(task: TaskMemory, sessionId: string, input: MemberAssignment): MemberBrief;
+/** Version matching is necessary for review, not proof that the work is correct. */
+export declare function validateMemberResult(task: TaskMemory, sessionId: string, assignment: MemberAssignment, input: MemberResult): MemberReview;
+//# sourceMappingURL=delegation.d.ts.map

package/lib/core/delegation.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"delegation.d.ts","sourceRoot":"","sources":["../../src/core/delegation.ts"],"names":[],"mappings":"AAEA,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,kBAAkB,CAAA;AAElD,MAAM,WAAW,gBAAgB;IAC/B,KAAK,EAAE,MAAM,CAAC;IAAC,SAAS,EAAE,MAAM,CAAC;IAAC,SAAS,EAAE,MAAM,CAAC;IAAC,UAAU,EAAE,MAAM,EAAE,CAAC;IAAC,UAAU,EAAE,MAAM,EAAE,CAAC;IAAC,SAAS,EAAE,MAAM,CAAA;CACnH;AACD,MAAM,WAAW,WAAY,SAAQ,gBAAgB;IACnD,OAAO,EAAE,MAAM,CAAC;IAAC,SAAS,EAAE,MAAM,CAAC;IAAC,MAAM,EAAE,MAAM,CAAC;IAAC,YAAY,EAAE,MAAM,CAAC;IAAC,oBAAoB,EAAE,MAAM,CAAC;IAAC,kBAAkB,EAAE,MAAM,CAAA;IAClI,SAAS,EAAE,MAAM,CAAC;IAAC,aAAa,EAAE,MAAM,CAAC;IAAC,WAAW,EAAE,MAAM,EAAE,CAAC;IAAC,QAAQ,EAAE,MAAM,CAAC;IAAC,cAAc,EAAE,MAAM,EAAE,CAAA;CAC5G;AACD,MAAM,WAAW,YAAY;IAC3B,OAAO,EAAE,MAAM,CAAC;IAAC,KAAK,EAAE,MAAM,CAAC;IAAC,SAAS,EAAE,MAAM,CAAA;IACjD,OAAO,EAAE,WAAW,GAAG,QAAQ,GAAG,WAAW,GAAG,cAAc,CAAA;IAC9D,OAAO,EAAE,MAAM,CAAC;IAAC,SAAS,EAAE,MAAM,EAAE,CAAC;IAAC,MAAM,EAAE,MAAM,EAAE,CAAC;IAAC,QAAQ,EAAE,MAAM,EAAE,CAAA;IAC1E,6EAA6E;IAC7E,IAAI,CAAC,EAAE;QAAE,WAAW,EAAE,MAAM,CAAC;QAAC,YAAY,EAAE,MAAM,CAAC;QAAC,YAAY,EAAE,MAAM,CAAC;QAAC,SAAS,EAAE,MAAM,CAAC;QAAC,QAAQ,EAAE,OAAO,CAAA;KAAE,CAAA;CACjH;AACD,MAAM,MAAM,gBAAgB,GAAG;IAAE,IAAI,EAAE,aAAa,CAAC;IAAC,MAAM,EAAE,SAAS,CAAA;CAAE,GACrE;IAAE,IAAI,EAAE,gBAAgB,CAAC;IAAC,MAAM,EAAE,SAAS,CAAC;IAAC,cAAc,EAAE,QAAQ,GAAG,UAAU,GAAG,YAAY,GAAG,SAAS,CAAA;CAAE,CAAA;AACnH,MAAM,WAAW,YAAY;IAAG,UAAU,EAAE,OAAO,CAAC;IAAC,OAAO,EAAE,MAAM,EAAE,CAAC;IAAC,UAAU,EAAE,gBAAgB,CAAC;IAAC,MAAM,EAAE,YAAY,CAAA;CAAE;AAM5H,6EAA6E;AAC7E,wBAAgB,iBAAiB,CAAC,IAAI,EAAE,UAAU,EAAE,SAAS,EAAE,MAAM,EAAE,KAAK,EAAE,gBAAgB,GAAG,WAAW,CAQ3G;AAED,oFAAoF;AACpF,wBAAgB,oBAAoB,CAAC,IAAI,EAAE,UAAU,EAAE,SAAS,EAAE,MAAM,EAAE,UAAU,EAAE,gBAAgB,EAAE,KAAK,EAAE,YAAY,GAAG,YAAY,CAiBzI"}

package/lib/core/delegation.js

@@ -0,0 +1,41 @@
+import { createHash } from 'node:crypto';
+import { z } from 'zod';
+const text = z.string().trim().min(1).max(4000);
+const count = z.number().int().nonnegative().safe();
+const assignmentSchema = z.object({ owner: text, attemptId: text, objective: text, ownedPaths: z.array(text).max(64), acceptance: z.array(text).min(1).max(32), maxTokens: z.number().int().positive() }).strict();
+const resultSchema = z.object({ binding: text, owner: text, attemptId: text, outcome: z.enum(['completed', 'failed', 'cancelled', 'stop_unknown']), summary: text, artifacts: z.array(text).max(64), checks: z.array(text).max(64), unknowns: z.array(text).max(32), cost: z.object({ inputTokens: count, outputTokens: count, cachedTokens: count, toolCalls: count, complete: z.boolean() }).strict().optional() }).strict();
+/** Compact, version-bound inputs for the existing host subagent provider. */
+export function createMemberBrief(task, sessionId, input) {
+    const assignment = assignmentSchema.parse(input);
+    if (sessionId.trim() === '')
+        throw new Error('sessionId is required');
+    if (task.status !== 'active')
+        throw new Error('Only active tasks may delegate work');
+    const fields = { ...assignment, sessionId, taskId: task.id, createdAtSeq: task.createdAtSeq, requirementsRevision: task.requirementsRevision,
+        delegationRevision: task.delegationRevision, taskGoal: task.goal, taskAcceptance: [...task.acceptance],
+        workspace: task.workspace, sourceVersion: task.sourceVersion, constraints: task.requirements.map(item => item.text) };
+    return { ...fields, binding: createHash('sha256').update(JSON.stringify(fields)).digest('hex') };
+}
+/** Version matching is necessary for review, not proof that the work is correct. */
+export function validateMemberResult(task, sessionId, assignment, input) {
+    assignment = assignmentSchema.parse(assignment);
+    const result = resultSchema.parse(input);
+    const reasons = [];
+    if (task.status !== 'active')
+        reasons.push('parent task is not active');
+    else {
+        const current = createMemberBrief(task, sessionId, assignment);
+        if (result.binding !== current.binding)
+            reasons.push('stale task, requirements, workspace, source or assignment');
+    }
+    if (result.owner !== assignment.owner || result.attemptId !== assignment.attemptId)
+        reasons.push('owner or attempt differs');
+    if (result.outcome !== 'completed')
+        reasons.push(result.outcome);
+    const cost = result.cost;
+    const accounting = cost === undefined ? { kind: 'unavailable', budget: 'unknown' }
+        : { kind: 'model-reported', budget: 'unknown', reportedBudget: cost.cachedTokens > cost.inputTokens || !Number.isSafeInteger(cost.inputTokens + cost.outputTokens) ? 'invalid'
+                : !cost.complete ? 'incomplete' : cost.inputTokens + cost.outputTokens > assignment.maxTokens ? 'exceeded' : 'within' };
+    return { reviewable: reasons.length === 0, reasons, accounting, result };
+}
+//# sourceMappingURL=delegation.js.map

package/lib/core/delegation.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"delegation.js","sourceRoot":"","sources":["../../src/core/delegation.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,UAAU,EAAE,MAAM,aAAa,CAAA;AACxC,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAA;AAoBvB,MAAM,IAAI,GAAG,CAAC,CAAC,MAAM,EAAE,CAAC,IAAI,EAAE,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,GAAG,CAAC,IAAI,CAAC,CAAA;AAC/C,MAAM,KAAK,GAAG,CAAC,CAAC,MAAM,EAAE,CAAC,GAAG,EAAE,CAAC,WAAW,EAAE,CAAC,IAAI,EAAE,CAAA;AACnD,MAAM,gBAAgB,GAAgC,CAAC,CAAC,MAAM,CAAC,EAAE,KAAK,EAAE,IAAI,EAAE,SAAS,EAAE,IAAI,EAAE,SAAS,EAAE,IAAI,EAAE,UAAU,EAAE,CAAC,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC,GAAG,CAAC,EAAE,CAAC,EAAE,UAAU,EAAE,CAAC,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,GAAG,CAAC,EAAE,CAAC,EAAE,SAAS,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,GAAG,EAAE,CAAC,QAAQ,EAAE,EAAE,CAAC,CAAC,MAAM,EAAE,CAAA;AAC/O,MAAM,YAAY,GAA4B,CAAC,CAAC,MAAM,CAAC,EAAE,OAAO,EAAE,IAAI,EAAE,KAAK,EAAE,IAAI,EAAE,SAAS,EAAE,IAAI,EAAE,OAAO,EAAE,CAAC,CAAC,IAAI,CAAC,CAAC,WAAW,EAAE,QAAQ,EAAE,WAAW,EAAE,cAAc,CAAC,CAAC,EAAE,OAAO,EAAE,IAAI,EAAE,SAAS,EAAE,CAAC,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC,GAAG,CAAC,EAAE,CAAC,EAAE,MAAM,EAAE,CAAC,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC,GAAG,CAAC,EAAE,CAAC,EAAE,QAAQ,EAAE,CAAC,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC,GAAG,CAAC,EAAE,CAAC,EAAE,IAAI,EAAE,CAAC,CAAC,MAAM,CAAC,EAAE,WAAW,EAAE,KAAK,EAAE,YAAY,EAAE,KAAK,EAAE,YAAY,EAAE,KAAK,EAAE,SAAS,EAAE,KAAK,EAAE,QAAQ,EAAE,CAAC,CAAC,OAAO,EAAE,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE,EAAE,CAAC,CAAC,MAAM,EAAE,CAAA;AAEvb,6EAA6E;AAC7E,MAAM,UAAU,iBAAiB,CAAC,IAAgB,EAAE,SAAiB,EAAE,KAAuB;IAC5F,MAAM,UAAU,GAAG,gBAAgB,CAAC,KAAK,CAAC,KAAK,CAAC,CAAA;IAChD,IAAI,SAAS,CAAC,IAAI,EAAE,KAAK,EAAE;QAAE,MAAM,IAAI,KAAK,CAAC,uBAAuB,CAAC,CAAA;IACrE,IAAI,IAAI,CAAC,MAAM,KAAK,QAAQ;QAAE,MAAM,IAAI,KAAK,CAAC,qCAAqC,CAAC,CAAA;IACpF,MAAM,MAAM,GAAG,EAAE,GAAG,UAAU,EAAE,SAAS,EAAE,MAAM,EAAE,IAAI,CAAC,EAAE,EAAE,YAAY,EAAE,IAAI,CAAC,YAAY,EAAE,oBAAoB,EAAE,IAAI,CAAC,oBAAoB;QAC1I,kBAAkB,EAAE,IAAI,CAAC,kBAAkB,EAAE,QAAQ,EAAE,IAAI,CAAC,IAAI,EAAE,cAAc,EAAE,CAAC,GAAG,IAAI,CAAC,UAAU,CAAC;QACtG,SAAS,EAAE,IAAI,CAAC,SAAS,EAAE,aAAa,EAAE,IAAI,CAAC,aAAa,EAAE,WAAW,EAAE,IAAI,CAAC,YAAY,CAAC,GAAG,CAAC,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,IAAI,CAAC,EAAE,CAAA;IACvH,OAAO,EAAE,GAAG,MAAM,EAAE,OAAO,EAAE,UAAU,CAAC,QAAQ,CAAC,CAAC,MAAM,CAAC,IAAI,CAAC,SAAS,CAAC,MAAM,CAAC,CAAC,CAAC,MAAM,CAAC,KAAK,CAAC,EAAE,CAAA;AAClG,CAAC;AAED,oFAAoF;AACpF,MAAM,UAAU,oBAAoB,CAAC,IAAgB,EAAE,SAAiB,EAAE,UAA4B,EAAE,KAAmB;IACzH,UAAU,GAAG,gBAAgB,CAAC,KAAK,CAAC,UAAU,CAAC,CAAA;IAC/C,MAAM,MAAM,GAAG,YAAY,CAAC,KAAK,CAAC,KAAK,CAAC,CAAA;IACxC,MAAM,OAAO,GAAa,EAAE,CAAA;IAC5B,IAAI,IAAI,CAAC,MAAM,KAAK,QAAQ;QAAE,OAAO,CAAC,IAAI,CAAC,2BAA2B,CAAC,CAAA;SAClE,CAAC;QACJ,MAAM,OAAO,GAAG,iBAAiB,CAAC,IAAI,EAAE,SAAS,EAAE,UAAU,CAAC,CAAA;QAC9D,IAAI,MAAM,CAAC,OAAO,KAAK,OAAO,CAAC,OAAO;YAAE,OAAO,CAAC,IAAI,CAAC,2DAA2D,CAAC,CAAA;IACnH,CAAC;IACD,IAAI,MAAM,CAAC,KAAK,KAAK,UAAU,CAAC,KAAK,IAAI,MAAM,CAAC,SAAS,KAAK,UAAU,CAAC,SAAS;QAAE,OAAO,CAAC,IAAI,CAAC,0BAA0B,CAAC,CAAA;IAC5H,IAAI,MAAM,CAAC,OAAO,KAAK,WAAW;QAAE,OAAO,CAAC,IAAI,CAAC,MAAM,CAAC,OAAO,CAAC,CAAA;IAChE,MAAM,IAAI,GAAG,MAAM,CAAC,IAAI,CAAA;IACxB,MAAM,UAAU,GAAqB,IAAI,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,IAAI,EAAE,aAAa,EAAE,MAAM,EAAE,SAAS,EAAE;QAClG,CAAC,CAAC,EAAE,IAAI,EAAE,gBAAgB,EAAE,MAAM,EAAE,SAAS,EAAE,cAAc,EAC3D,IAAI,CAAC,YAAY,GAAG,IAAI,CAAC,WAAW,IAAI,CAAC,MAAM,CAAC,aAAa,CAAC,IAAI,CAAC,WAAW,GAAG,IAAI,CAAC,YAAY,CAAC,CAAC,CAAC,CAAC,SAAS;gBAC7G,CAAC,CAAC,CAAC,IAAI,CAAC,QAAQ,CAAC,CAAC,CAAC,YAAY,CAAC,CAAC,CAAC,IAAI,CAAC,WAAW,GAAG,IAAI,CAAC,YAAY,GAAG,UAAU,CAAC,SAAS,CAAC,CAAC,CAAC,UAAU,CAAC,CAAC,CAAC,QAAQ,EAAE,CAAA;IAC7H,OAAO,EAAE,UAAU,EAAE,OAAO,CAAC,MAAM,KAAK,CAAC,EAAE,OAAO,EAAE,UAAU,EAAE,MAAM,EAAE,CAAA;AAC1E,CAAC"}

package/lib/core/evaluation.d.ts

@@ -0,0 +1,123 @@
+import { z } from 'zod';
+export interface EvaluationUsage {
+    uncachedInput: number;
+    cacheRead: number;
+    cacheWrite: number;
+    output: number;
+}
+export interface EvaluationIdentity {
+    preset: string;
+    version: string;
+    artifact: string;
+}
+export interface EvaluationConditions {
+    task: string;
+    evaluator: string;
+    model: string;
+    reasoning: string;
+    tools: string;
+    resources: string;
+    environment: string;
+    harness: string;
+}
+export interface EvaluationRate extends EvaluationUsage {
+    id: string;
+    provider: string;
+    model: string;
+    currency: string;
+    source: string;
+    effectiveAt: string;
+    perRequest: number;
+}
+export interface EvaluationRequest {
+    id: string;
+    sessionId: string;
+    attemptId: string;
+    kind: 'root' | 'member' | 'retry' | 'compaction';
+    provider: string;
+    model: string;
+    rateId?: string;
+    usage?: EvaluationUsage;
+    toolCharge?: number;
+}
+export interface SuperCodeManifest {
+    protocol: 'super-code/v3';
+    baseline: EvaluationIdentity;
+    candidate: EvaluationIdentity;
+    cases: {
+        taskId: string;
+        category: string;
+        conditions: EvaluationConditions;
+    }[];
+    gates: {
+        minAccuracyUplift: number;
+        maxLostSuccesses: number;
+        minCostReduction?: number;
+        minLatencyReduction?: number;
+    };
+    rates: EvaluationRate[];
+}
+export interface SuperCodeMeasurement {
+    protocol: 'super-code/v3';
+    side: 'baseline' | 'candidate';
+    identity: EvaluationIdentity;
+    taskId: string;
+    category: string;
+    conditions: EvaluationConditions;
+    mode: 'real' | 'mock' | 'replay';
+    outcome: 'passed' | 'failed' | 'agent_failed' | 'audit_failed' | 'infrastructure_error' | 'evaluation_error' | 'unknown';
+    latencyMs: number;
+    expectedRequests: string[];
+    requestManifestComplete: boolean;
+    requests: EvaluationRequest[];
+}
+export declare const evaluationManifestSchema: z.ZodType<SuperCodeManifest>;
+export declare const evaluationMeasurementSchema: z.ZodType<SuperCodeMeasurement>;
+export interface EvaluationSide {
+    recorded: number;
+    scored: number;
+    passed: number;
+    attempts: number;
+    requests: number;
+    expectedRequests: number;
+    latencyMs: number;
+    tokens: EvaluationUsage;
+    outcomes: Record<string, number>;
+    accounting: {
+        status: 'measured';
+        currency: string;
+        amount: number;
+    } | {
+        status: 'unmeasured';
+        reasons: string[];
+    };
+}
+export interface EvaluationPairs {
+    paired: number;
+    gained: string[];
+    lost: string[];
+    unchanged: string[];
+    unscored: string[];
+}
+export interface SuperCodeBatchResult {
+    status: 'incomplete' | 'unmatched' | 'unmeasured' | 'measured';
+    accepted: boolean;
+    reasons: string[];
+    planned: number;
+    baseline: EvaluationSide;
+    candidate: EvaluationSide;
+    pairs: EvaluationPairs;
+    categories: Record<string, {
+        paired: number;
+        gained: number;
+        lost: number;
+    }>;
+    checks: {
+        quality: boolean;
+        cost: boolean;
+        latency: boolean;
+    };
+}
+/** Fixed sides, fixed gates, full coverage. Partial reports never imply a release pass. */
+export declare function evaluateSuperCodeBatch(input: SuperCodeManifest, measurements: readonly SuperCodeMeasurement[]): SuperCodeBatchResult;
+//# sourceMappingURL=evaluation.d.ts.map

package/lib/core/evaluation.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"evaluation.d.ts","sourceRoot":"","sources":["../../src/core/evaluation.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAA;AAEvB,MAAM,WAAW,eAAe;IAAG,aAAa,EAAE,MAAM,CAAC;IAAC,SAAS,EAAE,MAAM,CAAC;IAAC,UAAU,EAAE,MAAM,CAAC;IAAC,MAAM,EAAE,MAAM,CAAA;CAAE;AACjH,MAAM,WAAW,kBAAkB;IAAG,MAAM,EAAE,MAAM,CAAC;IAAC,OAAO,EAAE,MAAM,CAAC;IAAC,QAAQ,EAAE,MAAM,CAAA;CAAE;AACzF,MAAM,WAAW,oBAAoB;IAAG,IAAI,EAAE,MAAM,CAAC;IAAC,SAAS,EAAE,MAAM,CAAC;IAAC,KAAK,EAAE,MAAM,CAAC;IAAC,SAAS,EAAE,MAAM,CAAC;IAAC,KAAK,EAAE,MAAM,CAAC;IAAC,SAAS,EAAE,MAAM,CAAC;IAAC,WAAW,EAAE,MAAM,CAAC;IAAC,OAAO,EAAE,MAAM,CAAA;CAAE;AACnL,MAAM,WAAW,cAAe,SAAQ,eAAe;IAAG,EAAE,EAAE,MAAM,CAAC;IAAC,QAAQ,EAAE,MAAM,CAAC;IAAC,KAAK,EAAE,MAAM,CAAC;IAAC,QAAQ,EAAE,MAAM,CAAC;IAAC,MAAM,EAAE,MAAM,CAAC;IAAC,WAAW,EAAE,MAAM,CAAC;IAAC,UAAU,EAAE,MAAM,CAAA;CAAE;AAClL,MAAM,WAAW,iBAAiB;IAChC,EAAE,EAAE,MAAM,CAAC;IAAC,SAAS,EAAE,MAAM,CAAC;IAAC,SAAS,EAAE,MAAM,CAAC;IAAC,IAAI,EAAE,MAAM,GAAG,QAAQ,GAAG,OAAO,GAAG,YAAY,CAAA;IAClG,QAAQ,EAAE,MAAM,CAAC;IAAC,KAAK,EAAE,MAAM,CAAC;IAAC,MAAM,CAAC,EAAE,MAAM,CAAC;IAAC,KAAK,CAAC,EAAE,eAAe,CAAC;IAAC,UAAU,CAAC,EAAE,MAAM,CAAA;CAC/F;AACD,MAAM,WAAW,iBAAiB;IAChC,QAAQ,EAAE,eAAe,CAAC;IAAC,QAAQ,EAAE,kBAAkB,CAAC;IAAC,SAAS,EAAE,kBAAkB,CAAA;IACtF,KAAK,EAAE;QAAE,MAAM,EAAE,MAAM,CAAC;QAAC,QAAQ,EAAE,MAAM,CAAC;QAAC,UAAU,EAAE,oBAAoB,CAAA;KAAE,EAAE,CAAA;IAC/E,KAAK,EAAE;QAAE,iBAAiB,EAAE,MAAM,CAAC;QAAC,gBAAgB,EAAE,MAAM,CAAC;QAAC,gBAAgB,CAAC,EAAE,MAAM,CAAC;QAAC,mBAAmB,CAAC,EAAE,MAAM,CAAA;KAAE,CAAA;IACvH,KAAK,EAAE,cAAc,EAAE,CAAA;CACxB;AACD,MAAM,WAAW,oBAAoB;IACnC,QAAQ,EAAE,eAAe,CAAC;IAAC,IAAI,EAAE,UAAU,GAAG,WAAW,CAAC;IAAC,QAAQ,EAAE,kBAAkB,CAAA;IACvF,MAAM,EAAE,MAAM,CAAC;IAAC,QAAQ,EAAE,MAAM,CAAC;IAAC,UAAU,EAAE,oBAAoB,CAAC;IAAC,IAAI,EAAE,MAAM,GAAG,MAAM,GAAG,QAAQ,CAAA;IACpG,OAAO,EAAE,QAAQ,GAAG,QAAQ,GAAG,cAAc,GAAG,cAAc,GAAG,sBAAsB,GAAG,kBAAkB,GAAG,SAAS,CAAA;IACxH,SAAS,EAAE,MAAM,CAAC;IAAC,gBAAgB,EAAE,MAAM,EAAE,CAAC;IAAC,uBAAuB,EAAE,OAAO,CAAC;IAAC,QAAQ,EAAE,iBAAiB,EAAE,CAAA;CAC/G;AAcD,eAAO,MAAM,wBAAwB,EAAE,CAAC,CAAC,OAAO,CAAC,iBAAiB,CAOvD,CAAA;AAMX,eAAO,MAAM,2BAA2B,EAAE,CAAC,CAAC,OAAO,CAAC,oBAAoB,CAI7D,CAAA;AACX,MAAM,WAAW,cAAc;IAC7B,QAAQ,EAAE,MAAM,CAAC;IAAC,MAAM,EAAE,MAAM,CAAC;IAAC,MAAM,EAAE,MAAM,CAAC;IAAC,QAAQ,EAAE,MAAM,CAAC;IAAC,QAAQ,EAAE,MAAM,CAAC;IAAC,gBAAgB,EAAE,MAAM,CAAA;IAC9G,SAAS,EAAE,MAAM,CAAC;IAAC,MAAM,EAAE,eAAe,CAAC;IAAC,QAAQ,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAA;IAC5E,UAAU,EAAE;QAAE,MAAM,EAAE,UAAU,CAAC;QAAC,QAAQ,EAAE,MAAM,CAAC;QAAC,MAAM,EAAE,MAAM,CAAA;KAAE,GAAG;QAAE,MAAM,EAAE,YAAY,CAAC;QAAC,OAAO,EAAE,MAAM,EAAE,CAAA;KAAE,CAAA;CACnH;AACD,MAAM,WAAW,eAAe;IAAG,MAAM,EAAE,MAAM,CAAC;IAAC,MAAM,EAAE,MAAM,EAAE,CAAC;IAAC,IAAI,EAAE,MAAM,EAAE,CAAC;IAAC,SAAS,EAAE,MAAM,EAAE,CAAC;IAAC,QAAQ,EAAE,MAAM,EAAE,CAAA;CAAE;AAC9H,MAAM,WAAW,oBAAoB;IACnC,MAAM,EAAE,YAAY,GAAG,WAAW,GAAG,YAAY,GAAG,UAAU,CAAC;IAAC,QAAQ,EAAE,OAAO,CAAC;IAAC,OAAO,EAAE,MAAM,EAAE,CAAA;IACpG,OAAO,EAAE,MAAM,CAAC;IAAC,QAAQ,EAAE,cAAc,CAAC;IAAC,SAAS,EAAE,cAAc,CAAC;IAAC,KAAK,EAAE,eAAe,CAAA;IAC5F,UAAU,EAAE,MAAM,CAAC,MAAM,EAAE;QAAE,MAAM,EAAE,MAAM,CAAC;QAAC,MAAM,EAAE,MAAM,CAAC;QAAC,IAAI,EAAE,MAAM,CAAA;KAAE,CAAC,CAAA;IAC5E,MAAM,EAAE;QAAE,OAAO,EAAE,OAAO,CAAC;QAAC,IAAI,EAAE,OAAO,CAAC;QAAC,OAAO,EAAE,OAAO,CAAA;KAAE,CAAA;CAC9D;AA4DD,2FAA2F;AAC3F,wBAAgB,sBAAsB,CAAC,KAAK,EAAE,iBAAiB,EAAE,YAAY,EAAE,SAAS,oBAAoB,EAAE,GAAG,oBAAoB,CAuDpI"}

package/lib/core/evaluation.js

@@ -0,0 +1,180 @@
+import { z } from 'zod';
+const label = z.string().trim().min(1);
+const count = z.number().int().nonnegative().safe();
+const amount = z.number().finite().nonnegative();
+const ratio = z.number().min(0).max(1);
+const sides = ['baseline', 'candidate'];
+const buckets = ['uncachedInput', 'cacheRead', 'cacheWrite', 'output'];
+const usageSchema = z.object({ uncachedInput: count, cacheRead: count, cacheWrite: count, output: count }).strict();
+const identitySchema = z.object({ preset: label, version: label, artifact: label }).strict();
+const conditionsSchema = z.object({ task: label, evaluator: label, model: label, reasoning: label, tools: label, resources: label, environment: label, harness: label }).strict();
+const rateSchema = z.object({ id: label, provider: label, model: label, currency: label, source: label, effectiveAt: label,
+    uncachedInput: amount, cacheRead: amount, cacheWrite: amount, output: amount, perRequest: amount }).strict();
+export const evaluationManifestSchema = z.object({
+    protocol: z.literal('super-code/v3'),
+    baseline: identitySchema, candidate: identitySchema,
+    cases: z.array(z.object({ taskId: label, category: label, conditions: conditionsSchema }).strict()).min(1),
+    // Explicit before the run: equivalence refactors need not claim a quality uplift.
+    gates: z.object({ minAccuracyUplift: ratio, maxLostSuccesses: count, minCostReduction: ratio.optional(), minLatencyReduction: ratio.optional() }).strict(),
+    rates: z.array(rateSchema),
+}).strict();
+const requestSchema = z.object({ id: label, sessionId: label, attemptId: label,
+    kind: z.enum(['root', 'member', 'retry', 'compaction']), provider: label, model: label,
+    rateId: label.optional(), usage: usageSchema.optional(), toolCharge: amount.optional(),
+}).strict();
+const outcomeSchema = z.enum(['passed', 'failed', 'agent_failed', 'audit_failed', 'infrastructure_error', 'evaluation_error', 'unknown']);
+export const evaluationMeasurementSchema = z.object({
+    protocol: z.literal('super-code/v3'), side: z.enum(sides), identity: identitySchema,
+    taskId: label, category: label, conditions: conditionsSchema, mode: z.enum(['real', 'mock', 'replay']), outcome: outcomeSchema,
+    latencyMs: amount, expectedRequests: z.array(label), requestManifestComplete: z.boolean(), requests: z.array(requestSchema),
+}).strict();
+function unique(values, subject) {
+    const result = new Set(values);
+    if (result.size !== values.length)
+        throw new Error(`Duplicate ${subject}`);
+    return result;
+}
+/** The adapter supplies a complete request manifest, including failed and maintenance calls. */
+function summarize(rows, rates) {
+    const tokens = { uncachedInput: 0, cacheRead: 0, cacheWrite: 0, output: 0 };
+    const attempts = new Set(), currencies = new Set(), reasons = [];
+    const outcomes = {};
+    const rateById = new Map(rates.map(rate => [rate.id, rate]));
+    let total = 0, requests = 0, expectedRequests = 0, latencyMs = 0, scored = 0, passed = 0;
+    for (const row of rows) {
+        outcomes[row.outcome] = (outcomes[row.outcome] ?? 0) + 1;
+        if (isScored(row)) {
+            scored++;
+            if (row.outcome === 'passed')
+                passed++;
+        }
+        latencyMs += row.latencyMs;
+        const expected = unique(row.expectedRequests, 'expected request id');
+        const observed = unique(row.requests.map(request => request.id), 'request id');
+        expectedRequests += expected.size;
+        requests += observed.size;
+        if (!row.requestManifestComplete)
+            reasons.push(`${row.taskId}: incomplete request manifest`);
+        if (row.outcome === 'passed' && expected.size === 0)
+            reasons.push(`${row.taskId}: passed without model request evidence`);
+        for (const id of expected)
+            if (!observed.has(id))
+                reasons.push(`${row.taskId}: missing request ${id}`);
+        for (const request of row.requests) {
+            if (!expected.has(request.id))
+                throw new Error(`Unexpected request ${request.id}`);
+            attempts.add(JSON.stringify([row.taskId, request.sessionId, request.attemptId]));
+            if (request.usage)
+                for (const key of buckets)
+                    tokens[key] += request.usage[key];
+            const rate = request.rateId === undefined ? undefined : rateById.get(request.rateId);
+            if (!rate || rate.provider !== request.provider || rate.model !== request.model) {
+                reasons.push(`${row.taskId}/${request.id}: missing or mismatched price identity`);
+                continue;
+            }
+            currencies.add(rate.currency);
+            if (!request.usage || request.toolCharge === undefined) {
+                reasons.push(`${row.taskId}/${request.id}: incomplete usage or tool charges`);
+                continue;
+            }
+            total += buckets.reduce((sum, key) => sum + request.usage[key] * rate[key] / 1_000_000, rate.perRequest + request.toolCharge);
+        }
+    }
+    // Empty cost is not automatically free. A single declared currency establishes the unit.
+    if (!currencies.size && !requests && rates.length)
+        for (const rate of rates)
+            currencies.add(rate.currency);
+    if (currencies.size !== 1)
+        reasons.push('No single known currency; no implicit exchange rates');
+    if (!Number.isFinite(total) || !Number.isFinite(latencyMs) || buckets.some(key => !Number.isSafeInteger(tokens[key])))
+        throw new Error('Evaluation totals exceed numeric capacity');
+    return { recorded: rows.length, scored, passed, attempts: attempts.size, requests, expectedRequests, latencyMs, tokens, outcomes,
+        accounting: reasons.length ? { status: 'unmeasured', reasons } : { status: 'measured', currency: [...currencies][0], amount: total } };
+}
+function isScored(row) {
+    return row.mode === 'real' && ['passed', 'failed', 'agent_failed', 'audit_failed'].includes(row.outcome);
+}
+function same(left, right) {
+    return Object.keys(left).every(key => left[key] === right[key]);
+}
+function meetsReduction(baseline, candidate, required) {
+    return baseline === 0 ? candidate === 0 && required === 0 : (baseline - candidate) / baseline + 1e-12 >= required;
+}
+/** Fixed sides, fixed gates, full coverage. Partial reports never imply a release pass. */
+export function evaluateSuperCodeBatch(input, measurements) {
+    const manifest = evaluationManifestSchema.parse(input);
+    unique(manifest.cases.map(item => item.taskId), 'manifest task id');
+    unique(manifest.rates.map(item => item.id), 'price id');
+    const rows = measurements.map(row => evaluationMeasurementSchema.parse(row));
+    const cases = new Map(manifest.cases.map(item => [item.taskId, item]));
+    const groups = new Map();
+    const mismatches = [], missing = [], unmeasured = [];
+    const requestIdentities = new Set();
+    for (const row of rows) {
+        const expected = cases.get(row.taskId);
+        if (!expected)
+            throw new Error(`Unexpected task ${row.taskId}`);
+        const group = groups.get(row.taskId) ?? {};
+        if (group[row.side])
+            throw new Error(`Duplicate ${row.side} measurement for ${row.taskId}`);
+        group[row.side] = row;
+        groups.set(row.taskId, group);
+        if (row.category !== expected.category || !same(row.conditions, expected.conditions) || !same(row.identity, manifest[row.side]))
+            mismatches.push(`${row.taskId}/${row.side}: comparison identity differs`);
+        if (!isScored(row))
+            unmeasured.push(`${row.taskId}/${row.side}: ${row.mode}/${row.outcome}`);
+        for (const request of row.requests) {
+            const identity = JSON.stringify([request.sessionId, request.id]);
+            if (requestIdentities.has(identity))
+                throw new Error('Duplicate session request across measurements');
+            requestIdentities.add(identity);
+        }
+    }
+    const baseline = summarize(rows.filter(row => row.side === 'baseline'), manifest.rates);
+    const candidate = summarize(rows.filter(row => row.side === 'candidate'), manifest.rates);
+    const pairs = { paired: 0, gained: [], lost: [], unchanged: [], unscored: [] };
+    const categories = Object.create(null);
+    for (const item of manifest.cases) {
+        const group = groups.get(item.taskId);
+        for (const side of sides)
+            if (!group?.[side])
+                missing.push(`${item.taskId}/${side}`);
+        const a = group?.baseline, b = group?.candidate;
+        if (!a || !b || !isScored(a) || !isScored(b)) {
+            pairs.unscored.push(item.taskId);
+            continue;
+        }
+        pairs.paired++;
+        const category = categories[item.category] ??= { paired: 0, gained: 0, lost: 0 };
+        category.paired++;
+        if (a.outcome !== 'passed' && b.outcome === 'passed') {
+            pairs.gained.push(item.taskId);
+            category.gained++;
+        }
+        else if (a.outcome === 'passed' && b.outcome !== 'passed') {
+            pairs.lost.push(item.taskId);
+            category.lost++;
+        }
+        else
+            pairs.unchanged.push(item.taskId);
+    }
+    for (const [side, summary] of [['baseline', baseline], ['candidate', candidate]]) {
+        if (summary.accounting.status === 'unmeasured')
+            unmeasured.push(...summary.accounting.reasons.map(reason => `${side}: ${reason}`));
+    }
+    const a = baseline.accounting, b = candidate.accounting;
+    const comparableCost = a.status === 'measured' && b.status === 'measured' && a.currency === b.currency;
+    if (a.status === 'measured' && b.status === 'measured' && a.currency !== b.currency)
+        unmeasured.push('Comparison currencies differ');
+    const status = mismatches.length ? 'unmatched' : missing.length ? 'incomplete' : unmeasured.length ? 'unmeasured' : 'measured';
+    const gates = manifest.gates;
+    const checks = {
+        quality: pairs.paired === manifest.cases.length && (pairs.gained.length - pairs.lost.length) / pairs.paired + 1e-12 >= gates.minAccuracyUplift
+            && pairs.lost.length <= gates.maxLostSuccesses && Object.values(categories).every(category => category.gained >= category.lost),
+        cost: comparableCost && (gates.minCostReduction === undefined || meetsReduction(a.amount, b.amount, gates.minCostReduction)),
+        latency: gates.minLatencyReduction === undefined || meetsReduction(baseline.latencyMs, candidate.latencyMs, gates.minLatencyReduction),
+    };
+    return { status, accepted: status === 'measured' && Object.values(checks).every(Boolean), reasons: [...mismatches, ...missing, ...unmeasured],
+        planned: manifest.cases.length, baseline, candidate, pairs, categories, checks };
+}
+//# sourceMappingURL=evaluation.js.map

package/lib/core/evaluation.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"evaluation.js","sourceRoot":"","sources":["../../src/core/evaluation.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAA;AAuBvB,MAAM,KAAK,GAAG,CAAC,CAAC,MAAM,EAAE,CAAC,IAAI,EAAE,CAAC,GAAG,CAAC,CAAC,CAAC,CAAA;AACtC,MAAM,KAAK,GAAG,CAAC,CAAC,MAAM,EAAE,CAAC,GAAG,EAAE,CAAC,WAAW,EAAE,CAAC,IAAI,EAAE,CAAA;AACnD,MAAM,MAAM,GAAG,CAAC,CAAC,MAAM,EAAE,CAAC,MAAM,EAAE,CAAC,WAAW,EAAE,CAAA;AAChD,MAAM,KAAK,GAAG,CAAC,CAAC,MAAM,EAAE,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAA;AACtC,MAAM,KAAK,GAAG,CAAC,UAAU,EAAE,WAAW,CAAU,CAAA;AAEhD,MAAM,OAAO,GAAG,CAAC,eAAe,EAAE,WAAW,EAAE,YAAY,EAAE,QAAQ,CAAU,CAAA;AAC/E,MAAM,WAAW,GAAG,CAAC,CAAC,MAAM,CAAC,EAAE,aAAa,EAAE,KAAK,EAAE,SAAS,EAAE,KAAK,EAAE,UAAU,EAAE,KAAK,EAAE,MAAM,EAAE,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE,CAAA;AACnH,MAAM,cAAc,GAAG,CAAC,CAAC,MAAM,CAAC,EAAE,MAAM,EAAE,KAAK,EAAE,OAAO,EAAE,KAAK,EAAE,QAAQ,EAAE,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE,CAAA;AAC5F,MAAM,gBAAgB,GAAG,CAAC,CAAC,MAAM,CAAC,EAAE,IAAI,EAAE,KAAK,EAAE,SAAS,EAAE,KAAK,EAAE,KAAK,EAAE,KAAK,EAAE,SAAS,EAAE,KAAK,EAAE,KAAK,EAAE,KAAK,EAAE,SAAS,EAAE,KAAK,EAAE,WAAW,EAAE,KAAK,EAAE,OAAO,EAAE,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE,CAAA;AACjL,MAAM,UAAU,GAAG,CAAC,CAAC,MAAM,CAAC,EAAE,EAAE,EAAE,KAAK,EAAE,QAAQ,EAAE,KAAK,EAAE,KAAK,EAAE,KAAK,EAAE,QAAQ,EAAE,KAAK,EAAE,MAAM,EAAE,KAAK,EAAE,WAAW,EAAE,KAAK;IACxH,aAAa,EAAE,MAAM,EAAE,SAAS,EAAE,MAAM,EAAE,UAAU,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,EAAE,UAAU,EAAE,MAAM,EAAE,CAAC,CAAC,MAAM,EAAE,CAAA;AAC9G,MAAM,CAAC,MAAM,wBAAwB,GAAiC,CAAC,CAAC,MAAM,CAAC;IAC7E,QAAQ,EAAE,CAAC,CAAC,OAAO,CAAC,eAAe,CAAC;IACpC,QAAQ,EAAE,cAAc,EAAE,SAAS,EAAE,cAAc;IACnD,KAAK,EAAE,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,MAAM,CAAC,EAAE,MAAM,EAAE,KAAK,EAAE,QAAQ,EAAE,KAAK,EAAE,UAAU,EAAE,gBAAgB,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC;IAC1G,kFAAkF;IAClF,KAAK,EAAE,CAAC,CAAC,MAAM,CAAC,EAAE,iBAAiB,EAAE,KAAK,EAAE,gBAAgB,EAAE,KAAK,EAAE,gBAAgB,EAAE,KAAK,CAAC,QAAQ,EAAE,EAAE,mBAAmB,EAAE,KAAK,CAAC,QAAQ,EAAE,EAAE,CAAC,CAAC,MAAM,EAAE;IAC1J,KAAK,EAAE,CAAC,CAAC,KAAK,CAAC,UAAU,CAAC;CAC3B,CAAC,CAAC,MAAM,EAAE,CAAA;AACX,MAAM,aAAa,GAAG,CAAC,CAAC,MAAM,CAAC,EAAE,EAAE,EAAE,KAAK,EAAE,SAAS,EAAE,KAAK,EAAE,SAAS,EAAE,KAAK;IAC5E,IAAI,EAAE,CAAC,CAAC,IAAI,CAAC,CAAC,MAAM,EAAE,QAAQ,EAAE,OAAO,EAAE,YAAY,CAAC,CAAC,EAAE,QAAQ,EAAE,KAAK,EAAE,KAAK,EAAE,KAAK;IACtF,MAAM,EAAE,KAAK,CAAC,QAAQ,EAAE,EAAE,KAAK,EAAE,WAAW,CAAC,QAAQ,EAAE,EAAE,UAAU,EAAE,MAAM,CAAC,QAAQ,EAAE;CACvF,CAAC,CAAC,MAAM,EAAE,CAAA;AACX,MAAM,aAAa,GAAG,CAAC,CAAC,IAAI,CAAC,CAAC,QAAQ,EAAE,QAAQ,EAAE,cAAc,EAAE,cAAc,EAAE,sBAAsB,EAAE,kBAAkB,EAAE,SAAS,CAAC,CAAC,CAAA;AACzI,MAAM,CAAC,MAAM,2BAA2B,GAAoC,CAAC,CAAC,MAAM,CAAC;IACnF,QAAQ,EAAE,CAAC,CAAC,OAAO,CAAC,eAAe,CAAC,EAAE,IAAI,EAAE,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,EAAE,QAAQ,EAAE,cAAc;IACnF,MAAM,EAAE,KAAK,EAAE,QAAQ,EAAE,KAAK,EAAE,UAAU,EAAE,gBAAgB,EAAE,IAAI,EAAE,CAAC,CAAC,IAAI,CAAC,CAAC,MAAM,EAAE,MAAM,EAAE,QAAQ,CAAC,CAAC,EAAE,OAAO,EAAE,aAAa;IAC9H,SAAS,EAAE,MAAM,EAAE,gBAAgB,EAAE,CAAC,CAAC,KAAK,CAAC,KAAK,CAAC,EAAE,uBAAuB,EAAE,CAAC,CAAC,OAAO,EAAE,EAAE,QAAQ,EAAE,CAAC,CAAC,KAAK,CAAC,aAAa,CAAC;CAC5H,CAAC,CAAC,MAAM,EAAE,CAAA;AAcX,SAAS,MAAM,CAAC,MAAyB,EAAE,OAAe;IACxD,MAAM,MAAM,GAAG,IAAI,GAAG,CAAC,MAAM,CAAC,CAAA;IAC9B,IAAI,MAAM,CAAC,IAAI,KAAK,MAAM,CAAC,MAAM;QAAE,MAAM,IAAI,KAAK,CAAC,aAAa,OAAO,EAAE,CAAC,CAAA;IAC1E,OAAO,MAAM,CAAA;AACf,CAAC;AAED,gGAAgG;AAChG,SAAS,SAAS,CAAC,IAAqC,EAAE,KAAiC;IACzF,MAAM,MAAM,GAAoB,EAAE,aAAa,EAAE,CAAC,EAAE,SAAS,EAAE,CAAC,EAAE,UAAU,EAAE,CAAC,EAAE,MAAM,EAAE,CAAC,EAAE,CAAA;IAC5F,MAAM,QAAQ,GAAG,IAAI,GAAG,EAAU,EAAE,UAAU,GAAG,IAAI,GAAG,EAAU,EAAE,OAAO,GAAa,EAAE,CAAA;IAC1F,MAAM,QAAQ,GAA2B,EAAE,CAAA;IAC3C,MAAM,QAAQ,GAAG,IAAI,GAAG,CAAC,KAAK,CAAC,GAAG,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC,IAAI,CAAC,EAAE,EAAE,IAAI,CAAC,CAAC,CAAC,CAAA;IAC5D,IAAI,KAAK,GAAG,CAAC,EAAE,QAAQ,GAAG,CAAC,EAAE,gBAAgB,GAAG,CAAC,EAAE,SAAS,GAAG,CAAC,EAAE,MAAM,GAAG,CAAC,EAAE,MAAM,GAAG,CAAC,CAAA;IACxF,KAAK,MAAM,GAAG,IAAI,IAAI,EAAE,CAAC;QACvB,QAAQ,CAAC,GAAG,CAAC,OAAO,CAAC,GAAG,CAAC,QAAQ,CAAC,GAAG,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC,GAAG,CAAC,CAAA;QACxD,IAAI,QAAQ,CAAC,GAAG,CAAC,EAAE,CAAC;YAAC,MAAM,EAAE,CAAC;YAAC,IAAI,GAAG,CAAC,OAAO,KAAK,QAAQ;gBAAE,MAAM,EAAE,CAAA;QAAC,CAAC;QACvE,SAAS,IAAI,GAAG,CAAC,SAAS,CAAA;QAC1B,MAAM,QAAQ,GAAG,MAAM,CAAC,GAAG,CAAC,gBAAgB,EAAE,qBAAqB,CAAC,CAAA;QACpE,MAAM,QAAQ,GAAG,MAAM,CAAC,GAAG,CAAC,QAAQ,CAAC,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,YAAY,CAAC,CAAA;QAC9E,gBAAgB,IAAI,QAAQ,CAAC,IAAI,CAAC;QAAC,QAAQ,IAAI,QAAQ,CAAC,IAAI,CAAA;QAC5D,IAAI,CAAC,GAAG,CAAC,uBAAuB;YAAE,OAAO,CAAC,IAAI,CAAC,GAAG,GAAG,CAAC,MAAM,+BAA+B,CAAC,CAAA;QAC5F,IAAI,GAAG,CAAC,OAAO,KAAK,QAAQ,IAAI,QAAQ,CAAC,IAAI,KAAK,CAAC;YAAE,OAAO,CAAC,IAAI,CAAC,GAAG,GAAG,CAAC,MAAM,yCAAyC,CAAC,CAAA;QACzH,KAAK,MAAM,EAAE,IAAI,QAAQ;YAAE,IAAI,CAAC,QAAQ,CAAC,GAAG,CAAC,EAAE,CAAC;gBAAE,OAAO,CAAC,IAAI,CAAC,GAAG,GAAG,CAAC,MAAM,qBAAqB,EAAE,EAAE,CAAC,CAAA;QACtG,KAAK,MAAM,OAAO,IAAI,GAAG,CAAC,QAAQ,EAAE,CAAC;YACnC,IAAI,CAAC,QAAQ,CAAC,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC;gBAAE,MAAM,IAAI,KAAK,CAAC,sBAAsB,OAAO,CAAC,EAAE,EAAE,CAAC,CAAA;YAClF,QAAQ,CAAC,GAAG,CAAC,IAAI,CAAC,SAAS,CAAC,CAAC,GAAG,CAAC,MAAM,EAAE,OAAO,CAAC,SAAS,EAAE,OAAO,CAAC,SAAS,CAAC,CAAC,CAAC,CAAA;YAChF,IAAI,OAAO,CAAC,KAAK;gBAAE,KAAK,MAAM,GAAG,IAAI,OAAO;oBAAE,MAAM,CAAC,GAAG,CAAC,IAAI,OAAO,CAAC,KAAK,CAAC,GAAG,CAAC,CAAA;YAC/E,MAAM,IAAI,GAAG,OAAO,CAAC,MAAM,KAAK,SAAS,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,QAAQ,CAAC,GAAG,CAAC,OAAO,CAAC,MAAM,CAAC,CAAA;YACpF,IAAI,CAAC,IAAI,IAAI,IAAI,CAAC,QAAQ,KAAK,OAAO,CAAC,QAAQ,IAAI,IAAI,CAAC,KAAK,KAAK,OAAO,CAAC,KAAK,EAAE,CAAC;gBAChF,OAAO,CAAC,IAAI,CAAC,GAAG,GAAG,CAAC,MAAM,IAAI,OAAO,CAAC,EAAE,wCAAwC,CAAC,CAAC;gBAAC,SAAQ;YAC7F,CAAC;YACD,UAAU,CAAC,GAAG,CAAC,IAAI,CAAC,QAAQ,CAAC,CAAA;YAC7B,IAAI,CAAC,OAAO,CAAC,KAAK,IAAI,OAAO,CAAC,UAAU,KAAK,SAAS,EAAE,CAAC;gBACvD,OAAO,CAAC,IAAI,CAAC,GAAG,GAAG,CAAC,MAAM,IAAI,OAAO,CAAC,EAAE,oCAAoC,CAAC,CAAC;gBAAC,SAAQ;YACzF,CAAC;YACD,KAAK,IAAI,OAAO,CAAC,MAAM,CAAC,CAAC,GAAG,EAAE,GAAG,EAAE,EAAE,CAAC,GAAG,GAAG,OAAO,CAAC,KAAM,CAAC,GAAG,CAAC,GAAG,IAAI,CAAC,GAAG,CAAC,GAAG,SAAS,EAAE,IAAI,CAAC,UAAU,GAAG,OAAO,CAAC,UAAU,CAAC,CAAA;QAChI,CAAC;IACH,CAAC;IACD,yFAAyF;IACzF,IAAI,CAAC,UAAU,CAAC,IAAI,IAAI,CAAC,QAAQ,IAAI,KAAK,CAAC,MAAM;QAAE,KAAK,MAAM,IAAI,IAAI,KAAK;YAAE,UAAU,CAAC,GAAG,CAAC,IAAI,CAAC,QAAQ,CAAC,CAAA;IAC1G,IAAI,UAAU,CAAC,IAAI,KAAK,CAAC;QAAE,OAAO,CAAC,IAAI,CAAC,sDAAsD,CAAC,CAAA;IAC/F,IAAI,CAAC,MAAM,CAAC,QAAQ,CAAC,KAAK,CAAC,IAAI,CAAC,MAAM,CAAC,QAAQ,CAAC,SAAS,CAAC,IAAI,OAAO,CAAC,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,CAAC,MAAM,CAAC,aAAa,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC;QAAE,MAAM,IAAI,KAAK,CAAC,2CAA2C,CAAC,CAAA;IACnL,OAAO,EAAE,QAAQ,EAAE,IAAI,CAAC,MAAM,EAAE,MAAM,EAAE,MAAM,EAAE,QAAQ,EAAE,QAAQ,CAAC,IAAI,EAAE,QAAQ,EAAE,gBAAgB,EAAE,SAAS,EAAE,MAAM,EAAE,QAAQ;QAC9H,UAAU,EAAE,OAAO,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,MAAM,EAAE,YAAY,EAAE,OAAO,EAAE,CAAC,CAAC,CAAC,EAAE,MAAM,EAAE,UAAU,EAAE,QAAQ,EAAE,CAAC,GAAG,UAAU,CAAC,CAAC,CAAC,CAAE,EAAE,MAAM,EAAE,KAAK,EAAE,EAAE,CAAA;AAC3I,CAAC;AAED,SAAS,QAAQ,CAAC,GAAyB;IACzC,OAAO,GAAG,CAAC,IAAI,KAAK,MAAM,IAAI,CAAC,QAAQ,EAAE,QAAQ,EAAE,cAAc,EAAE,cAAc,CAAC,CAAC,QAAQ,CAAC,GAAG,CAAC,OAAO,CAAC,CAAA;AAC1G,CAAC;AAED,SAAS,IAAI,CAAmB,IAAO,EAAE,KAAQ;IAC/C,OAAQ,MAAM,CAAC,IAAI,CAAC,IAAI,CAAiB,CAAC,KAAK,CAAC,GAAG,CAAC,EAAE,CAAC,IAAI,CAAC,GAAG,CAAC,KAAK,KAAK,CAAC,GAAG,CAAC,CAAC,CAAA;AAClF,CAAC;AAED,SAAS,cAAc,CAAC,QAAgB,EAAE,SAAiB,EAAE,QAAgB;IAC3E,OAAO,QAAQ,KAAK,CAAC,CAAC,CAAC,CAAC,SAAS,KAAK,CAAC,IAAI,QAAQ,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,QAAQ,GAAG,SAAS,CAAC,GAAG,QAAQ,GAAG,KAAK,IAAI,QAAQ,CAAA;AACnH,CAAC;AAED,2FAA2F;AAC3F,MAAM,UAAU,sBAAsB,CAAC,KAAwB,EAAE,YAA6C;IAC5G,MAAM,QAAQ,GAAG,wBAAwB,CAAC,KAAK,CAAC,KAAK,CAAC,CAAA;IACtD,MAAM,CAAC,QAAQ,CAAC,KAAK,CAAC,GAAG,CAAC,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,MAAM,CAAC,EAAE,kBAAkB,CAAC,CAAA;IACnE,MAAM,CAAC,QAAQ,CAAC,KAAK,CAAC,GAAG,CAAC,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,EAAE,CAAC,EAAE,UAAU,CAAC,CAAA;IACvD,MAAM,IAAI,GAAG,YAAY,CAAC,GAAG,CAAC,GAAG,CAAC,EAAE,CAAC,2BAA2B,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,CAAA;IAC5E,MAAM,KAAK,GAAG,IAAI,GAAG,CAAC,QAAQ,CAAC,KAAK,CAAC,GAAG,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC,IAAI,CAAC,MAAM,EAAE,IAAI,CAAC,CAAC,CAAC,CAAA;IACtE,MAAM,MAAM,GAAG,IAAI,GAAG,EAAuD,CAAA;IAC7E,MAAM,UAAU,GAAa,EAAE,EAAE,OAAO,GAAa,EAAE,EAAE,UAAU,GAAa,EAAE,CAAA;IAClF,MAAM,iBAAiB,GAAG,IAAI,GAAG,EAAU,CAAA;IAC3C,KAAK,MAAM,GAAG,IAAI,IAAI,EAAE,CAAC;QACvB,MAAM,QAAQ,GAAG,KAAK,CAAC,GAAG,CAAC,GAAG,CAAC,MAAM,CAAC,CAAA;QACtC,IAAI,CAAC,QAAQ;YAAE,MAAM,IAAI,KAAK,CAAC,mBAAmB,GAAG,CAAC,MAAM,EAAE,CAAC,CAAA;QAC/D,MAAM,KAAK,GAAG,MAAM,CAAC,GAAG,CAAC,GAAG,CAAC,MAAM,CAAC,IAAI,EAAE,CAAA;QAC1C,IAAI,KAAK,CAAC,GAAG,CAAC,IAAI,CAAC;YAAE,MAAM,IAAI,KAAK,CAAC,aAAa,GAAG,CAAC,IAAI,oBAAoB,GAAG,CAAC,MAAM,EAAE,CAAC,CAAA;QAC3F,KAAK,CAAC,GAAG,CAAC,IAAI,CAAC,GAAG,GAAG,CAAC;QAAC,MAAM,CAAC,GAAG,CAAC,GAAG,CAAC,MAAM,EAAE,KAAK,CAAC,CAAA;QACpD,IAAI,GAAG,CAAC,QAAQ,KAAK,QAAQ,CAAC,QAAQ,IAAI,CAAC,IAAI,CAAC,GAAG,CAAC,UAAU,EAAE,QAAQ,CAAC,UAAU,CAAC,IAAI,CAAC,IAAI,CAAC,GAAG,CAAC,QAAQ,EAAE,QAAQ,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC;YAAE,UAAU,CAAC,IAAI,CAAC,GAAG,GAAG,CAAC,MAAM,IAAI,GAAG,CAAC,IAAI,+BAA+B,CAAC,CAAA;QAC1M,IAAI,CAAC,QAAQ,CAAC,GAAG,CAAC;YAAE,UAAU,CAAC,IAAI,CAAC,GAAG,GAAG,CAAC,MAAM,IAAI,GAAG,CAAC,IAAI,KAAK,GAAG,CAAC,IAAI,IAAI,GAAG,CAAC,OAAO,EAAE,CAAC,CAAA;QAC5F,KAAK,MAAM,OAAO,IAAI,GAAG,CAAC,QAAQ,EAAE,CAAC;YACnC,MAAM,QAAQ,GAAG,IAAI,CAAC,SAAS,CAAC,CAAC,OAAO,CAAC,SAAS,EAAE,OAAO,CAAC,EAAE,CAAC,CAAC,CAAA;YAChE,IAAI,iBAAiB,CAAC,GAAG,CAAC,QAAQ,CAAC;gBAAE,MAAM,IAAI,KAAK,CAAC,+CAA+C,CAAC,CAAA;YACrG,iBAAiB,CAAC,GAAG,CAAC,QAAQ,CAAC,CAAA;QACjC,CAAC;IACH,CAAC;IACD,MAAM,QAAQ,GAAG,SAAS,CAAC,IAAI,CAAC,MAAM,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,CAAC,IAAI,KAAK,UAAU,CAAC,EAAE,QAAQ,CAAC,KAAK,CAAC,CAAA;IACvF,MAAM,SAAS,GAAG,SAAS,CAAC,IAAI,CAAC,MAAM,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,CAAC,IAAI,KAAK,WAAW,CAAC,EAAE,QAAQ,CAAC,KAAK,CAAC,CAAA;IACzF,MAAM,KAAK,GAAoB,EAAE,MAAM,EAAE,CAAC,EAAE,MAAM,EAAE,EAAE,EAAE,IAAI,EAAE,EAAE,EAAE,SAAS,EAAE,EAAE,EAAE,QAAQ,EAAE,EAAE,EAAE,CAAA;IAC/F,MAAM,UAAU,GAAuC,MAAM,CAAC,MAAM,CAAC,IAAI,CAAC,CAAA;IAC1E,KAAK,MAAM,IAAI,IAAI,QAAQ,CAAC,KAAK,EAAE,CAAC;QAClC,MAAM,KAAK,GAAG,MAAM,CAAC,GAAG,CAAC,IAAI,CAAC,MAAM,CAAC,CAAA;QACrC,KAAK,MAAM,IAAI,IAAI,KAAK;YAAE,IAAI,CAAC,KAAK,EAAE,CAAC,IAAI,CAAC;gBAAE,OAAO,CAAC,IAAI,CAAC,GAAG,IAAI,CAAC,MAAM,IAAI,IAAI,EAAE,CAAC,CAAA;QACpF,MAAM,CAAC,GAAG,KAAK,EAAE,QAAQ,EAAE,CAAC,GAAG,KAAK,EAAE,SAAS,CAAA;QAC/C,IAAI,CAAC,CAAC,IAAI,CAAC,CAAC,IAAI,CAAC,QAAQ,CAAC,CAAC,CAAC,IAAI,CAAC,QAAQ,CAAC,CAAC,CAAC,EAAE,CAAC;YAAC,KAAK,CAAC,QAAQ,CAAC,IAAI,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;YAAC,SAAQ;QAAC,CAAC;QAC5F,KAAK,CAAC,MAAM,EAAE,CAAA;QACd,MAAM,QAAQ,GAAG,UAAU,CAAC,IAAI,CAAC,QAAQ,CAAC,KAAK,EAAE,MAAM,EAAE,CAAC,EAAE,MAAM,EAAE,CAAC,EAAE,IAAI,EAAE,CAAC,EAAE,CAAA;QAChF,QAAQ,CAAC,MAAM,EAAE,CAAA;QACjB,IAAI,CAAC,CAAC,OAAO,KAAK,QAAQ,IAAI,CAAC,CAAC,OAAO,KAAK,QAAQ,EAAE,CAAC;YAAC,KAAK,CAAC,MAAM,CAAC,IAAI,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;YAAC,QAAQ,CAAC,MAAM,EAAE,CAAA;QAAC,CAAC;aACtG,IAAI,CAAC,CAAC,OAAO,KAAK,QAAQ,IAAI,CAAC,CAAC,OAAO,KAAK,QAAQ,EAAE,CAAC;YAAC,KAAK,CAAC,IAAI,CAAC,IAAI,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;YAAC,QAAQ,CAAC,IAAI,EAAE,CAAA;QAAC,CAAC;;YACvG,KAAK,CAAC,SAAS,CAAC,IAAI,CAAC,IAAI,CAAC,MAAM,CAAC,CAAA;IACxC,CAAC;IACD,KAAK,MAAM,CAAC,IAAI,EAAE,OAAO,CAAC,IAAI,CAAC,CAAC,UAAU,EAAE,QAAQ,CAAC,EAAE,CAAC,WAAW,EAAE,SAAS,CAAC,CAAU,EAAE,CAAC;QAC1F,IAAI,OAAO,CAAC,UAAU,CAAC,MAAM,KAAK,YAAY;YAAE,UAAU,CAAC,IAAI,CAAC,GAAG,OAAO,CAAC,UAAU,CAAC,OAAO,CAAC,GAAG,CAAC,MAAM,CAAC,EAAE,CAAC,GAAG,IAAI,KAAK,MAAM,EAAE,CAAC,CAAC,CAAA;IACpI,CAAC;IACD,MAAM,CAAC,GAAG,QAAQ,CAAC,UAAU,EAAE,CAAC,GAAG,SAAS,CAAC,UAAU,CAAA;IACvD,MAAM,cAAc,GAAG,CAAC,CAAC,MAAM,KAAK,UAAU,IAAI,CAAC,CAAC,MAAM,KAAK,UAAU,IAAI,CAAC,CAAC,QAAQ,KAAK,CAAC,CAAC,QAAQ,CAAA;IACtG,IAAI,CAAC,CAAC,MAAM,KAAK,UAAU,IAAI,CAAC,CAAC,MAAM,KAAK,UAAU,IAAI,CAAC,CAAC,QAAQ,KAAK,CAAC,CAAC,QAAQ;QAAE,UAAU,CAAC,IAAI,CAAC,8BAA8B,CAAC,CAAA;IACpI,MAAM,MAAM,GAAG,UAAU,CAAC,MAAM,CAAC,CAAC,CAAC,WAAW,CAAC,CAAC,CAAC,OAAO,CAAC,MAAM,CAAC,CAAC,CAAC,YAAY,CAAC,CAAC,CAAC,UAAU,CAAC,MAAM,CAAC,CAAC,CAAC,YAAY,CAAC,CAAC,CAAC,UAAU,CAAA;IAC9H,MAAM,KAAK,GAAG,QAAQ,CAAC,KAAK,CAAA;IAC5B,MAAM,MAAM,GAAG;QACb,OAAO,EAAE,KAAK,CAAC,MAAM,KAAK,QAAQ,CAAC,KAAK,CAAC,MAAM,IAAI,CAAC,KAAK,CAAC,MAAM,CAAC,MAAM,GAAG,KAAK,CAAC,IAAI,CAAC,MAAM,CAAC,GAAG,KAAK,CAAC,MAAM,GAAG,KAAK,IAAI,KAAK,CAAC,iBAAiB;eACzI,KAAK,CAAC,IAAI,CAAC,MAAM,IAAI,KAAK,CAAC,gBAAgB,IAAI,MAAM,CAAC,MAAM,CAAC,UAAU,CAAC,CAAC,KAAK,CAAC,QAAQ,CAAC,EAAE,CAAC,QAAQ,CAAC,MAAM,IAAI,QAAQ,CAAC,IAAI,CAAC;QACjI,IAAI,EAAE,cAAc,IAAI,CAAC,KAAK,CAAC,gBAAgB,KAAK,SAAS,IAAI,cAAc,CAAC,CAAC,CAAC,MAAM,EAAE,CAAC,CAAC,MAAM,EAAE,KAAK,CAAC,gBAAgB,CAAC,CAAC;QAC5H,OAAO,EAAE,KAAK,CAAC,mBAAmB,KAAK,SAAS,IAAI,cAAc,CAAC,QAAQ,CAAC,SAAS,EAAE,SAAS,CAAC,SAAS,EAAE,KAAK,CAAC,mBAAmB,CAAC;KACvI,CAAA;IACD,OAAO,EAAE,MAAM,EAAE,QAAQ,EAAE,MAAM,KAAK,UAAU,IAAI,MAAM,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC,KAAK,CAAC,OAAO,CAAC,EAAE,OAAO,EAAE,CAAC,GAAG,UAAU,EAAE,GAAG,OAAO,EAAE,GAAG,UAAU,CAAC;QAC3I,OAAO,EAAE,QAAQ,CAAC,KAAK,CAAC,MAAM,EAAE,QAAQ,EAAE,SAAS,EAAE,KAAK,EAAE,UAAU,EAAE,MAAM,EAAE,CAAA;AACpF,CAAC"}

package/lib/core/experiment-recipes.d.ts

@@ -0,0 +1,5 @@
+/** Documentation only: experiments run in task-owned shell scripts, never here. */
+export declare const EXPERIMENT_RECIPES: readonly ["reduce", "differential", "interleavings", "metamorphic"];
+export type ExperimentRecipe = typeof EXPERIMENT_RECIPES[number];
+export declare function readExperimentRecipe(recipe: ExperimentRecipe): string;
+//# sourceMappingURL=experiment-recipes.d.ts.map

package/lib/core/experiment-recipes.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"experiment-recipes.d.ts","sourceRoot":"","sources":["../../src/core/experiment-recipes.ts"],"names":[],"mappings":"AAAA,mFAAmF;AACnF,eAAO,MAAM,kBAAkB,qEAAsE,CAAA;AACrG,MAAM,MAAM,gBAAgB,GAAG,OAAO,kBAAkB,CAAC,MAAM,CAAC,CAAA;AAqDhE,wBAAgB,oBAAoB,CAAC,MAAM,EAAE,gBAAgB,GAAG,MAAM,CAYrE"}

package/lib/core/experiment-recipes.js

@@ -0,0 +1,67 @@
+/** Documentation only: experiments run in task-owned shell scripts, never here. */
+export const EXPERIMENT_RECIPES = ['reduce', 'differential', 'interleavings', 'metamorphic'];
+const examples = {
+    reduce: `const probe = xs => xs.filter(x => x === 2).length > 1
+  ? {kind:'fail', signature:'duplicate-two'} : {kind:'pass'};
+const result = await reduceFailure([0,2,8,2,9], 'duplicate-two', probe, budget);
+const {input,...summary}=result;
+console.log(JSON.stringify({...summary,inputLength:input.length,inputPreview:input.slice(0,8)}));`,
+    differential: `// Independent reference: sum retains duplicates; this candidate incorrectly drops them.
+const sum = xs => xs.reduce((a,b) => a+b, 0);
+const probe = differential(sum, xs => sum([...new Set(xs)]));
+const result = await findCounterexample([[0,1], [0,2,8,2,9]], probe, budget);
+// A general difference does not identify the original fault; do not auto-reduce it.
+const {input,verdict,...summary}=result;
+console.log(JSON.stringify({...summary,...(input ? {inputLength:input.length,inputPreview:input.slice(0,8)} : {}),
+  ...(verdict ? {signature:verdict.signature} : {})}));`,
+    interleavings: `// Explicit start/completion boundaries, two requests, fresh state per trial.
+async function observe(schedule, guarded) {
+  let version=0, latest=-1, visible=-1;
+  const gates=[Promise.withResolvers(), Promise.withResolvers()], jobs=[];
+  for (const {lane,index} of schedule) {
+    if (index===0) {
+      const token=++version; latest=lane;
+      jobs[lane]=(async()=>{ await gates[lane].promise;
+        if (!guarded || token===version) visible=lane; })();
+    } else { gates[lane].resolve(); await jobs[lane]; }
+  }
+  return {latest,visible};
+}
+const schedules=interleavings(2,2,6);
+const probe=differential(s=>observe(s,true), s=>observe(s,false));
+const result=await findCounterexample(schedules,probe,budget);
+const {input,verdict,...summary}=result;
+console.log(JSON.stringify({coverage:schedules.coverage,...summary,
+  ...(input ? {inputLength:input.length,inputPreview:input.slice(0,8)} : {}),...(verdict ? {signature:verdict.signature} : {})}));`,
+    metamorphic: `// Contract: a sum is invariant under permutation. The candidate violates it.
+const probe=metamorphic('sum-permutation', xs=>xs[0], xs=>xs.reverse(), (a,b)=>a===b);
+const result=await findCounterexample([[1,2,3]],probe,budget);
+const {input,verdict,...summary}=result;
+console.log(JSON.stringify({...summary,...(input ? {inputLength:input.length,inputPreview:input.slice(0,8)} : {}),
+  ...(verdict ? {signature:verdict.signature} : {})}));`,
+};
+const imports = {
+    reduce: 'reduceFailure', differential: 'differential, findCounterexample',
+    interleavings: 'interleavings, differential, findCounterexample', metamorphic: 'metamorphic, findCounterexample',
+};
+const notes = {
+    reduce: 'Use a task-specific failure predicate/signature. Confirmation runs twice and counts toward budget. one-removal assumes repeatable probes; it is not a global minimum or proof of the same root cause.',
+    differential: 'Use an independent, trusted reference. Each probe runs two observations. checked covers supplied samples only. behavior-difference identifies any mismatch, not one fault; reduce only with a task-specific predicate/signature, otherwise report a difference witness.',
+    interleavings: 'Control real await boundaries with barriers, not sleeps; reset state per schedule. Two ordered lanes, at most 256 steps. coverage describes enumeration, not checked schedules; search stops at the first difference. Each differential probe runs two observations.',
+    metamorphic: 'Derive the transform/relation from a real contract. Each probe runs two observations. checked covers supplied samples only; a relation violation does not identify its root cause.',
+};
+export function readExperimentRecipe(recipe) {
+    if (!Object.hasOwn(examples, recipe))
+        throw new Error('Unknown experiment recipe');
+    const moduleUrl = new URL('../experiments.js', import.meta.url).href;
+    return `Experiment: ${recipe}
+Module: ${moduleUrl}
+Demo, not project evidence. Adapt to authorized, repeatable checks with isolated state and cloneable data. Run in foreground bash with timeoutMs (e.g. 10000); library deadlines/abort are cooperative, not hard interruption. Share one probe budget per batch. Exceptions propagate; unknown/stopped are not pass. Array previews show at most 8 entries; full data stays in result. Save large reports to task files; avoid printing large entries.
+${notes[recipe]}
+\`\`\`js
+import { ExperimentBudget, ${imports[recipe]} } from ${JSON.stringify(moduleUrl)};
+const budget=new ExperimentBudget({maxChecks:100,timeoutMs:2000});
+${examples[recipe]}
+\`\`\``;
+}
+//# sourceMappingURL=experiment-recipes.js.map

package/lib/core/experiment-recipes.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"experiment-recipes.js","sourceRoot":"","sources":["../../src/core/experiment-recipes.ts"],"names":[],"mappings":"AAAA,mFAAmF;AACnF,MAAM,CAAC,MAAM,kBAAkB,GAAG,CAAC,QAAQ,EAAE,cAAc,EAAE,eAAe,EAAE,aAAa,CAAU,CAAA;AAGrG,MAAM,QAAQ,GAA+C;IAC3D,MAAM,EAAE;;;;kGAIwF;IAChG,YAAY,EAAE;;;;;;;wDAOwC;IACtD,aAAa,EAAE;;;;;;;;;;;;;;;;;;mIAkBkH;IACjI,WAAW,EAAE;;;;;wDAKyC;CACvD,CAAA;AACD,MAAM,OAAO,GAA+C;IAC1D,MAAM,EAAE,eAAe,EAAE,YAAY,EAAE,kCAAkC;IACzE,aAAa,EAAE,iDAAiD,EAAE,WAAW,EAAE,iCAAiC;CACjH,CAAA;AACD,MAAM,KAAK,GAA+C;IACxD,MAAM,EAAE,uMAAuM;IAC/M,YAAY,EAAE,yQAAyQ;IACvR,aAAa,EAAE,sQAAsQ;IACrR,WAAW,EAAE,oLAAoL;CAClM,CAAA;AAED,MAAM,UAAU,oBAAoB,CAAC,MAAwB;IAC3D,IAAI,CAAC,MAAM,CAAC,MAAM,CAAC,QAAQ,EAAE,MAAM,CAAC;QAAE,MAAM,IAAI,KAAK,CAAC,2BAA2B,CAAC,CAAA;IAClF,MAAM,SAAS,GAAG,IAAI,GAAG,CAAC,mBAAmB,EAAE,MAAM,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,IAAI,CAAA;IACpE,OAAO,eAAe,MAAM;UACpB,SAAS;;EAEjB,KAAK,CAAC,MAAM,CAAC;;6BAEc,OAAO,CAAC,MAAM,CAAC,WAAW,IAAI,CAAC,SAAS,CAAC,SAAS,CAAC;;EAE9E,QAAQ,CAAC,MAAM,CAAC;OACX,CAAA;AACP,CAAC"}

package/lib/core/guidance.d.ts

@@ -0,0 +1,10 @@
+export type GuidanceMode = 'standard' | 'guided';
+export interface GuidedRoute {
+    readonly provider: string;
+    readonly model: string;
+}
+/** Exact deployment policy, not a guess about capability from a model name. */
+export declare function createGuidanceSelector(routes?: readonly GuidedRoute[]): (provider: unknown, model: unknown) => GuidanceMode;
+/** Only selected routes pay for concrete reminders; standards remain identical. */
+export declare const GUIDED_INSTRUCTIONS = "Inspect definitions and working callers for uncertain APIs; do not add aliases to satisfy imagined tests. A cause remains unconfirmed until an observation distinguishes it. Run the relevant check and inspect the result: a started command, exit code alone, zero tests or a teammate's claim does not establish acceptance. Before calling a failure pre-existing or environmental, compare the same check on the baseline; otherwise report it as unresolved. Never assume hidden tests will fix a failure. When stuck, change the hypothesis or reduce the reproduction, not the agent count. Use only necessary checks; no mandatory checklist or extra narration.";
+//# sourceMappingURL=guidance.d.ts.map

package/lib/core/guidance.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"guidance.d.ts","sourceRoot":"","sources":["../../src/core/guidance.ts"],"names":[],"mappings":"AAEA,MAAM,MAAM,YAAY,GAAG,UAAU,GAAG,QAAQ,CAAA;AAChD,MAAM,WAAW,WAAW;IAAG,QAAQ,CAAC,QAAQ,EAAE,MAAM,CAAC;IAAC,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAA;CAAE;AAOlF,+EAA+E;AAC/E,wBAAgB,sBAAsB,CAAC,MAAM,GAAE,SAAS,WAAW,EAAO,GAAG,CAAC,QAAQ,EAAE,OAAO,EAAE,KAAK,EAAE,OAAO,KAAK,YAAY,CAK/H;AAED,mFAAmF;AACnF,eAAO,MAAM,mBAAmB,8oBAA8oB,CAAA"}

package/lib/core/guidance.js

@@ -0,0 +1,16 @@
+import { z } from 'zod';
+const guidedRoutesSchema = z.array(z.object({
+    provider: z.string().trim().min(1).max(256),
+    model: z.string().trim().min(1).max(256),
+}).strict()).max(128);
+/** Exact deployment policy, not a guess about capability from a model name. */
+export function createGuidanceSelector(routes = []) {
+    const configured = guidedRoutesSchema.parse(routes);
+    const keys = new Set(configured.map(route => JSON.stringify([route.provider, route.model])));
+    if (keys.size !== configured.length)
+        throw new Error('Duplicate guided model route');
+    return (provider, model) => typeof provider === 'string' && typeof model === 'string' && keys.has(JSON.stringify([provider, model])) ? 'guided' : 'standard';
+}
+/** Only selected routes pay for concrete reminders; standards remain identical. */
+export const GUIDED_INSTRUCTIONS = `Inspect definitions and working callers for uncertain APIs; do not add aliases to satisfy imagined tests. A cause remains unconfirmed until an observation distinguishes it. Run the relevant check and inspect the result: a started command, exit code alone, zero tests or a teammate's claim does not establish acceptance. Before calling a failure pre-existing or environmental, compare the same check on the baseline; otherwise report it as unresolved. Never assume hidden tests will fix a failure. When stuck, change the hypothesis or reduce the reproduction, not the agent count. Use only necessary checks; no mandatory checklist or extra narration.`;
+//# sourceMappingURL=guidance.js.map

package/lib/core/guidance.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"guidance.js","sourceRoot":"","sources":["../../src/core/guidance.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAA;AAKvB,MAAM,kBAAkB,GAAG,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,MAAM,CAAC;IAC1C,QAAQ,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,IAAI,EAAE,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,GAAG,CAAC,GAAG,CAAC;IAC3C,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,IAAI,EAAE,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,GAAG,CAAC,GAAG,CAAC;CACzC,CAAC,CAAC,MAAM,EAAE,CAAC,CAAC,GAAG,CAAC,GAAG,CAAC,CAAA;AAErB,+EAA+E;AAC/E,MAAM,UAAU,sBAAsB,CAAC,SAAiC,EAAE;IACxE,MAAM,UAAU,GAAG,kBAAkB,CAAC,KAAK,CAAC,MAAM,CAAC,CAAA;IACnD,MAAM,IAAI,GAAG,IAAI,GAAG,CAAC,UAAU,CAAC,GAAG,CAAC,KAAK,CAAC,EAAE,CAAC,IAAI,CAAC,SAAS,CAAC,CAAC,KAAK,CAAC,QAAQ,EAAE,KAAK,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,CAAA;IAC5F,IAAI,IAAI,CAAC,IAAI,KAAK,UAAU,CAAC,MAAM;QAAE,MAAM,IAAI,KAAK,CAAC,8BAA8B,CAAC,CAAA;IACpF,OAAO,CAAC,QAAQ,EAAE,KAAK,EAAE,EAAE,CAAC,OAAO,QAAQ,KAAK,QAAQ,IAAI,OAAO,KAAK,KAAK,QAAQ,IAAI,IAAI,CAAC,GAAG,CAAC,IAAI,CAAC,SAAS,CAAC,CAAC,QAAQ,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,QAAQ,CAAC,CAAC,CAAC,UAAU,CAAA;AAC9J,CAAC;AAED,mFAAmF;AACnF,MAAM,CAAC,MAAM,mBAAmB,GAAG,2oBAA2oB,CAAA"}

package/lib/core/session-scan.d.ts

@@ -0,0 +1,8 @@
+import type { SessionEvent } from '@deepseek-ai/dsh-session';
+/** Bounded backward lookup; yield between chunks so cancellation stays responsive. */
+export declare function scanSessionPage(end: number, before: number, eventAt: (seq: number) => SessionEvent | undefined, visit: (event: SessionEvent) => boolean, signal: AbortSignal): Promise<{
+    nextBeforeSeq: number;
+    done: boolean;
+    scanned: number;
+}>;
+//# sourceMappingURL=session-scan.d.ts.map

package/lib/core/session-scan.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"session-scan.d.ts","sourceRoot":"","sources":["../../src/core/session-scan.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,0BAA0B,CAAA;AAE5D,sFAAsF;AACtF,wBAAsB,eAAe,CAAC,GAAG,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,EAAE,OAAO,EAAE,CAAC,GAAG,EAAE,MAAM,KAAK,YAAY,GAAG,SAAS,EACnH,KAAK,EAAE,CAAC,KAAK,EAAE,YAAY,KAAK,OAAO,EAAE,MAAM,EAAE,WAAW,GAAG,OAAO,CAAC;IAAE,aAAa,EAAE,MAAM,CAAC;IAAC,IAAI,EAAE,OAAO,CAAC;IAAC,OAAO,EAAE,MAAM,CAAA;CAAE,CAAC,CAelI"}