@mcptoolshop/research-os 0.4.0 → 0.5.0

package/dist/calibration/aggregate-receipt-schema.d.ts

@@ -0,0 +1,509 @@
+import { z } from 'zod';
+
+declare const AggregateMetricSchema: z.ZodObject<{
+    median: z.ZodNumber;
+    min: z.ZodNumber;
+    max: z.ZodNumber;
+    values: z.ZodArray<z.ZodNumber, "many">;
+}, "strip", z.ZodTypeAny, {
+    median: number;
+    min: number;
+    max: number;
+    values: number[];
+}, {
+    median: number;
+    min: number;
+    max: number;
+    values: number[];
+}>;
+declare const PerCategoryAggregateEntrySchema: z.ZodObject<{
+    median_ratio: z.ZodNumber;
+    min_ratio: z.ZodNumber;
+    max_ratio: z.ZodNumber;
+    total: z.ZodNumber;
+    per_run_ratios: z.ZodArray<z.ZodNumber, "many">;
+}, "strip", z.ZodTypeAny, {
+    median_ratio: number;
+    min_ratio: number;
+    max_ratio: number;
+    total: number;
+    per_run_ratios: number[];
+}, {
+    median_ratio: number;
+    min_ratio: number;
+    max_ratio: number;
+    total: number;
+    per_run_ratios: number[];
+}>;
+declare const PerCategoryAggregateSchema: z.ZodRecord<z.ZodString, z.ZodObject<{
+    median_ratio: z.ZodNumber;
+    min_ratio: z.ZodNumber;
+    max_ratio: z.ZodNumber;
+    total: z.ZodNumber;
+    per_run_ratios: z.ZodArray<z.ZodNumber, "many">;
+}, "strip", z.ZodTypeAny, {
+    median_ratio: number;
+    min_ratio: number;
+    max_ratio: number;
+    total: number;
+    per_run_ratios: number[];
+}, {
+    median_ratio: number;
+    min_ratio: number;
+    max_ratio: number;
+    total: number;
+    per_run_ratios: number[];
+}>>;
+declare const AggregatePassFailSchema: z.ZodObject<{
+    fp_ceiling: z.ZodEnum<["PASS", "FAIL"]>;
+    any_flag_recall_floor: z.ZodEnum<["PASS", "FAIL"]>;
+    per_category_any_flag_floor: z.ZodEnum<["PASS", "FAIL"]>;
+    strict_recall_floor: z.ZodEnum<["PASS", "FAIL"]>;
+    decision_vocab_completeness: z.ZodEnum<["PASS", "FAIL"]>;
+    latency_soft: z.ZodEnum<["PASS", "WARN"]>;
+    latency_hard: z.ZodEnum<["PASS", "FAIL"]>;
+    empty_or_malformed: z.ZodEnum<["PASS", "FAIL"]>;
+    overall: z.ZodEnum<["PASS", "FAIL"]>;
+}, "strip", z.ZodTypeAny, {
+    fp_ceiling: "PASS" | "FAIL";
+    any_flag_recall_floor: "PASS" | "FAIL";
+    per_category_any_flag_floor: "PASS" | "FAIL";
+    strict_recall_floor: "PASS" | "FAIL";
+    decision_vocab_completeness: "PASS" | "FAIL";
+    latency_soft: "PASS" | "WARN";
+    latency_hard: "PASS" | "FAIL";
+    empty_or_malformed: "PASS" | "FAIL";
+    overall: "PASS" | "FAIL";
+}, {
+    fp_ceiling: "PASS" | "FAIL";
+    any_flag_recall_floor: "PASS" | "FAIL";
+    per_category_any_flag_floor: "PASS" | "FAIL";
+    strict_recall_floor: "PASS" | "FAIL";
+    decision_vocab_completeness: "PASS" | "FAIL";
+    latency_soft: "PASS" | "WARN";
+    latency_hard: "PASS" | "FAIL";
+    empty_or_malformed: "PASS" | "FAIL";
+    overall: "PASS" | "FAIL";
+}>;
+declare const AggregateDecisionVocabBarSchema: z.ZodObject<{
+    architecture: z.ZodEnum<["single-pass", "two-pass"]>;
+    required: z.ZodNumber;
+    median_produced: z.ZodNumber;
+    passed: z.ZodBoolean;
+}, "strip", z.ZodTypeAny, {
+    required: number;
+    architecture: "single-pass" | "two-pass";
+    median_produced: number;
+    passed: boolean;
+}, {
+    required: number;
+    architecture: "single-pass" | "two-pass";
+    median_produced: number;
+    passed: boolean;
+}>;
+declare const AggregateCalibrationReceiptSchema: z.ZodObject<{
+    schema_version: z.ZodLiteral<1>;
+    receipt_kind: z.ZodLiteral<"aggregate">;
+    profile_name: z.ZodString;
+    status: z.ZodEnum<["trusted_baseline", "conditional_pass", "failed", "comparison_only"]>;
+    model: z.ZodString;
+    architecture: z.ZodEnum<["single-pass", "two-pass"]>;
+    fixture: z.ZodString;
+    fixture_total_claims: z.ZodNumber;
+    fixture_good_claims: z.ZodNumber;
+    fixture_bad_claims: z.ZodNumber;
+    runs_count: z.ZodNumber;
+    run_files: z.ZodArray<z.ZodString, "many">;
+    aggregated_at: z.ZodString;
+    research_os_version: z.ZodString;
+    good_fp_count: z.ZodObject<{
+        median: z.ZodNumber;
+        min: z.ZodNumber;
+        max: z.ZodNumber;
+        values: z.ZodArray<z.ZodNumber, "many">;
+    }, "strip", z.ZodTypeAny, {
+        median: number;
+        min: number;
+        max: number;
+        values: number[];
+    }, {
+        median: number;
+        min: number;
+        max: number;
+        values: number[];
+    }>;
+    any_flag_recall_ratio: z.ZodObject<{
+        median: z.ZodNumber;
+        min: z.ZodNumber;
+        max: z.ZodNumber;
+        values: z.ZodArray<z.ZodNumber, "many">;
+    }, "strip", z.ZodTypeAny, {
+        median: number;
+        min: number;
+        max: number;
+        values: number[];
+    }, {
+        median: number;
+        min: number;
+        max: number;
+        values: number[];
+    }>;
+    strict_recall_ratio: z.ZodObject<{
+        median: z.ZodNumber;
+        min: z.ZodNumber;
+        max: z.ZodNumber;
+        values: z.ZodArray<z.ZodNumber, "many">;
+    }, "strip", z.ZodTypeAny, {
+        median: number;
+        min: number;
+        max: number;
+        values: number[];
+    }, {
+        median: number;
+        min: number;
+        max: number;
+        values: number[];
+    }>;
+    decisions_produced_count: z.ZodObject<{
+        median: z.ZodNumber;
+        min: z.ZodNumber;
+        max: z.ZodNumber;
+        values: z.ZodArray<z.ZodNumber, "many">;
+    }, "strip", z.ZodTypeAny, {
+        median: number;
+        min: number;
+        max: number;
+        values: number[];
+    }, {
+        median: number;
+        min: number;
+        max: number;
+        values: number[];
+    }>;
+    runtime_ms: z.ZodObject<{
+        median: z.ZodNumber;
+        min: z.ZodNumber;
+        max: z.ZodNumber;
+        values: z.ZodArray<z.ZodNumber, "many">;
+    }, "strip", z.ZodTypeAny, {
+        median: number;
+        min: number;
+        max: number;
+        values: number[];
+    }, {
+        median: number;
+        min: number;
+        max: number;
+        values: number[];
+    }>;
+    empty_or_malformed_responses: z.ZodObject<{
+        median: z.ZodNumber;
+        min: z.ZodNumber;
+        max: z.ZodNumber;
+        values: z.ZodArray<z.ZodNumber, "many">;
+    }, "strip", z.ZodTypeAny, {
+        median: number;
+        min: number;
+        max: number;
+        values: number[];
+    }, {
+        median: number;
+        min: number;
+        max: number;
+        values: number[];
+    }>;
+    per_category_any_flag: z.ZodRecord<z.ZodString, z.ZodObject<{
+        median_ratio: z.ZodNumber;
+        min_ratio: z.ZodNumber;
+        max_ratio: z.ZodNumber;
+        total: z.ZodNumber;
+        per_run_ratios: z.ZodArray<z.ZodNumber, "many">;
+    }, "strip", z.ZodTypeAny, {
+        median_ratio: number;
+        min_ratio: number;
+        max_ratio: number;
+        total: number;
+        per_run_ratios: number[];
+    }, {
+        median_ratio: number;
+        min_ratio: number;
+        max_ratio: number;
+        total: number;
+        per_run_ratios: number[];
+    }>>;
+    per_category_strict: z.ZodRecord<z.ZodString, z.ZodObject<{
+        median_ratio: z.ZodNumber;
+        min_ratio: z.ZodNumber;
+        max_ratio: z.ZodNumber;
+        total: z.ZodNumber;
+        per_run_ratios: z.ZodArray<z.ZodNumber, "many">;
+    }, "strip", z.ZodTypeAny, {
+        median_ratio: number;
+        min_ratio: number;
+        max_ratio: number;
+        total: number;
+        per_run_ratios: number[];
+    }, {
+        median_ratio: number;
+        min_ratio: number;
+        max_ratio: number;
+        total: number;
+        per_run_ratios: number[];
+    }>>;
+    decision_vocabulary: z.ZodRecord<z.ZodString, z.ZodObject<{
+        median: z.ZodNumber;
+        min: z.ZodNumber;
+        max: z.ZodNumber;
+        values: z.ZodArray<z.ZodNumber, "many">;
+    }, "strip", z.ZodTypeAny, {
+        median: number;
+        min: number;
+        max: number;
+        values: number[];
+    }, {
+        median: number;
+        min: number;
+        max: number;
+        values: number[];
+    }>>;
+    decision_vocab_bar: z.ZodObject<{
+        architecture: z.ZodEnum<["single-pass", "two-pass"]>;
+        required: z.ZodNumber;
+        median_produced: z.ZodNumber;
+        passed: z.ZodBoolean;
+    }, "strip", z.ZodTypeAny, {
+        required: number;
+        architecture: "single-pass" | "two-pass";
+        median_produced: number;
+        passed: boolean;
+    }, {
+        required: number;
+        architecture: "single-pass" | "two-pass";
+        median_produced: number;
+        passed: boolean;
+    }>;
+    unreachable_decisions: z.ZodArray<z.ZodString, "many">;
+    pass_fail: z.ZodObject<{
+        fp_ceiling: z.ZodEnum<["PASS", "FAIL"]>;
+        any_flag_recall_floor: z.ZodEnum<["PASS", "FAIL"]>;
+        per_category_any_flag_floor: z.ZodEnum<["PASS", "FAIL"]>;
+        strict_recall_floor: z.ZodEnum<["PASS", "FAIL"]>;
+        decision_vocab_completeness: z.ZodEnum<["PASS", "FAIL"]>;
+        latency_soft: z.ZodEnum<["PASS", "WARN"]>;
+        latency_hard: z.ZodEnum<["PASS", "FAIL"]>;
+        empty_or_malformed: z.ZodEnum<["PASS", "FAIL"]>;
+        overall: z.ZodEnum<["PASS", "FAIL"]>;
+    }, "strip", z.ZodTypeAny, {
+        fp_ceiling: "PASS" | "FAIL";
+        any_flag_recall_floor: "PASS" | "FAIL";
+        per_category_any_flag_floor: "PASS" | "FAIL";
+        strict_recall_floor: "PASS" | "FAIL";
+        decision_vocab_completeness: "PASS" | "FAIL";
+        latency_soft: "PASS" | "WARN";
+        latency_hard: "PASS" | "FAIL";
+        empty_or_malformed: "PASS" | "FAIL";
+        overall: "PASS" | "FAIL";
+    }, {
+        fp_ceiling: "PASS" | "FAIL";
+        any_flag_recall_floor: "PASS" | "FAIL";
+        per_category_any_flag_floor: "PASS" | "FAIL";
+        strict_recall_floor: "PASS" | "FAIL";
+        decision_vocab_completeness: "PASS" | "FAIL";
+        latency_soft: "PASS" | "WARN";
+        latency_hard: "PASS" | "FAIL";
+        empty_or_malformed: "PASS" | "FAIL";
+        overall: "PASS" | "FAIL";
+    }>;
+    recurring_bar_failures: z.ZodArray<z.ZodString, "many">;
+    notes: z.ZodArray<z.ZodString, "many">;
+}, "strip", z.ZodTypeAny, {
+    status: "trusted_baseline" | "conditional_pass" | "failed" | "comparison_only";
+    architecture: "single-pass" | "two-pass";
+    schema_version: 1;
+    receipt_kind: "aggregate";
+    profile_name: string;
+    model: string;
+    fixture: string;
+    fixture_total_claims: number;
+    fixture_good_claims: number;
+    fixture_bad_claims: number;
+    runs_count: number;
+    run_files: string[];
+    aggregated_at: string;
+    research_os_version: string;
+    good_fp_count: {
+        median: number;
+        min: number;
+        max: number;
+        values: number[];
+    };
+    any_flag_recall_ratio: {
+        median: number;
+        min: number;
+        max: number;
+        values: number[];
+    };
+    strict_recall_ratio: {
+        median: number;
+        min: number;
+        max: number;
+        values: number[];
+    };
+    decisions_produced_count: {
+        median: number;
+        min: number;
+        max: number;
+        values: number[];
+    };
+    runtime_ms: {
+        median: number;
+        min: number;
+        max: number;
+        values: number[];
+    };
+    empty_or_malformed_responses: {
+        median: number;
+        min: number;
+        max: number;
+        values: number[];
+    };
+    per_category_any_flag: Record<string, {
+        median_ratio: number;
+        min_ratio: number;
+        max_ratio: number;
+        total: number;
+        per_run_ratios: number[];
+    }>;
+    per_category_strict: Record<string, {
+        median_ratio: number;
+        min_ratio: number;
+        max_ratio: number;
+        total: number;
+        per_run_ratios: number[];
+    }>;
+    decision_vocabulary: Record<string, {
+        median: number;
+        min: number;
+        max: number;
+        values: number[];
+    }>;
+    decision_vocab_bar: {
+        required: number;
+        architecture: "single-pass" | "two-pass";
+        median_produced: number;
+        passed: boolean;
+    };
+    unreachable_decisions: string[];
+    pass_fail: {
+        fp_ceiling: "PASS" | "FAIL";
+        any_flag_recall_floor: "PASS" | "FAIL";
+        per_category_any_flag_floor: "PASS" | "FAIL";
+        strict_recall_floor: "PASS" | "FAIL";
+        decision_vocab_completeness: "PASS" | "FAIL";
+        latency_soft: "PASS" | "WARN";
+        latency_hard: "PASS" | "FAIL";
+        empty_or_malformed: "PASS" | "FAIL";
+        overall: "PASS" | "FAIL";
+    };
+    recurring_bar_failures: string[];
+    notes: string[];
+}, {
+    status: "trusted_baseline" | "conditional_pass" | "failed" | "comparison_only";
+    architecture: "single-pass" | "two-pass";
+    schema_version: 1;
+    receipt_kind: "aggregate";
+    profile_name: string;
+    model: string;
+    fixture: string;
+    fixture_total_claims: number;
+    fixture_good_claims: number;
+    fixture_bad_claims: number;
+    runs_count: number;
+    run_files: string[];
+    aggregated_at: string;
+    research_os_version: string;
+    good_fp_count: {
+        median: number;
+        min: number;
+        max: number;
+        values: number[];
+    };
+    any_flag_recall_ratio: {
+        median: number;
+        min: number;
+        max: number;
+        values: number[];
+    };
+    strict_recall_ratio: {
+        median: number;
+        min: number;
+        max: number;
+        values: number[];
+    };
+    decisions_produced_count: {
+        median: number;
+        min: number;
+        max: number;
+        values: number[];
+    };
+    runtime_ms: {
+        median: number;
+        min: number;
+        max: number;
+        values: number[];
+    };
+    empty_or_malformed_responses: {
+        median: number;
+        min: number;
+        max: number;
+        values: number[];
+    };
+    per_category_any_flag: Record<string, {
+        median_ratio: number;
+        min_ratio: number;
+        max_ratio: number;
+        total: number;
+        per_run_ratios: number[];
+    }>;
+    per_category_strict: Record<string, {
+        median_ratio: number;
+        min_ratio: number;
+        max_ratio: number;
+        total: number;
+        per_run_ratios: number[];
+    }>;
+    decision_vocabulary: Record<string, {
+        median: number;
+        min: number;
+        max: number;
+        values: number[];
+    }>;
+    decision_vocab_bar: {
+        required: number;
+        architecture: "single-pass" | "two-pass";
+        median_produced: number;
+        passed: boolean;
+    };
+    unreachable_decisions: string[];
+    pass_fail: {
+        fp_ceiling: "PASS" | "FAIL";
+        any_flag_recall_floor: "PASS" | "FAIL";
+        per_category_any_flag_floor: "PASS" | "FAIL";
+        strict_recall_floor: "PASS" | "FAIL";
+        decision_vocab_completeness: "PASS" | "FAIL";
+        latency_soft: "PASS" | "WARN";
+        latency_hard: "PASS" | "FAIL";
+        empty_or_malformed: "PASS" | "FAIL";
+        overall: "PASS" | "FAIL";
+    };
+    recurring_bar_failures: string[];
+    notes: string[];
+}>;
+type AggregateMetric = z.infer<typeof AggregateMetricSchema>;
+type PerCategoryAggregateEntry = z.infer<typeof PerCategoryAggregateEntrySchema>;
+type PerCategoryAggregate = z.infer<typeof PerCategoryAggregateSchema>;
+type AggregatePassFail = z.infer<typeof AggregatePassFailSchema>;
+type AggregateDecisionVocabBar = z.infer<typeof AggregateDecisionVocabBarSchema>;
+type AggregateCalibrationReceipt = z.infer<typeof AggregateCalibrationReceiptSchema>;
+
+export { type AggregateCalibrationReceipt, AggregateCalibrationReceiptSchema, type AggregateDecisionVocabBar, AggregateDecisionVocabBarSchema, type AggregateMetric, AggregateMetricSchema, type AggregatePassFail, AggregatePassFailSchema, type PerCategoryAggregate, type PerCategoryAggregateEntry, PerCategoryAggregateEntrySchema, PerCategoryAggregateSchema };

package/dist/calibration/aggregate-receipt-schema.js

@@ -0,0 +1,143 @@
+// src/calibration/aggregate-receipt-schema.ts
+import { z as z2 } from "zod";
+
+// src/calibration/receipt-schema.ts
+import { z } from "zod";
+var StatusLabelSchema = z.enum([
+  "trusted_baseline",
+  "conditional_pass",
+  "failed",
+  "comparison_only"
+]);
+var ArchitectureSchema = z.enum(["single-pass", "two-pass"]);
+var RecallSchema = z.object({
+  matched: z.number().int().nonnegative(),
+  total: z.number().int().nonnegative(),
+  ratio: z.number().min(0).max(1)
+});
+var PerCategoryRecallSchema = z.record(z.string(), RecallSchema);
+var PassFailSchema = z.object({
+  fp_ceiling: z.enum(["PASS", "FAIL"]),
+  any_flag_recall_floor: z.enum(["PASS", "FAIL"]),
+  per_category_any_flag_floor: z.enum(["PASS", "FAIL"]),
+  strict_recall_floor: z.enum(["PASS", "FAIL"]),
+  decision_vocab_completeness: z.enum(["PASS", "FAIL"]),
+  latency_soft: z.enum(["PASS", "WARN"]),
+  latency_hard: z.enum(["PASS", "FAIL"]),
+  empty_or_malformed: z.enum(["PASS", "FAIL"]),
+  overall: z.enum(["PASS", "FAIL"])
+});
+var DecisionVocabBarSchema = z.object({
+  architecture: ArchitectureSchema,
+  required: z.number().int().positive(),
+  produced: z.number().int().nonnegative(),
+  passed: z.boolean()
+});
+var CalibrationReceiptSchema = z.object({
+  schema_version: z.literal(1),
+  profile_name: z.string(),
+  status: StatusLabelSchema,
+  model: z.string(),
+  architecture: ArchitectureSchema,
+  fixture: z.string(),
+  fixture_total_claims: z.number().int().positive(),
+  fixture_good_claims: z.number().int().nonnegative(),
+  fixture_bad_claims: z.number().int().nonnegative(),
+  calibrated_at: z.string(),
+  research_os_version: z.string(),
+  runtime_ms: z.number().int().nonnegative(),
+  good_fp_count: z.number().int().nonnegative(),
+  any_flag_recall: RecallSchema,
+  strict_recall: RecallSchema,
+  per_category_any_flag: PerCategoryRecallSchema,
+  per_category_strict: PerCategoryRecallSchema,
+  decision_vocabulary: z.record(z.string(), z.number().int().nonnegative()),
+  decisions_produced_count: z.number().int().nonnegative(),
+  decision_vocab_bar: DecisionVocabBarSchema,
+  unreachable_decisions: z.array(z.string()),
+  empty_or_malformed_responses: z.number().int().nonnegative(),
+  pass_fail: PassFailSchema,
+  notes: z.array(z.string())
+});
+
+// src/calibration/aggregate-receipt-schema.ts
+var AggregateMetricSchema = z2.object({
+  median: z2.number(),
+  min: z2.number(),
+  max: z2.number(),
+  values: z2.array(z2.number())
+  // per-run values in run order (run-001, run-002, ...)
+});
+var PerCategoryAggregateEntrySchema = z2.object({
+  median_ratio: z2.number().min(0).max(1),
+  min_ratio: z2.number().min(0).max(1),
+  max_ratio: z2.number().min(0).max(1),
+  total: z2.number().int().nonnegative(),
+  // seed count — same across all runs
+  per_run_ratios: z2.array(z2.number())
+});
+var PerCategoryAggregateSchema = z2.record(z2.string(), PerCategoryAggregateEntrySchema);
+var AggregatePassFailSchema = z2.object({
+  fp_ceiling: z2.enum(["PASS", "FAIL"]),
+  any_flag_recall_floor: z2.enum(["PASS", "FAIL"]),
+  per_category_any_flag_floor: z2.enum(["PASS", "FAIL"]),
+  strict_recall_floor: z2.enum(["PASS", "FAIL"]),
+  decision_vocab_completeness: z2.enum(["PASS", "FAIL"]),
+  latency_soft: z2.enum(["PASS", "WARN"]),
+  latency_hard: z2.enum(["PASS", "FAIL"]),
+  empty_or_malformed: z2.enum(["PASS", "FAIL"]),
+  overall: z2.enum(["PASS", "FAIL"])
+});
+var AggregateDecisionVocabBarSchema = z2.object({
+  architecture: ArchitectureSchema,
+  required: z2.number().int().positive(),
+  median_produced: z2.number(),
+  // float — median of per-run decisions_produced_count
+  passed: z2.boolean()
+});
+var AggregateCalibrationReceiptSchema = z2.object({
+  schema_version: z2.literal(1),
+  receipt_kind: z2.literal("aggregate"),
+  // discriminates from single-run receipt
+  profile_name: z2.string(),
+  status: StatusLabelSchema,
+  model: z2.string(),
+  architecture: ArchitectureSchema,
+  fixture: z2.string(),
+  fixture_total_claims: z2.number().int().positive(),
+  fixture_good_claims: z2.number().int().nonnegative(),
+  fixture_bad_claims: z2.number().int().nonnegative(),
+  runs_count: z2.number().int().min(2),
+  run_files: z2.array(z2.string()),
+  // relative paths: runs/run-001.json, etc.
+  aggregated_at: z2.string(),
+  // ISO 8601
+  research_os_version: z2.string(),
+  // Aggregate metrics — median + min + max + per-run values in run order
+  good_fp_count: AggregateMetricSchema,
+  any_flag_recall_ratio: AggregateMetricSchema,
+  strict_recall_ratio: AggregateMetricSchema,
+  decisions_produced_count: AggregateMetricSchema,
+  runtime_ms: AggregateMetricSchema,
+  empty_or_malformed_responses: AggregateMetricSchema,
+  per_category_any_flag: PerCategoryAggregateSchema,
+  per_category_strict: PerCategoryAggregateSchema,
+  // Decision vocabulary — union of all decisions seen across runs, median count each
+  decision_vocabulary: z2.record(z2.string(), AggregateMetricSchema),
+  decision_vocab_bar: AggregateDecisionVocabBarSchema,
+  unreachable_decisions: z2.array(z2.string()),
+  pass_fail: AggregatePassFailSchema,
+  // Bars that FAILed in >= ceil(runs_count/2) individual runs.
+  // Non-empty list demotes trusted_baseline to conditional_pass.
+  recurring_bar_failures: z2.array(z2.string()),
+  notes: z2.array(z2.string())
+});
+export {
+  AggregateCalibrationReceiptSchema,
+  AggregateDecisionVocabBarSchema,
+  AggregateMetricSchema,
+  AggregatePassFailSchema,
+  PerCategoryAggregateEntrySchema,
+  PerCategoryAggregateSchema
+};
+//# sourceMappingURL=aggregate-receipt-schema.js.map

package/dist/calibration/aggregate-receipt-schema.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../src/calibration/aggregate-receipt-schema.ts","../../src/calibration/receipt-schema.ts"],"sourcesContent":["import { z } from 'zod';\nimport { StatusLabelSchema, ArchitectureSchema } from './receipt-schema.js';\n\nexport const AggregateMetricSchema = z.object({\n  median: z.number(),\n  min: z.number(),\n  max: z.number(),\n  values: z.array(z.number()), // per-run values in run order (run-001, run-002, ...)\n});\n\nexport const PerCategoryAggregateEntrySchema = z.object({\n  median_ratio: z.number().min(0).max(1),\n  min_ratio: z.number().min(0).max(1),\n  max_ratio: z.number().min(0).max(1),\n  total: z.number().int().nonnegative(), // seed count — same across all runs\n  per_run_ratios: z.array(z.number()),\n});\n\nexport const PerCategoryAggregateSchema = z.record(z.string(), PerCategoryAggregateEntrySchema);\n\nexport const AggregatePassFailSchema = z.object({\n  fp_ceiling: z.enum(['PASS', 'FAIL']),\n  any_flag_recall_floor: z.enum(['PASS', 'FAIL']),\n  per_category_any_flag_floor: z.enum(['PASS', 'FAIL']),\n  strict_recall_floor: z.enum(['PASS', 'FAIL']),\n  decision_vocab_completeness: z.enum(['PASS', 'FAIL']),\n  latency_soft: z.enum(['PASS', 'WARN']),\n  latency_hard: z.enum(['PASS', 'FAIL']),\n  empty_or_malformed: z.enum(['PASS', 'FAIL']),\n  overall: z.enum(['PASS', 'FAIL']),\n});\n\nexport const AggregateDecisionVocabBarSchema = z.object({\n  architecture: ArchitectureSchema,\n  required: z.number().int().positive(),\n  median_produced: z.number(), // float — median of per-run decisions_produced_count\n  passed: z.boolean(),\n});\n\nexport const AggregateCalibrationReceiptSchema = z.object({\n  schema_version: z.literal(1),\n  receipt_kind: z.literal('aggregate'), // discriminates from single-run receipt\n  profile_name: z.string(),\n  status: StatusLabelSchema,\n  model: z.string(),\n  architecture: ArchitectureSchema,\n  fixture: z.string(),\n  fixture_total_claims: z.number().int().positive(),\n  fixture_good_claims: z.number().int().nonnegative(),\n  fixture_bad_claims: z.number().int().nonnegative(),\n  runs_count: z.number().int().min(2),\n  run_files: z.array(z.string()), // relative paths: runs/run-001.json, etc.\n  aggregated_at: z.string(), // ISO 8601\n  research_os_version: z.string(),\n\n  // Aggregate metrics — median + min + max + per-run values in run order\n  good_fp_count: AggregateMetricSchema,\n  any_flag_recall_ratio: AggregateMetricSchema,\n  strict_recall_ratio: AggregateMetricSchema,\n  decisions_produced_count: AggregateMetricSchema,\n  runtime_ms: AggregateMetricSchema,\n  empty_or_malformed_responses: AggregateMetricSchema,\n\n  per_category_any_flag: PerCategoryAggregateSchema,\n  per_category_strict: PerCategoryAggregateSchema,\n\n  // Decision vocabulary — union of all decisions seen across runs, median count each\n  decision_vocabulary: z.record(z.string(), AggregateMetricSchema),\n  decision_vocab_bar: AggregateDecisionVocabBarSchema,\n  unreachable_decisions: z.array(z.string()),\n\n  pass_fail: AggregatePassFailSchema,\n  // Bars that FAILed in >= ceil(runs_count/2) individual runs.\n  // Non-empty list demotes trusted_baseline to conditional_pass.\n  recurring_bar_failures: z.array(z.string()),\n\n  notes: z.array(z.string()),\n});\n\nexport type AggregateMetric = z.infer<typeof AggregateMetricSchema>;\nexport type PerCategoryAggregateEntry = z.infer<typeof PerCategoryAggregateEntrySchema>;\nexport type PerCategoryAggregate = z.infer<typeof PerCategoryAggregateSchema>;\nexport type AggregatePassFail = z.infer<typeof AggregatePassFailSchema>;\nexport type AggregateDecisionVocabBar = z.infer<typeof AggregateDecisionVocabBarSchema>;\nexport type AggregateCalibrationReceipt = z.infer<typeof AggregateCalibrationReceiptSchema>;\n","import { z } from 'zod';\n\nexport const StatusLabelSchema = z.enum([\n  'trusted_baseline',\n  'conditional_pass',\n  'failed',\n  'comparison_only',\n]);\n\nexport const ArchitectureSchema = z.enum(['single-pass', 'two-pass']);\n\nexport const RecallSchema = z.object({\n  matched: z.number().int().nonnegative(),\n  total: z.number().int().nonnegative(),\n  ratio: z.number().min(0).max(1),\n});\n\nexport const PerCategoryRecallSchema = z.record(z.string(), RecallSchema);\n\nexport const PassFailSchema = z.object({\n  fp_ceiling: z.enum(['PASS', 'FAIL']),\n  any_flag_recall_floor: z.enum(['PASS', 'FAIL']),\n  per_category_any_flag_floor: z.enum(['PASS', 'FAIL']),\n  strict_recall_floor: z.enum(['PASS', 'FAIL']),\n  decision_vocab_completeness: z.enum(['PASS', 'FAIL']),\n  latency_soft: z.enum(['PASS', 'WARN']),\n  latency_hard: z.enum(['PASS', 'FAIL']),\n  empty_or_malformed: z.enum(['PASS', 'FAIL']),\n  overall: z.enum(['PASS', 'FAIL']),\n});\n\nexport const DecisionVocabBarSchema = z.object({\n  architecture: ArchitectureSchema,\n  required: z.number().int().positive(),\n  produced: z.number().int().nonnegative(),\n  passed: z.boolean(),\n});\n\nexport const CalibrationReceiptSchema = z.object({\n  schema_version: z.literal(1),\n  profile_name: z.string(),\n  status: StatusLabelSchema,\n  model: z.string(),\n  architecture: ArchitectureSchema,\n  fixture: z.string(),\n  fixture_total_claims: z.number().int().positive(),\n  fixture_good_claims: z.number().int().nonnegative(),\n  fixture_bad_claims: z.number().int().nonnegative(),\n  calibrated_at: z.string(),\n  research_os_version: z.string(),\n  runtime_ms: z.number().int().nonnegative(),\n  good_fp_count: z.number().int().nonnegative(),\n  any_flag_recall: RecallSchema,\n  strict_recall: RecallSchema,\n  per_category_any_flag: PerCategoryRecallSchema,\n  per_category_strict: PerCategoryRecallSchema,\n  decision_vocabulary: z.record(z.string(), z.number().int().nonnegative()),\n  decisions_produced_count: z.number().int().nonnegative(),\n  decision_vocab_bar: DecisionVocabBarSchema,\n  unreachable_decisions: z.array(z.string()),\n  empty_or_malformed_responses: z.number().int().nonnegative(),\n  pass_fail: PassFailSchema,\n  notes: z.array(z.string()),\n});\n\nexport type StatusLabel = z.infer<typeof StatusLabelSchema>;\nexport type Architecture = z.infer<typeof ArchitectureSchema>;\nexport type Recall = z.infer<typeof RecallSchema>;\nexport type PerCategoryRecall = z.infer<typeof PerCategoryRecallSchema>;\nexport type PassFail = z.infer<typeof PassFailSchema>;\nexport type DecisionVocabBar = z.infer<typeof DecisionVocabBarSchema>;\nexport type CalibrationReceipt = z.infer<typeof CalibrationReceiptSchema>;\n"],"mappings":";AAAA,SAAS,KAAAA,UAAS;;;ACAlB,SAAS,SAAS;AAEX,IAAM,oBAAoB,EAAE,KAAK;AAAA,EACtC;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF,CAAC;AAEM,IAAM,qBAAqB,EAAE,KAAK,CAAC,eAAe,UAAU,CAAC;AAE7D,IAAM,eAAe,EAAE,OAAO;AAAA,EACnC,SAAS,EAAE,OAAO,EAAE,IAAI,EAAE,YAAY;AAAA,EACtC,OAAO,EAAE,OAAO,EAAE,IAAI,EAAE,YAAY;AAAA,EACpC,OAAO,EAAE,OAAO,EAAE,IAAI,CAAC,EAAE,IAAI,CAAC;AAChC,CAAC;AAEM,IAAM,0BAA0B,EAAE,OAAO,EAAE,OAAO,GAAG,YAAY;AAEjE,IAAM,iBAAiB,EAAE,OAAO;AAAA,EACrC,YAAY,EAAE,KAAK,CAAC,QAAQ,MAAM,CAAC;AAAA,EACnC,uBAAuB,EAAE,KAAK,CAAC,QAAQ,MAAM,CAAC;AAAA,EAC9C,6BAA6B,EAAE,KAAK,CAAC,QAAQ,MAAM,CAAC;AAAA,EACpD,qBAAqB,EAAE,KAAK,CAAC,QAAQ,MAAM,CAAC;AAAA,EAC5C,6BAA6B,EAAE,KAAK,CAAC,QAAQ,MAAM,CAAC;AAAA,EACpD,cAAc,EAAE,KAAK,CAAC,QAAQ,MAAM,CAAC;AAAA,EACrC,cAAc,EAAE,KAAK,CAAC,QAAQ,MAAM,CAAC;AAAA,EACrC,oBAAoB,EAAE,KAAK,CAAC,QAAQ,MAAM,CAAC;AAAA,EAC3C,SAAS,EAAE,KAAK,CAAC,QAAQ,MAAM,CAAC;AAClC,CAAC;AAEM,IAAM,yBAAyB,EAAE,OAAO;AAAA,EAC7C,cAAc;AAAA,EACd,UAAU,EAAE,OAAO,EAAE,IAAI,EAAE,SAAS;AAAA,EACpC,UAAU,EAAE,OAAO,EAAE,IAAI,EAAE,YAAY;AAAA,EACvC,QAAQ,EAAE,QAAQ;AACpB,CAAC;AAEM,IAAM,2BAA2B,EAAE,OAAO;AAAA,EAC/C,gBAAgB,EAAE,QAAQ,CAAC;AAAA,EAC3B,cAAc,EAAE,OAAO;AAAA,EACvB,QAAQ;AAAA,EACR,OAAO,EAAE,OAAO;AAAA,EAChB,cAAc;AAAA,EACd,SAAS,EAAE,OAAO;AAAA,EAClB,sBAAsB,EAAE,OAAO,EAAE,IAAI,EAAE,SAAS;AAAA,EAChD,qBAAqB,EAAE,OAAO,EAAE,IAAI,EAAE,YAAY;AAAA,EAClD,oBAAoB,EAAE,OAAO,EAAE,IAAI,EAAE,YAAY;AAAA,EACjD,eAAe,EAAE,OAAO;AAAA,EACxB,qBAAqB,EAAE,OAAO;AAAA,EAC9B,YAAY,EAAE,OAAO,EAAE,IAAI,EAAE,YAAY;AAAA,EACzC,eAAe,EAAE,OAAO,EAAE,IAAI,EAAE,YAAY;AAAA,EAC5C,iBAAiB;AAAA,EACjB,eAAe;AAAA,EACf,uBAAuB;AAAA,EACvB,qBAAqB;AAAA,EACrB,qBAAqB,EAAE,OAAO,EAAE,OAAO,GAAG,EAAE,OAAO,EAAE,IAAI,EAAE,YAAY,CAAC;AAAA,EACxE,0BAA0B,EAAE,OAAO,EAAE,IAAI,EAAE,YAAY;AAAA,EACvD,oBAAoB;AAAA,EACpB,uBAAuB,EAAE,MAAM,EAAE,OAAO,CAAC;AAAA,EACzC,8BAA8B,EAAE,OAAO,EAAE,IAAI,EAAE,YAAY;AAAA,EAC3D,WAAW;AAAA,EACX,OAAO,EAAE,MAAM,EAAE,OAAO,CAAC;AAC3B,CAAC;;;AD5DM,IAAM,wBAAwBC,GAAE,OAAO;AAAA,EAC5C,QAAQA,GAAE,OAAO;AAAA,EACjB,KAAKA,GAAE,OAAO;AAAA,EACd,KAAKA,GAAE,OAAO;AAAA,EACd,QAAQA,GAAE,MAAMA,GAAE,OAAO,CAAC;AAAA;AAC5B,CAAC;AAEM,IAAM,kCAAkCA,GAAE,OAAO;AAAA,EACtD,cAAcA,GAAE,OAAO,EAAE,IAAI,CAAC,EAAE,IAAI,CAAC;AAAA,EACrC,WAAWA,GAAE,OAAO,EAAE,IAAI,CAAC,EAAE,IAAI,CAAC;AAAA,EAClC,WAAWA,GAAE,OAAO,EAAE,IAAI,CAAC,EAAE,IAAI,CAAC;AAAA,EAClC,OAAOA,GAAE,OAAO,EAAE,IAAI,EAAE,YAAY;AAAA;AAAA,EACpC,gBAAgBA,GAAE,MAAMA,GAAE,OAAO,CAAC;AACpC,CAAC;AAEM,IAAM,6BAA6BA,GAAE,OAAOA,GAAE,OAAO,GAAG,+BAA+B;AAEvF,IAAM,0BAA0BA,GAAE,OAAO;AAAA,EAC9C,YAAYA,GAAE,KAAK,CAAC,QAAQ,MAAM,CAAC;AAAA,EACnC,uBAAuBA,GAAE,KAAK,CAAC,QAAQ,MAAM,CAAC;AAAA,EAC9C,6BAA6BA,GAAE,KAAK,CAAC,QAAQ,MAAM,CAAC;AAAA,EACpD,qBAAqBA,GAAE,KAAK,CAAC,QAAQ,MAAM,CAAC;AAAA,EAC5C,6BAA6BA,GAAE,KAAK,CAAC,QAAQ,MAAM,CAAC;AAAA,EACpD,cAAcA,GAAE,KAAK,CAAC,QAAQ,MAAM,CAAC;AAAA,EACrC,cAAcA,GAAE,KAAK,CAAC,QAAQ,MAAM,CAAC;AAAA,EACrC,oBAAoBA,GAAE,KAAK,CAAC,QAAQ,MAAM,CAAC;AAAA,EAC3C,SAASA,GAAE,KAAK,CAAC,QAAQ,MAAM,CAAC;AAClC,CAAC;AAEM,IAAM,kCAAkCA,GAAE,OAAO;AAAA,EACtD,cAAc;AAAA,EACd,UAAUA,GAAE,OAAO,EAAE,IAAI,EAAE,SAAS;AAAA,EACpC,iBAAiBA,GAAE,OAAO;AAAA;AAAA,EAC1B,QAAQA,GAAE,QAAQ;AACpB,CAAC;AAEM,IAAM,oCAAoCA,GAAE,OAAO;AAAA,EACxD,gBAAgBA,GAAE,QAAQ,CAAC;AAAA,EAC3B,cAAcA,GAAE,QAAQ,WAAW;AAAA;AAAA,EACnC,cAAcA,GAAE,OAAO;AAAA,EACvB,QAAQ;AAAA,EACR,OAAOA,GAAE,OAAO;AAAA,EAChB,cAAc;AAAA,EACd,SAASA,GAAE,OAAO;AAAA,EAClB,sBAAsBA,GAAE,OAAO,EAAE,IAAI,EAAE,SAAS;AAAA,EAChD,qBAAqBA,GAAE,OAAO,EAAE,IAAI,EAAE,YAAY;AAAA,EAClD,oBAAoBA,GAAE,OAAO,EAAE,IAAI,EAAE,YAAY;AAAA,EACjD,YAAYA,GAAE,OAAO,EAAE,IAAI,EAAE,IAAI,CAAC;AAAA,EAClC,WAAWA,GAAE,MAAMA,GAAE,OAAO,CAAC;AAAA;AAAA,EAC7B,eAAeA,GAAE,OAAO;AAAA;AAAA,EACxB,qBAAqBA,GAAE,OAAO;AAAA;AAAA,EAG9B,eAAe;AAAA,EACf,uBAAuB;AAAA,EACvB,qBAAqB;AAAA,EACrB,0BAA0B;AAAA,EAC1B,YAAY;AAAA,EACZ,8BAA8B;AAAA,EAE9B,uBAAuB;AAAA,EACvB,qBAAqB;AAAA;AAAA,EAGrB,qBAAqBA,GAAE,OAAOA,GAAE,OAAO,GAAG,qBAAqB;AAAA,EAC/D,oBAAoB;AAAA,EACpB,uBAAuBA,GAAE,MAAMA,GAAE,OAAO,CAAC;AAAA,EAEzC,WAAW;AAAA;AAAA;AAAA,EAGX,wBAAwBA,GAAE,MAAMA,GAAE,OAAO,CAAC;AAAA,EAE1C,OAAOA,GAAE,MAAMA,GAAE,OAAO,CAAC;AAC3B,CAAC;","names":["z","z"]}