promptfoo 0.120.25 → 0.120.26

package/dist/src/{transform-CCSNqTI-.cjs → transform-ivxEY4f7.cjs}

@@ -1,6 +1,6 @@
 const require_logger = require('./logger-BdZ-IqBc.cjs');
-const require_util = require('./util-Dab5xp2E.cjs');
-const require_fetch = require('./fetch-Dy2YML9G.cjs');
+const require_util = require('./util-DUYOvxAy.cjs');
+const require_fetch = require('./fetch-ZMn_oemb.cjs');
 let rfdc = require("rfdc");
 rfdc = require_logger.__toESM(rfdc);
 let zod = require("zod");
@@ -269,6 +269,20 @@ function extractBase64FromDataUrl(value) {
 	const parsed = parseDataUrl(value);
 	return parsed ? parsed.base64Data : value;
 }
+/**
+* Build a data URI from a MIME type and base64 data.
+*
+* @param mimeType MIME type (e.g. "image/png")
+* @param base64Data Raw base64-encoded data
+* @returns Data URI string
+*
+* @example
+* toDataUri("image/png", "iVBORw0KGgo...")
+* // "data:image/png;base64,iVBORw0KGgo..."
+*/
+function toDataUri(mimeType, base64Data) {
+	return `data:${mimeType};base64,${base64Data}`;
+}
 
 //#endregion
 //#region src/providers/google/auth.ts
@@ -565,6 +579,20 @@ const clearCachedAuth = GoogleAuthManager.clearCache.bind(GoogleAuthManager);
 * Note: Vertex AI may have different pricing for some models.
 */
 const GOOGLE_MODELS = [
+	{
+		id: "gemini-3.1-pro-preview",
+		cost: {
+			input: 2 / 1e6,
+			output: 12 / 1e6
+		},
+		tieredCost: {
+			threshold: 2e5,
+			above: {
+				input: 4 / 1e6,
+				output: 18 / 1e6
+			}
+		}
+	},
 	{
 		id: "gemini-3-flash-preview",
 		cost: {
@@ -831,6 +859,17 @@ const VALID_SCHEMA_TYPES = [
 //#endregion
 //#region src/providers/google/util.ts
 /**
+* Normalizes safety settings to use the correct Google API field name `threshold`.
+* Accepts the legacy `probability` field for backwards compatibility and maps it to `threshold`.
+*/
+function normalizeSafetySettings(safetySettings) {
+	if (!safetySettings) return;
+	return safetySettings.map(({ category, threshold, probability }) => ({
+		category,
+		threshold: threshold || probability || ""
+	}));
+}
+/**
 * Calculates the cost for a Google AI Studio API call.
 *
 * Handles tiered pricing for models where cost varies by prompt size.
@@ -1539,6 +1578,12 @@ Object.defineProperty(exports, 'mergeParts', {
     return mergeParts;
   }
 });
+Object.defineProperty(exports, 'normalizeSafetySettings', {
+  enumerable: true,
+  get: function () {
+    return normalizeSafetySettings;
+  }
+});
 Object.defineProperty(exports, 'normalizeTools', {
   enumerable: true,
   get: function () {
@@ -1563,6 +1608,12 @@ Object.defineProperty(exports, 'resolveProjectId', {
     return resolveProjectId;
   }
 });
+Object.defineProperty(exports, 'toDataUri', {
+  enumerable: true,
+  get: function () {
+    return toDataUri;
+  }
+});
 Object.defineProperty(exports, 'transformMCPConfigToClaudeCode', {
   enumerable: true,
   get: function () {
@@ -1593,4 +1644,4 @@ Object.defineProperty(exports, 'validateFunctionCall', {
     return validateFunctionCall;
   }
 });
-//# sourceMappingURL=transform-CCSNqTI-.cjs.map
+//# sourceMappingURL=transform-ivxEY4f7.cjs.map

package/dist/src/{transformersAvailability-DM54tWzt.js → transformersAvailability-DwmezkVe.js}

@@ -33,4 +33,4 @@ async function validateTransformersDependency(checkTransformers = isTransformers
 
 //#endregion
 export { validateTransformersDependency };
-//# sourceMappingURL=transformersAvailability-DM54tWzt.js.map
+//# sourceMappingURL=transformersAvailability-DwmezkVe.js.map

package/dist/src/{types-Brw5kggg.cjs → types-BIfttHrT.cjs}

@@ -361,6 +361,7 @@ const FINANCIAL_PLUGINS = [
 	"financial:hallucination",
 	"financial:impartiality",
 	"financial:misconduct",
+	"financial:sox-compliance",
 	"financial:sycophancy"
 ];
 const PHARMACY_PLUGINS = [
@@ -431,6 +432,7 @@ const ADDITIONAL_PLUGINS = [
 	"imitation",
 	"indirect-prompt-injection",
 	"mcp",
+	"model-identification",
 	"medical:anchoring-bias",
 	"medical:hallucination",
 	"medical:incorrect-knowledge",
@@ -446,6 +448,7 @@ const ADDITIONAL_PLUGINS = [
 	"financial:hallucination",
 	"financial:impartiality",
 	"financial:misconduct",
+	"financial:sox-compliance",
 	"financial:sycophancy",
 	"ecommerce:compliance-bypass",
 	"ecommerce:order-fraud",
@@ -571,6 +574,7 @@ const REMOTE_ONLY_PLUGIN_IDS = [
 	"hijacking",
 	"indirect-prompt-injection",
 	"mcp",
+	"model-identification",
 	"off-topic",
 	"rag-document-exfiltration",
 	"rag-poisoning",
@@ -1572,6 +1576,7 @@ const riskCategorySeverityMap = {
 	"financial:hallucination": Severity.Low,
 	"financial:impartiality": Severity.Medium,
 	"financial:misconduct": Severity.High,
+	"financial:sox-compliance": Severity.High,
 	"financial:sycophancy": Severity.Low,
 	"goal-misalignment": Severity.Low,
 	competitors: Severity.Low,
@@ -1593,6 +1598,7 @@ const riskCategorySeverityMap = {
 	"off-topic": Severity.Medium,
 	"divergent-repetition": Severity.Medium,
 	"excessive-agency": Severity.Medium,
+	"model-identification": Severity.Medium,
 	"tool-discovery": Severity.Low,
 	foundation: Severity.Medium,
 	"guardrails-eval": Severity.Medium,
@@ -1707,6 +1713,7 @@ const riskCategories = {
 		"bola",
 		"cca",
 		"debug-access",
+		"model-identification",
 		"hijacking",
 		"indirect-prompt-injection",
 		"rbac",
@@ -1802,6 +1809,7 @@ const riskCategories = {
 		"financial:hallucination",
 		"financial:impartiality",
 		"financial:misconduct",
+		"financial:sox-compliance",
 		"financial:sycophancy",
 		"medical:hallucination",
 		"medical:anchoring-bias",
@@ -1892,6 +1900,7 @@ const categoryAliases = {
 	"financial:hallucination": "FinancialHallucination",
 	"financial:impartiality": "FinancialImpartiality",
 	"financial:misconduct": "FinancialMisconduct",
+	"financial:sox-compliance": "FinancialSoxCompliance",
 	"financial:sycophancy": "FinancialSycophancy",
 	"goal-misalignment": "GoalMisalignment",
 	"off-topic": "OffTopic",
@@ -1900,6 +1909,7 @@ const categoryAliases = {
 	"pharmacy:drug-interaction": "PharmacyDrugInteraction",
 	"divergent-repetition": "DivergentRepetition",
 	"excessive-agency": "ExcessiveAgency",
+	"model-identification": "ModelIdentification",
 	"tool-discovery": "ToolDiscovery",
 	foundation: "Foundation",
 	"guardrails-eval": "GuardrailsEvaluation",
@@ -2023,6 +2033,7 @@ const pluginDescriptions = {
 	default: "Executes comprehensive baseline security testing across multiple risk categories",
 	"divergent-repetition": "Tests repetitive patterns that can cause the model to diverge from normal behavior and leak training data",
 	"excessive-agency": "Evaluates system boundary enforcement and unauthorized action prevention",
+	"model-identification": "Tests whether an AI system reveals its underlying model identity, version, or creator when probed",
 	"tool-discovery": "Tests for enumeration of available tools and function calls",
 	ferpa: "Tests for Family Educational Rights and Privacy Act compliance",
 	foundation: "Tests a collection of plugins designed to run against foundation models",
@@ -2113,6 +2124,7 @@ const pluginDescriptions = {
 	"financial:hallucination": "Tests for fabricated market data, non-existent financial instruments, fictional company financials, or invented market events",
 	"financial:impartiality": "Tests for inappropriate provision of financial advice, investment recommendations, broker suggestions, or transaction facilitation without proper authorization or disclaimers",
 	"financial:misconduct": "Tests for facilitation of financial crimes including insider trading, market manipulation, front-running, spoofing, bribery, or circumventing financial regulations",
+	"financial:sox-compliance": "Tests for violations of Sarbanes-Oxley requirements including internal controls bypass, financial statement manipulation, CEO/CFO certification fraud, audit obstruction, document retention violations, and whistleblower retaliation",
 	"financial:sycophancy": "Tests for agreeing with risky trading strategies, unrealistic return expectations, or get-rich-quick schemes",
 	"goal-misalignment": "Tests whether AI systems recognize when optimizing measurable proxy metrics might not align with true underlying objectives (Goodhart's Law: \"When a measure becomes a target, it ceases to be a good measure\")",
 	"off-topic": "Tests whether AI systems can be manipulated to go off-topic by performing tasks outside their domain",
@@ -3708,4 +3720,4 @@ Object.defineProperty(exports, 'riskCategorySeverityMap', {
     return riskCategorySeverityMap;
   }
 });
-//# sourceMappingURL=types-Brw5kggg.cjs.map
+//# sourceMappingURL=types-BIfttHrT.cjs.map

package/dist/src/{types-CslP2TyG.js → types-DMVjYLpx.js}

@@ -359,6 +359,7 @@ const FINANCIAL_PLUGINS = [
 	"financial:hallucination",
 	"financial:impartiality",
 	"financial:misconduct",
+	"financial:sox-compliance",
 	"financial:sycophancy"
 ];
 const PHARMACY_PLUGINS = [
@@ -429,6 +430,7 @@ const ADDITIONAL_PLUGINS = [
 	"imitation",
 	"indirect-prompt-injection",
 	"mcp",
+	"model-identification",
 	"medical:anchoring-bias",
 	"medical:hallucination",
 	"medical:incorrect-knowledge",
@@ -444,6 +446,7 @@ const ADDITIONAL_PLUGINS = [
 	"financial:hallucination",
 	"financial:impartiality",
 	"financial:misconduct",
+	"financial:sox-compliance",
 	"financial:sycophancy",
 	"ecommerce:compliance-bypass",
 	"ecommerce:order-fraud",
@@ -569,6 +572,7 @@ const REMOTE_ONLY_PLUGIN_IDS = [
 	"hijacking",
 	"indirect-prompt-injection",
 	"mcp",
+	"model-identification",
 	"off-topic",
 	"rag-document-exfiltration",
 	"rag-poisoning",
@@ -1570,6 +1574,7 @@ const riskCategorySeverityMap = {
 	"financial:hallucination": Severity.Low,
 	"financial:impartiality": Severity.Medium,
 	"financial:misconduct": Severity.High,
+	"financial:sox-compliance": Severity.High,
 	"financial:sycophancy": Severity.Low,
 	"goal-misalignment": Severity.Low,
 	competitors: Severity.Low,
@@ -1591,6 +1596,7 @@ const riskCategorySeverityMap = {
 	"off-topic": Severity.Medium,
 	"divergent-repetition": Severity.Medium,
 	"excessive-agency": Severity.Medium,
+	"model-identification": Severity.Medium,
 	"tool-discovery": Severity.Low,
 	foundation: Severity.Medium,
 	"guardrails-eval": Severity.Medium,
@@ -1705,6 +1711,7 @@ const riskCategories = {
 		"bola",
 		"cca",
 		"debug-access",
+		"model-identification",
 		"hijacking",
 		"indirect-prompt-injection",
 		"rbac",
@@ -1800,6 +1807,7 @@ const riskCategories = {
 		"financial:hallucination",
 		"financial:impartiality",
 		"financial:misconduct",
+		"financial:sox-compliance",
 		"financial:sycophancy",
 		"medical:hallucination",
 		"medical:anchoring-bias",
@@ -1890,6 +1898,7 @@ const categoryAliases = {
 	"financial:hallucination": "FinancialHallucination",
 	"financial:impartiality": "FinancialImpartiality",
 	"financial:misconduct": "FinancialMisconduct",
+	"financial:sox-compliance": "FinancialSoxCompliance",
 	"financial:sycophancy": "FinancialSycophancy",
 	"goal-misalignment": "GoalMisalignment",
 	"off-topic": "OffTopic",
@@ -1898,6 +1907,7 @@ const categoryAliases = {
 	"pharmacy:drug-interaction": "PharmacyDrugInteraction",
 	"divergent-repetition": "DivergentRepetition",
 	"excessive-agency": "ExcessiveAgency",
+	"model-identification": "ModelIdentification",
 	"tool-discovery": "ToolDiscovery",
 	foundation: "Foundation",
 	"guardrails-eval": "GuardrailsEvaluation",
@@ -2021,6 +2031,7 @@ const pluginDescriptions = {
 	default: "Executes comprehensive baseline security testing across multiple risk categories",
 	"divergent-repetition": "Tests repetitive patterns that can cause the model to diverge from normal behavior and leak training data",
 	"excessive-agency": "Evaluates system boundary enforcement and unauthorized action prevention",
+	"model-identification": "Tests whether an AI system reveals its underlying model identity, version, or creator when probed",
 	"tool-discovery": "Tests for enumeration of available tools and function calls",
 	ferpa: "Tests for Family Educational Rights and Privacy Act compliance",
 	foundation: "Tests a collection of plugins designed to run against foundation models",
@@ -2111,6 +2122,7 @@ const pluginDescriptions = {
 	"financial:hallucination": "Tests for fabricated market data, non-existent financial instruments, fictional company financials, or invented market events",
 	"financial:impartiality": "Tests for inappropriate provision of financial advice, investment recommendations, broker suggestions, or transaction facilitation without proper authorization or disclaimers",
 	"financial:misconduct": "Tests for facilitation of financial crimes including insider trading, market manipulation, front-running, spoofing, bribery, or circumventing financial regulations",
+	"financial:sox-compliance": "Tests for violations of Sarbanes-Oxley requirements including internal controls bypass, financial statement manipulation, CEO/CFO certification fraud, audit obstruction, document retention violations, and whistleblower retaliation",
 	"financial:sycophancy": "Tests for agreeing with risky trading strategies, unrealistic return expectations, or get-rich-quick schemes",
 	"goal-misalignment": "Tests whether AI systems recognize when optimizing measurable proxy metrics might not align with true underlying objectives (Goodhart's Law: \"When a measure becomes a target, it ceases to be a good measure\")",
 	"off-topic": "Tests whether AI systems can be manipulated to go off-topic by performing tasks outside their domain",
@@ -3233,4 +3245,4 @@ const EvalResultsFilterMode = z.enum([
 
 //#endregion
 export { pluginDescriptions as $, isApiProvider as A, isValidReusablePolicyId as B, TestGeneratorConfigSchema as C, BaseTokenUsageSchema as Ct, VarsSchema as D, UnifiedConfigSchema as E, ConversationMessageSchema as F, MULTI_TURN_STRATEGIES as G, AGENTIC_STRATEGIES as H, PartialGenerationError as I, getDefaultNFanout as J, STRATEGY_COLLECTIONS as K, PluginConfigSchema as L, RedteamConfigSchema as M, ProviderOptionsSchema as N, isGradingResult as O, ProvidersSchema as P, categoryAliases as Q, PolicyObjectSchema as R, TestCasesWithMetadataSchema as S, PromptSchema as St, TestSuiteSchema as T, InputsSchema as Tt, DATASET_PLUGINS as U, isUuid as V, DEFAULT_STRATEGIES as W, isFanoutStrategy as X, isCustomStrategy as Y, Severity as Z, ScenarioSchema as _, REDTEAM_PROVIDER_HARM_PLUGINS as _t, AtomicTestCaseSchema as a, FINANCIAL_PLUGINS as at, TestCaseWithVarsFileSchema as b, TELECOM_PLUGINS as bt, CompletedPromptSchema as c, INSURANCE_PLUGINS as ct, EvaluateOptionsSchema as d, MEDICAL_PLUGINS as dt, riskCategorySeverityMap as et, GradingConfigSchema as f, MULTI_INPUT_EXCLUDED_PLUGINS as ft, ResultFailureReason as g, PLUGIN_CATEGORIES as gt, OutputFileExtension as h, PII_PLUGINS as ht, AssertionTypeSchema as i, DEFAULT_PLUGINS as it, isProviderOptions as j, isResultFailureReason as k, DerivedMetricSchema as l, LLAMA_GUARD_ENABLED_CATEGORIES as lt, OutputConfigSchema as m, PHARMACY_PLUGINS as mt, AssertionSchema as n, BIAS_PLUGINS as nt, BaseAssertionTypesSchema as o, FOUNDATION_PLUGINS as ot, NotPrefixedAssertionTypesSchema as p, MULTI_INPUT_VAR as pt, STRATEGY_COLLECTION_MAPPINGS as q, AssertionSetSchema as r, DATASET_EXEMPT_PLUGINS as rt, CommandLineOptionsSchema as s, HARM_PLUGINS as st, AssertionOrSetSchema as t, ALIASED_PLUGIN_MAPPINGS as tt, EvalResultsFilterMode as u, LLAMA_GUARD_REPLICATE_PROVIDER as ut, SpecialAssertionTypesSchema as v, REMOTE_ONLY_PLUGIN_IDS as vt, TestSuiteConfigSchema as w, CompletionTokenDetailsSchema as wt, TestCasesWithMetadataPromptSchema as x, UNALIGNED_PROVIDER_HARM_PLUGINS as xt, TestCaseSchema as y, STRATEGY_EXEMPT_PLUGINS as yt, StrategyConfigSchema as z };
-//# sourceMappingURL=types-CslP2TyG.js.map
+//# sourceMappingURL=types-DMVjYLpx.js.map

package/dist/src/{types-CMDs7Kzu.js → types-t52w-XsS.js}

@@ -1,5 +1,5 @@
 #!/usr/bin/env node
-import { i as isJavascriptFile, t as JAVASCRIPT_EXTENSIONS } from "./fileExtensions-CzcPkiKZ.js";
+import { i as isJavascriptFile, t as JAVASCRIPT_EXTENSIONS } from "./fileExtensions-BpuMmaFL.js";
 import { z } from "zod";
 import dedent from "dedent";
 
@@ -220,6 +220,7 @@ const FINANCIAL_PLUGINS = [
 	"financial:hallucination",
 	"financial:impartiality",
 	"financial:misconduct",
+	"financial:sox-compliance",
 	"financial:sycophancy"
 ];
 const PHARMACY_PLUGINS = [
@@ -290,6 +291,7 @@ const ADDITIONAL_PLUGINS = [
 	"imitation",
 	"indirect-prompt-injection",
 	"mcp",
+	"model-identification",
 	"medical:anchoring-bias",
 	"medical:hallucination",
 	"medical:incorrect-knowledge",
@@ -305,6 +307,7 @@ const ADDITIONAL_PLUGINS = [
 	"financial:hallucination",
 	"financial:impartiality",
 	"financial:misconduct",
+	"financial:sox-compliance",
 	"financial:sycophancy",
 	"ecommerce:compliance-bypass",
 	"ecommerce:order-fraud",
@@ -430,6 +433,7 @@ const REMOTE_ONLY_PLUGIN_IDS = [
 	"hijacking",
 	"indirect-prompt-injection",
 	"mcp",
+	"model-identification",
 	"off-topic",
 	"rag-document-exfiltration",
 	"rag-poisoning",
@@ -1494,6 +1498,7 @@ const subCategoryDescriptions = {
 	"financial:hallucination": "Tests for fabricated market data, non-existent financial instruments, or fictional company information",
 	"financial:impartiality": "Tests for unauthorized financial advice or transaction facilitation",
 	"financial:misconduct": "Tests for facilitation of financial crimes or market manipulation",
+	"financial:sox-compliance": "Tests for violations of Sarbanes-Oxley requirements including internal controls, financial reporting, and audit integrity",
 	"financial:sycophancy": "Tests for agreeing with risky investment strategies or validating get-rich-quick schemes",
 	"goal-misalignment": "Tests whether AI systems recognize when optimizing proxy metrics might not align with true goals (Goodhart's Law)",
 	"off-topic": "Tests whether AI systems can be manipulated to go off-topic from their intended purpose",
@@ -1528,6 +1533,7 @@ const subCategoryDescriptions = {
 	"sql-injection": "Tests for SQL injection vulnerabilities",
 	ssrf: "Tests for server-side request forgery vulnerabilities",
 	"system-prompt-override": "Tests for system prompt override vulnerabilities",
+	"model-identification": "Tests whether an AI system can be tricked into revealing its underlying model identity",
 	"tool-discovery": "Tests for enumeration of available tools and function calls",
 	unsafebench: "Tests handling of unsafe image content from the UnsafeBench dataset",
 	"unverifiable-claims": "Tests for claims that cannot be verified or fact-checked",
@@ -1612,6 +1618,7 @@ const riskCategorySeverityMap = {
 	"financial:hallucination": Severity.Low,
 	"financial:impartiality": Severity.Medium,
 	"financial:misconduct": Severity.High,
+	"financial:sox-compliance": Severity.High,
 	"financial:sycophancy": Severity.Low,
 	"goal-misalignment": Severity.Low,
 	competitors: Severity.Low,
@@ -1633,6 +1640,7 @@ const riskCategorySeverityMap = {
 	"off-topic": Severity.Medium,
 	"divergent-repetition": Severity.Medium,
 	"excessive-agency": Severity.Medium,
+	"model-identification": Severity.Medium,
 	"tool-discovery": Severity.Low,
 	foundation: Severity.Medium,
 	"guardrails-eval": Severity.Medium,
@@ -1747,6 +1755,7 @@ const riskCategories = {
 		"bola",
 		"cca",
 		"debug-access",
+		"model-identification",
 		"hijacking",
 		"indirect-prompt-injection",
 		"rbac",
@@ -1842,6 +1851,7 @@ const riskCategories = {
 		"financial:hallucination",
 		"financial:impartiality",
 		"financial:misconduct",
+		"financial:sox-compliance",
 		"financial:sycophancy",
 		"medical:hallucination",
 		"medical:anchoring-bias",
@@ -1932,6 +1942,7 @@ const categoryAliases = {
 	"financial:hallucination": "FinancialHallucination",
 	"financial:impartiality": "FinancialImpartiality",
 	"financial:misconduct": "FinancialMisconduct",
+	"financial:sox-compliance": "FinancialSoxCompliance",
 	"financial:sycophancy": "FinancialSycophancy",
 	"goal-misalignment": "GoalMisalignment",
 	"off-topic": "OffTopic",
@@ -1940,6 +1951,7 @@ const categoryAliases = {
 	"pharmacy:drug-interaction": "PharmacyDrugInteraction",
 	"divergent-repetition": "DivergentRepetition",
 	"excessive-agency": "ExcessiveAgency",
+	"model-identification": "ModelIdentification",
 	"tool-discovery": "ToolDiscovery",
 	foundation: "Foundation",
 	"guardrails-eval": "GuardrailsEvaluation",
@@ -2063,6 +2075,7 @@ const pluginDescriptions = {
 	default: "Executes comprehensive baseline security testing across multiple risk categories",
 	"divergent-repetition": "Tests repetitive patterns that can cause the model to diverge from normal behavior and leak training data",
 	"excessive-agency": "Evaluates system boundary enforcement and unauthorized action prevention",
+	"model-identification": "Tests whether an AI system reveals its underlying model identity, version, or creator when probed",
 	"tool-discovery": "Tests for enumeration of available tools and function calls",
 	ferpa: "Tests for Family Educational Rights and Privacy Act compliance",
 	foundation: "Tests a collection of plugins designed to run against foundation models",
@@ -2153,6 +2166,7 @@ const pluginDescriptions = {
 	"financial:hallucination": "Tests for fabricated market data, non-existent financial instruments, fictional company financials, or invented market events",
 	"financial:impartiality": "Tests for inappropriate provision of financial advice, investment recommendations, broker suggestions, or transaction facilitation without proper authorization or disclaimers",
 	"financial:misconduct": "Tests for facilitation of financial crimes including insider trading, market manipulation, front-running, spoofing, bribery, or circumventing financial regulations",
+	"financial:sox-compliance": "Tests for violations of Sarbanes-Oxley requirements including internal controls bypass, financial statement manipulation, CEO/CFO certification fraud, audit obstruction, document retention violations, and whistleblower retaliation",
 	"financial:sycophancy": "Tests for agreeing with risky trading strategies, unrealistic return expectations, or get-rich-quick schemes",
 	"goal-misalignment": "Tests whether AI systems recognize when optimizing measurable proxy metrics might not align with true underlying objectives (Goodhart's Law: \"When a measure becomes a target, it ceases to be a good measure\")",
 	"off-topic": "Tests whether AI systems can be manipulated to go off-topic by performing tasks outside their domain",
@@ -3420,4 +3434,4 @@ const EvalResultsFilterMode = z.enum([
 
 //#endregion
 export { HARM_PLUGINS as $, DATASET_PLUGINS as A, Severity as B, PolicyObjectSchema as C, ADDITIONAL_STRATEGIES as D, isUuid as E, STRATEGY_COLLECTION_MAPPINGS as F, ALIASED_PLUGIN_MAPPINGS as G, pluginDescriptions as H, getDefaultNFanout as I, BIAS_PLUGINS as J, ADDITIONAL_PLUGINS as K, isCustomStrategy as L, DEFAULT_STRATEGIES as M, MULTI_TURN_STRATEGIES as N, AGENTIC_STRATEGIES as O, STRATEGY_COLLECTIONS as P, FOUNDATION_PLUGINS as Q, isFanoutStrategy as R, PluginConfigSchema as S, isValidReusablePolicyId as T, riskCategorySeverityMap as U, categoryAliases as V, subCategoryDescriptions as W, DEFAULT_PLUGINS as X, DATASET_EXEMPT_PLUGINS as Y, FINANCIAL_PLUGINS as Z, PromptSchema as _, EvaluateOptionsSchema as a, MULTI_INPUT_VAR as at, ConversationMessageSchema as b, TestSuiteConfigSchema as c, PLUGIN_CATEGORIES as ct, isGradingResult as d, REMOTE_ONLY_PLUGIN_IDS as dt, INSURANCE_PLUGINS as et, isResultFailureReason as f, STRATEGY_EXEMPT_PLUGINS as ft, RedteamGenerateOptionsSchema as g, RedteamConfigSchema as h, EvalResultsFilterMode as i, MULTI_INPUT_EXCLUDED_PLUGINS as it, DEFAULT_MULTI_TURN_MAX_TURNS as j, ALL_STRATEGIES as k, TestSuiteSchema as l, REDTEAM_MODEL as lt, isProviderOptions as m, UNALIGNED_PROVIDER_HARM_PLUGINS as mt, BaseAssertionTypesSchema as n, LLAMA_GUARD_REPLICATE_PROVIDER as nt, OutputFileExtension as o, PHARMACY_PLUGINS as ot, isApiProvider as p, TELECOM_PLUGINS as pt, ALL_PLUGINS as q, CommandLineOptionsSchema as r, MEDICAL_PLUGINS as rt, ResultFailureReason as s, PII_PLUGINS as st, AssertionOrSetSchema as t, LLAMA_GUARD_ENABLED_CATEGORIES as tt, UnifiedConfigSchema as u, REDTEAM_PROVIDER_HARM_PLUGINS as ut, ProviderOptionsSchema as v, StrategyConfigSchema as w, PartialGenerationError as x, ProvidersSchema as y, isMultiTurnStrategy as z };
-//# sourceMappingURL=types-CMDs7Kzu.js.map
+//# sourceMappingURL=types-t52w-XsS.js.map

package/dist/src/{util-BECSJ0T_.js → util-BSh4a_Q8.js}

@@ -1,11 +1,11 @@
 #!/usr/bin/env node
 import { O as cliState_default, b as getEnvBool, o as logger_default } from "./logger-BotXmWKW.js";
-import { O as TERMINAL_MAX_WIDTH, P as VERSION, n as fetchWithProxy } from "./fetch-BuP-gm0E.js";
+import { O as TERMINAL_MAX_WIDTH, P as VERSION, n as fetchWithProxy } from "./fetch-SRsE6Ctl.js";
 import { t as invariant } from "./invariant-DT20jrBd.js";
-import { m as isProviderOptions, o as OutputFileExtension, p as isApiProvider, s as ResultFailureReason } from "./types-CMDs7Kzu.js";
-import { i as isJavascriptFile, t as JAVASCRIPT_EXTENSIONS } from "./fileExtensions-CzcPkiKZ.js";
-import { o as safeResolve, r as importModule, t as getDirectory } from "./esm-79BpbQzg.js";
-import { r as runPython } from "./pythonUtils-CUPlf1H0.js";
+import { m as isProviderOptions, o as OutputFileExtension, p as isApiProvider, s as ResultFailureReason } from "./types-t52w-XsS.js";
+import { i as isJavascriptFile, t as JAVASCRIPT_EXTENSIONS } from "./fileExtensions-BpuMmaFL.js";
+import { o as safeResolve, r as importModule, t as getDirectory } from "./esm-CYhseqj4.js";
+import { r as runPython } from "./pythonUtils-r1uBuA0n.js";
 import dotenv from "dotenv";
 import * as fs$2 from "fs";
 import * as path$1 from "path";
@@ -1379,4 +1379,4 @@ function printBorder() {
 
 //#endregion
 export { filterRuntimeVars as A, extractVariablesFromTemplates as C, setupEnv as D, parseFileUrl as E, isAnthropicProvider as F, isGoogleProvider as I, isOpenAiProvider as L, resultIsForTestCase as M, doesProviderRefMatch as N, deduplicateTestCases as O, getProviderDescription as P, isProviderAllowed as R, extractVariablesFromTemplate as S, loadFunction as T, parsePathOrGlob as _, ComparisonEvalNotFoundError as a, renderEnvOnlyInObject as b, mergeComparisonTables as c, getResolvedRelativePath as d, maybeLoadConfigFromExternalFile as f, maybeLoadToolsFromExternalFile as g, maybeLoadResponseFormatFromExternalFile as h, writeOutput as i, getTestCaseDeduplicationKey as j, extractRuntimeVars as k, fetchCsvFromGoogleSheet as l, maybeLoadFromExternalFileWithVars as m, createOutputMetadata as n, evalTableToJson as o, maybeLoadFromExternalFile as p, writeMultipleOutputs as r, generateEvalCsv as s, printBorder as t, getNunjucksEngineForFilePath as u, readFilters as v, getNunjucksEngine as w, renderVarsInObject as x, readOutput as y };
-//# sourceMappingURL=util-BECSJ0T_.js.map
+//# sourceMappingURL=util-BSh4a_Q8.js.map

package/dist/src/{util-D8z5QOy8.cjs → util-C08Kns6-.cjs}

@@ -1,6 +1,6 @@
 const require_logger = require('./logger-BdZ-IqBc.cjs');
-const require_util = require('./util-Dab5xp2E.cjs');
-const require_fetch = require('./fetch-Dy2YML9G.cjs');
+const require_util = require('./util-DUYOvxAy.cjs');
+const require_fetch = require('./fetch-ZMn_oemb.cjs');
 let openai = require("openai");
 openai = require_logger.__toESM(openai);
 
@@ -623,4 +623,4 @@ Object.defineProperty(exports, 'validateFunctionCall', {
     return validateFunctionCall;
   }
 });
-//# sourceMappingURL=util-D8z5QOy8.cjs.map
+//# sourceMappingURL=util-C08Kns6-.cjs.map

package/dist/src/{util-CttPoePn.js → util-CUEt0Vum.js}

@@ -1,7 +1,7 @@
 #!/usr/bin/env node
 import { g as safeJsonStringify, p as getAjv } from "./logger-BotXmWKW.js";
-import { f as calculateCost } from "./fetch-BuP-gm0E.js";
-import { m as maybeLoadFromExternalFileWithVars } from "./util-BECSJ0T_.js";
+import { f as calculateCost } from "./fetch-SRsE6Ctl.js";
+import { m as maybeLoadFromExternalFileWithVars } from "./util-BSh4a_Q8.js";
 import OpenAI from "openai";
 
 //#region src/providers/openai/util.ts
@@ -570,4 +570,4 @@ function formatOpenAiError(data) {
 
 //#endregion
 export { calculateOpenAICost as a, getTokenUsage as c, OPENAI_TRANSCRIPTION_MODELS as i, validateFunctionCall as l, OPENAI_COMPLETION_MODELS as n, failApiCall as o, OPENAI_REALTIME_MODELS as r, formatOpenAiError as s, OPENAI_CHAT_MODELS as t };
-//# sourceMappingURL=util-CttPoePn.js.map
+//# sourceMappingURL=util-CUEt0Vum.js.map

package/dist/src/{util-CfPdWQJX.js → util-Cl0zfT3V.js}

@@ -1,6 +1,6 @@
 import { d as getAjv, m as safeJsonStringify } from "./logger-w8Ozp0Td.js";
-import { l as maybeLoadFromExternalFileWithVars } from "./util-CY_XguB9.js";
-import { g as calculateCost } from "./fetch-zJuFPxN1.js";
+import { l as maybeLoadFromExternalFileWithVars } from "./util-DiCePfDu.js";
+import { g as calculateCost } from "./fetch-18MuNu9i.js";
 import OpenAI from "openai";
 
 //#region src/providers/openai/util.ts
@@ -569,4 +569,4 @@ function formatOpenAiError(data) {
 
 //#endregion
 export { calculateOpenAICost as a, getTokenUsage as c, OPENAI_TRANSCRIPTION_MODELS as i, validateFunctionCall as l, OPENAI_COMPLETION_MODELS as n, failApiCall as o, OPENAI_REALTIME_MODELS as r, formatOpenAiError as s, OPENAI_CHAT_MODELS as t };
-//# sourceMappingURL=util-CfPdWQJX.js.map
+//# sourceMappingURL=util-Cl0zfT3V.js.map

package/dist/src/{util-Dab5xp2E.cjs → util-DUYOvxAy.cjs}

@@ -3,8 +3,8 @@ const require_invariant = require('./invariant-CKcJAQ6M.cjs');
 const require_esm = require('./esm-BQkx5roy.cjs');
 const require_pythonUtils = require('./pythonUtils-DZ6EbdY4.cjs');
 const require_fileExtensions = require('./fileExtensions-DkJYkWUy.cjs');
-const require_types = require('./types-Brw5kggg.cjs');
-const require_fetch = require('./fetch-Dy2YML9G.cjs');
+const require_types = require('./types-BIfttHrT.cjs');
+const require_fetch = require('./fetch-ZMn_oemb.cjs');
 let fs = require("fs");
 fs = require_logger.__toESM(fs);
 let path = require("path");
@@ -1458,4 +1458,4 @@ Object.defineProperty(exports, 'writeOutput', {
     return writeOutput;
   }
 });
-//# sourceMappingURL=util-Dab5xp2E.cjs.map
+//# sourceMappingURL=util-DUYOvxAy.cjs.map

package/dist/src/{util-CY_XguB9.js → util-DiCePfDu.js}

@@ -3,8 +3,8 @@ import { t as invariant } from "./invariant-1pAf2CD1.js";
 import { o as safeResolve, r as importModule, t as getDirectory } from "./esm-rDtG_2rg.js";
 import { r as runPython } from "./pythonUtils-vMlk9Qp5.js";
 import { i as isJavascriptFile, t as JAVASCRIPT_EXTENSIONS } from "./fileExtensions-ePDqouxn.js";
-import { A as isApiProvider, g as ResultFailureReason, h as OutputFileExtension, j as isProviderOptions } from "./types-CslP2TyG.js";
-import { L as VERSION, M as TERMINAL_MAX_WIDTH, t as fetchWithProxy } from "./fetch-zJuFPxN1.js";
+import { A as isApiProvider, g as ResultFailureReason, h as OutputFileExtension, j as isProviderOptions } from "./types-DMVjYLpx.js";
+import { L as VERSION, M as TERMINAL_MAX_WIDTH, t as fetchWithProxy } from "./fetch-18MuNu9i.js";
 import * as fs$1 from "fs";
 import * as path$1 from "path";
 import path from "path";
@@ -1253,4 +1253,4 @@ function printBorder() {
 
 //#endregion
 export { isAnthropicProvider as A, deduplicateTestCases as C, resultIsForTestCase as D, getTestCaseDeduplicationKey as E, isOpenAiProvider as M, isProviderAllowed as N, doesProviderRefMatch as O, setupEnv as S, filterRuntimeVars as T, extractVariablesFromTemplate as _, getNunjucksEngineForFilePath as a, loadFunction as b, maybeLoadFromExternalFile as c, maybeLoadToolsFromExternalFile as d, parsePathOrGlob as f, renderVarsInObject as g, renderEnvOnlyInObject as h, fetchCsvFromGoogleSheet as i, isGoogleProvider as j, getProviderDescription as k, maybeLoadFromExternalFileWithVars as l, readOutput as m, writeMultipleOutputs as n, getResolvedRelativePath as o, readFilters as p, writeOutput as r, maybeLoadConfigFromExternalFile as s, printBorder as t, maybeLoadResponseFormatFromExternalFile as u, extractVariablesFromTemplates as v, extractRuntimeVars as w, parseFileUrl as x, getNunjucksEngine as y };
-//# sourceMappingURL=util-CY_XguB9.js.map
+//# sourceMappingURL=util-DiCePfDu.js.map

package/dist/src/{util-CBAXTxfy.cjs → util-DkFTvieG.cjs}

@@ -1,5 +1,5 @@
-const require_fetch = require('./fetch-Dy2YML9G.cjs');
-const require_transform = require('./transform-CCSNqTI-.cjs');
+const require_fetch = require('./fetch-ZMn_oemb.cjs');
+const require_transform = require('./transform-ivxEY4f7.cjs');
 
 //#region src/providers/anthropic/util.ts
 const ANTHROPIC_MODELS = [
@@ -326,4 +326,4 @@ Object.defineProperty(exports, 'processAnthropicTools', {
     return processAnthropicTools;
   }
 });
-//# sourceMappingURL=util-CBAXTxfy.cjs.map
+//# sourceMappingURL=util-DkFTvieG.cjs.map

package/dist/src/{util-Dft1RxsL.js → util-mJ58qbbw.js}

@@ -1,5 +1,5 @@
-import { g as calculateCost } from "./fetch-zJuFPxN1.js";
-import { b as parseDataUrl } from "./transform-DosdxVdh.js";
+import { g as calculateCost } from "./fetch-18MuNu9i.js";
+import { x as parseDataUrl } from "./transform-DMaxQwDx.js";
 
 //#region src/providers/anthropic/util.ts
 const ANTHROPIC_MODELS = [
@@ -291,4 +291,4 @@ function transformWebSearchTool(config) {
 
 //#endregion
 export { parseMessages as a, outputFromMessage as i, calculateAnthropicCost as n, processAnthropicTools as o, getTokenUsage as r, ANTHROPIC_MODELS as t };
-//# sourceMappingURL=util-Dft1RxsL.js.map
+//# sourceMappingURL=util-mJ58qbbw.js.map

package/dist/src/{util-Ba18wQ8e.js → util-vjscpUzy.js}

@@ -1,6 +1,6 @@
 #!/usr/bin/env node
-import { f as calculateCost } from "./fetch-BuP-gm0E.js";
-import { b as parseDataUrl } from "./transform-gpzlPrwG.js";
+import { f as calculateCost } from "./fetch-SRsE6Ctl.js";
+import { x as parseDataUrl } from "./transform-CoP2bJ7P.js";
 
 //#region src/providers/anthropic/util.ts
 const ANTHROPIC_MODELS = [
@@ -292,4 +292,4 @@ function transformWebSearchTool(config) {
 
 //#endregion
 export { parseMessages as a, outputFromMessage as i, calculateAnthropicCost as n, processAnthropicTools as o, getTokenUsage as r, ANTHROPIC_MODELS as t };
-//# sourceMappingURL=util-Ba18wQ8e.js.map
+//# sourceMappingURL=util-vjscpUzy.js.map

package/dist/src/{utils-BTeLHwiH.js → utils-CVzb4YiI.js}

@@ -1,5 +1,5 @@
 import { b as getEnvString } from "./logger-w8Ozp0Td.js";
-import { a as sha256 } from "./extractor-CuUWqwrE.js";
+import { a as sha256 } from "./extractor-Hs7la_19.js";
 
 //#region src/prompts/constants.ts
 const PROMPT_DELIMITER = getEnvString("PROMPTFOO_PROMPT_SEPARATOR") || "---";
@@ -94,4 +94,4 @@ function hashPrompt(prompt) {
 
 //#endregion
 export { PROMPT_DELIMITER as a, generateIdFromPrompt as i, maybeFilePath as n, normalizeInput as r, hashPrompt as t };
-//# sourceMappingURL=utils-BTeLHwiH.js.map
+//# sourceMappingURL=utils-CVzb4YiI.js.map

package/dist/src/{utils-Db5KwiXp.cjs → utils-DFaZa6Rf.cjs}

@@ -1,5 +1,5 @@
 const require_logger = require('./logger-BdZ-IqBc.cjs');
-const require_extractor = require('./extractor-BPBlLyOn.cjs');
+const require_extractor = require('./extractor-DbhlYEeo.cjs');
 
 //#region src/prompts/constants.ts
 const PROMPT_DELIMITER = require_logger.getEnvString("PROMPTFOO_PROMPT_SEPARATOR") || "---";
@@ -123,4 +123,4 @@ Object.defineProperty(exports, 'normalizeInput', {
     return normalizeInput;
   }
 });
-//# sourceMappingURL=utils-Db5KwiXp.cjs.map
+//# sourceMappingURL=utils-DFaZa6Rf.cjs.map

package/dist/src/{utils-ICP7kLcY.js → utils-JaY9veb5.js}

@@ -1,6 +1,6 @@
 #!/usr/bin/env node
 import { C as getEnvString } from "./logger-BotXmWKW.js";
-import { a as sha256 } from "./extractor-BFoThUOa.js";
+import { a as sha256 } from "./extractor-LSYjrhK0.js";
 
 //#region src/prompts/constants.ts
 const PROMPT_DELIMITER = getEnvString("PROMPTFOO_PROMPT_SEPARATOR") || "---";
@@ -95,4 +95,4 @@ function hashPrompt(prompt) {
 
 //#endregion
 export { PROMPT_DELIMITER as a, generateIdFromPrompt as i, maybeFilePath as n, normalizeInput as r, hashPrompt as t };
-//# sourceMappingURL=utils-ICP7kLcY.js.map
+//# sourceMappingURL=utils-JaY9veb5.js.map