@agentv/core 3.7.0 → 3.9.0

package/dist/index.d.cts

@@ -23,8 +23,6 @@ interface ProviderStreamCallbacks {
 interface ProviderRequest {
     readonly question: string;
     readonly systemPrompt?: string;
-    readonly guidelines?: string;
-    readonly guideline_patterns?: readonly string[];
     readonly chatPrompt?: ChatPrompt;
     readonly inputFiles?: readonly string[];
     readonly evalCaseId?: string;
@@ -1016,15 +1014,13 @@ type EvaluatorConfig = CodeEvaluatorConfig | LlmGraderEvaluatorConfig | Composit
  */
 interface EvalTest {
     readonly id: string;
-    readonly dataset?: string;
+    readonly eval_set?: string;
     readonly conversation_id?: string;
     readonly question: string;
     readonly input: readonly TestMessage[];
     readonly input_segments: readonly JsonObject[];
     readonly expected_output: readonly JsonObject[];
     readonly reference_answer?: string;
-    readonly guideline_paths: readonly string[];
-    readonly guideline_patterns?: readonly string[];
     readonly file_paths: readonly string[];
     readonly criteria: string;
     readonly evaluator?: EvaluatorKind;
@@ -1128,11 +1124,10 @@ type FailOnError = boolean;
 interface EvaluationResult {
     readonly timestamp: string;
     readonly testId: string;
-    readonly dataset?: string;
+    readonly eval_set?: string;
     readonly conversationId?: string;
     readonly score: number;
     readonly assertions: readonly AssertionEntry[];
-    readonly outputText: string;
     readonly target: string;
     /** Token usage metrics from provider (optional) */
     readonly tokenUsage?: TokenUsage;
@@ -1155,10 +1150,10 @@ interface EvaluationResult {
     readonly trace?: TraceSummary;
     /** Path to the temporary workspace directory (included on failure for debugging) */
     readonly workspacePath?: string;
-    /** Input messages or prompt string sent to the agent */
-    readonly input?: readonly Message[] | string;
-    /** Full output messages from agent execution (only included when --trace flag is set) */
-    readonly output?: readonly Message[];
+    /** Input messages sent to the agent. Always Message[] for consistent shape with output. */
+    readonly input?: readonly Message[];
+    /** Output messages from agent execution. Always present — at minimum contains the final assistant message. */
+    readonly output: readonly Message[];
     /** Captured output from workspace before_all script */
     readonly beforeAllOutput?: string;
     /** Captured output from workspace before_each script */
@@ -1200,7 +1195,7 @@ interface EvaluatorResult {
     readonly verdict?: EvaluationVerdict;
     readonly assertions: readonly AssertionEntry[];
     readonly rawRequest?: JsonObject;
-    readonly evaluatorProviderRequest?: JsonObject;
+    readonly input?: JsonObject;
     readonly scores?: readonly EvaluatorResult[];
     /** Optional structured details from code graders (e.g., TP/TN/FP/FN counts). */
     readonly details?: JsonObject;
@@ -1262,7 +1257,6 @@ type ExecutionDefaults = {
 };
 type AgentVConfig$1 = {
     readonly required_version?: string;
-    readonly guideline_patterns?: readonly string[];
     readonly eval_patterns?: readonly string[];
     readonly execution?: ExecutionDefaults;
 };
@@ -1271,10 +1265,6 @@ type AgentVConfig$1 = {
  * Searches from eval file directory up to repo root.
  */
 declare function loadConfig(evalFilePath: string, repoRoot: string): Promise<AgentVConfig$1 | null>;
-/**
- * Determine whether a path references guideline content (instructions or prompts).
- */
-declare function isGuidelineFile(filePath: string, patterns?: readonly string[]): boolean;
 /**
  * Extract target name from parsed eval suite (checks execution.target then falls back to root-level target).
  */
@@ -1321,16 +1311,15 @@ declare function extractFailOnError(suite: JsonObject): FailOnError | undefined;
 type FormattingMode = 'agent' | 'lm';
 
 /**
- * Build prompt inputs by consolidating user request context and guideline content.
+ * Build prompt inputs by consolidating user request context.
  */
 interface PromptInputs {
     readonly question: string;
-    readonly guidelines: string;
     readonly chatPrompt?: ChatPrompt;
     readonly systemMessage?: string;
 }
 /**
- * Build prompt inputs by consolidating user request context and guideline content.
+ * Build prompt inputs by consolidating user request context.
  *
  * @param testCase - The evaluation test case
  * @param mode - Formatting mode: 'agent' for file references, 'lm' for embedded content (default: 'lm')
@@ -2019,7 +2008,6 @@ interface EvaluationContext {
     readonly attempt: number;
     readonly promptInputs: {
         readonly question: string;
-        readonly guidelines: string;
         readonly systemMessage?: string;
         readonly chatPrompt?: ChatPrompt;
     };
@@ -3500,7 +3488,7 @@ declare class OtelStreamingObserver {
     private rootCtx;
     constructor(tracer: Tracer, api: OtelApi, captureContent: boolean, parentCtx?: any | undefined);
     /** Create root eval span immediately (visible in backend right away) */
-    startEvalCase(testId: string, target: string, dataset?: string): void;
+    startEvalCase(testId: string, target: string, evalSet?: string): void;
     /** Create and immediately export a tool span */
     onToolCall(name: string, input: unknown, output: unknown, _durationMs: number, toolCallId?: string): void;
     /** Create and immediately export an LLM span */
@@ -3613,4 +3601,4 @@ type AgentKernel = {
 };
 declare function createAgentKernel(): AgentKernel;
 
-export { type AcquireWorkspaceOptions, type AgentKernel, type AgentVConfig$1 as AgentVConfig, type AgentVResolvedConfig, type AgentVConfig as AgentVTsConfig, type AnthropicResolvedConfig, type ArgsMatchMode, type AssertContext, type AssertEntry, type AssertFn, type AssertResult, type AssertionEntry, type AssertionResult, type AssistantTestMessage, type AzureResolvedConfig, type CacheConfig, type ChildEvaluatorResult, type ClaudeResolvedConfig, type CliResolvedConfig, CodeEvaluator, type CodeEvaluatorConfig, type CodeEvaluatorOptions, type CompositeAggregatorConfig, CompositeEvaluator, type CompositeEvaluatorConfig, type CompositeEvaluatorOptions, type ConfidenceIntervalAggregation, type ContainsAllEvaluatorConfig, type ContainsAnyEvaluatorConfig, type ContainsEvaluatorConfig, type CopilotCliResolvedConfig, type CopilotSdkResolvedConfig, CostEvaluator, type CostEvaluatorConfig, type CostEvaluatorOptions, DEFAULT_EVALUATOR_TEMPLATE, DEFAULT_EVAL_PATTERNS, DEFAULT_EXPLORATION_TOOLS, DeterministicAssertionEvaluator, type EndsWithEvaluatorConfig, type EnsureSubagentsOptions, type EnsureSubagentsResult, type EnvLookup, type EqualsEvaluatorConfig, type EvalAssertionInput, type EvalCase, type EvalConfig, type EvalMetadata, type EvalRunResult, type EvalSuiteResult, type EvalSummary, type EvalTest, type EvalTestInput, type EvalsJsonCase, type EvalsJsonFile, type EvaluationCache, type EvaluationContext, type EvaluationResult, type EvaluationScore, type EvaluationVerdict, type Evaluator, type EvaluatorConfig, type EvaluatorDispatchContext, type EvaluatorFactory, type EvaluatorFactoryFn, type EvaluatorKind, EvaluatorRegistry, type EvaluatorResult, type ExecutionDefaults, type ExecutionError, type ExecutionMetrics, ExecutionMetricsEvaluator, type ExecutionMetricsEvaluatorConfig, type ExecutionMetricsEvaluatorOptions, type ExecutionStatus, type FailOnError, type FailureStage, FieldAccuracyEvaluator, type FieldAccuracyEvaluatorConfig, type FieldAccuracyEvaluatorOptions, type FieldAggregationType, type FieldConfig, type FieldMatchType, type GeminiResolvedConfig, type GenerateRubricsOptions, type IcontainsAllEvaluatorConfig, type IcontainsAnyEvaluatorConfig, type IcontainsEvaluatorConfig, type InlineAssertEvaluatorConfig, type IsJsonEvaluatorConfig, type JsonObject, type JsonPrimitive, type JsonValue, LatencyEvaluator, type LatencyEvaluatorConfig, type LatencyEvaluatorOptions, LlmGraderEvaluator, type LlmGraderEvaluatorConfig, type LlmGraderEvaluatorOptions, type LlmGraderPromptAssembly, LlmGraderEvaluator as LlmJudgeEvaluator, type LlmJudgeEvaluatorConfig, type LlmGraderEvaluatorOptions as LlmJudgeEvaluatorOptions, type LlmGraderPromptAssembly as LlmJudgePromptAssembly, type LocalPathValidationError, type MeanAggregation, type Message, type MockResolvedConfig, OTEL_BACKEND_PRESETS, type OpenAIResolvedConfig, type OpenRouterResolvedConfig, type OtelBackendPreset, type OtelExportOptions, OtelStreamingObserver, OtelTraceExporter, OtlpJsonFileExporter, type OutputMessage, type PassAtKAggregation, type PiAgentSdkResolvedConfig, type PiCodingAgentResolvedConfig, type PoolSlot, type ProgressEvent, type PromptInputs, type PromptScriptConfig, type Provider, type ProviderFactoryFn, type ProviderKind, ProviderRegistry, type ProviderRequest, type ProviderResponse, type ProviderStreamCallbacks, type ProviderTokenUsage, type RegexEvaluatorConfig, type RepoCheckout, type RepoClone, type RepoConfig, RepoManager, type RepoSource, type ResolvedTarget, type ResolvedWorkspaceTemplate, ResponseCache, type RubricItem, type RubricsEvaluatorConfig, type RunEvalCaseOptions, type RunEvaluationOptions, type ScoreRange, type ScriptExecutionContext, SimpleTraceFileExporter, SkillTriggerEvaluator, type SkillTriggerEvaluatorConfig, type StartsWithEvaluatorConfig, type SystemTestMessage, TEST_MESSAGE_ROLES, type TargetAccessConfig, type TargetDefinition, TemplateNotDirectoryError, TemplateNotFoundError, type TestMessage, type TestMessageContent, type TestMessageRole, type TokenUsage, TokenUsageEvaluator, type TokenUsageEvaluatorConfig, type TokenUsageEvaluatorOptions, type ToolCall, type ToolTestMessage, ToolTrajectoryEvaluator, type ToolTrajectoryEvaluatorConfig, type ToolTrajectoryEvaluatorOptions, type ToolTrajectoryExpectedItem, type TraceComputeResult, type TraceSummary, type TranspileResult, type TrialAggregation, type TrialResult, type TrialStrategy, type TrialsConfig, type UserTestMessage, type VSCodeResolvedConfig, type WorkspaceConfig, WorkspaceCreationError, type WorkspaceHookConfig, type WorkspaceHooksConfig, WorkspacePoolManager, type WorkspaceScriptConfig, assembleLlmGraderPrompt, assembleLlmGraderPrompt as assembleLlmJudgePrompt, avgToolDurationMs, buildDirectoryChain, buildOutputSchema, buildPromptInputs, buildRubricOutputSchema, buildScoreRangeOutputSchema, buildSearchRoots, calculateRubricScore, captureFileChanges, clampScore, cleanupEvalWorkspaces, cleanupWorkspace, computeTraceSummary, computeWorkspaceFingerprint, consumeClaudeLogEntries, consumeCodexLogEntries, consumeCopilotCliLogEntries, consumeCopilotSdkLogEntries, consumePiLogEntries, createAgentKernel, createBuiltinProviderRegistry, createBuiltinRegistry, createProvider, createTempWorkspace, deepEqual, defineConfig, detectFormat, discoverAssertions, discoverGraders, discoverGraders as discoverJudges, discoverProviders, ensureVSCodeSubagents, evaluate, executeScript, executeWorkspaceScript, explorationRatio, extractCacheConfig, extractFailOnError, extractJsonBlob, extractTargetFromSuite, extractTargetsFromSuite, extractTargetsFromTestCase, extractTrialsConfig, fileExists, findGitRoot, freeformEvaluationSchema, generateRubrics, getAgentvHome, getOutputFilenames, getSubagentsRoot, getTraceStateRoot, getWorkspacePath, getWorkspacePoolRoot, getWorkspacesRoot, initializeBaseline, isAgentSkillsFormat, isEvaluatorKind, isGuidelineFile, isJsonObject, isJsonValue, isNonEmptyString, isTestMessage, isTestMessageRole, listTargetNames, loadConfig, loadEvalCaseById, loadEvalCases, loadEvalSuite, loadTestById, loadTestSuite, loadTests, loadTsConfig, mergeExecutionMetrics, negateScore, normalizeLineEndings, parseAgentSkillsEvals, parseJsonFromText, parseJsonSafe, readJsonFile, readTargetDefinitions, readTestSuiteMetadata, readTextFile, resolveAndCreateProvider, resolveFileReference, resolveTargetDefinition, resolveWorkspaceTemplate, rubricEvaluationSchema, runContainsAllAssertion, runContainsAnyAssertion, runContainsAssertion, runEndsWithAssertion, runEqualsAssertion, runEvalCase, runEvaluation, runIcontainsAllAssertion, runIcontainsAnyAssertion, runIcontainsAssertion, runIsJsonAssertion, runRegexAssertion, runStartsWithAssertion, scoreToVerdict, shouldEnableCache, shouldSkipCacheForTemperature, subscribeToClaudeLogEntries, subscribeToCodexLogEntries, subscribeToCopilotCliLogEntries, subscribeToCopilotSdkLogEntries, subscribeToPiLogEntries, substituteVariables, toCamelCaseDeep, toSnakeCaseDeep, tokensPerTool, transpileEvalYaml, transpileEvalYamlFile, trimBaselineResult };
+export { type AcquireWorkspaceOptions, type AgentKernel, type AgentVConfig$1 as AgentVConfig, type AgentVResolvedConfig, type AgentVConfig as AgentVTsConfig, type AnthropicResolvedConfig, type ArgsMatchMode, type AssertContext, type AssertEntry, type AssertFn, type AssertResult, type AssertionEntry, type AssertionResult, type AssistantTestMessage, type AzureResolvedConfig, type CacheConfig, type ChildEvaluatorResult, type ClaudeResolvedConfig, type CliResolvedConfig, CodeEvaluator, type CodeEvaluatorConfig, type CodeEvaluatorOptions, type CompositeAggregatorConfig, CompositeEvaluator, type CompositeEvaluatorConfig, type CompositeEvaluatorOptions, type ConfidenceIntervalAggregation, type ContainsAllEvaluatorConfig, type ContainsAnyEvaluatorConfig, type ContainsEvaluatorConfig, type CopilotCliResolvedConfig, type CopilotSdkResolvedConfig, CostEvaluator, type CostEvaluatorConfig, type CostEvaluatorOptions, DEFAULT_EVALUATOR_TEMPLATE, DEFAULT_EVAL_PATTERNS, DEFAULT_EXPLORATION_TOOLS, DeterministicAssertionEvaluator, type EndsWithEvaluatorConfig, type EnsureSubagentsOptions, type EnsureSubagentsResult, type EnvLookup, type EqualsEvaluatorConfig, type EvalAssertionInput, type EvalCase, type EvalConfig, type EvalMetadata, type EvalRunResult, type EvalSuiteResult, type EvalSummary, type EvalTest, type EvalTestInput, type EvalsJsonCase, type EvalsJsonFile, type EvaluationCache, type EvaluationContext, type EvaluationResult, type EvaluationScore, type EvaluationVerdict, type Evaluator, type EvaluatorConfig, type EvaluatorDispatchContext, type EvaluatorFactory, type EvaluatorFactoryFn, type EvaluatorKind, EvaluatorRegistry, type EvaluatorResult, type ExecutionDefaults, type ExecutionError, type ExecutionMetrics, ExecutionMetricsEvaluator, type ExecutionMetricsEvaluatorConfig, type ExecutionMetricsEvaluatorOptions, type ExecutionStatus, type FailOnError, type FailureStage, FieldAccuracyEvaluator, type FieldAccuracyEvaluatorConfig, type FieldAccuracyEvaluatorOptions, type FieldAggregationType, type FieldConfig, type FieldMatchType, type GeminiResolvedConfig, type GenerateRubricsOptions, type IcontainsAllEvaluatorConfig, type IcontainsAnyEvaluatorConfig, type IcontainsEvaluatorConfig, type InlineAssertEvaluatorConfig, type IsJsonEvaluatorConfig, type JsonObject, type JsonPrimitive, type JsonValue, LatencyEvaluator, type LatencyEvaluatorConfig, type LatencyEvaluatorOptions, LlmGraderEvaluator, type LlmGraderEvaluatorConfig, type LlmGraderEvaluatorOptions, type LlmGraderPromptAssembly, LlmGraderEvaluator as LlmJudgeEvaluator, type LlmJudgeEvaluatorConfig, type LlmGraderEvaluatorOptions as LlmJudgeEvaluatorOptions, type LlmGraderPromptAssembly as LlmJudgePromptAssembly, type LocalPathValidationError, type MeanAggregation, type Message, type MockResolvedConfig, OTEL_BACKEND_PRESETS, type OpenAIResolvedConfig, type OpenRouterResolvedConfig, type OtelBackendPreset, type OtelExportOptions, OtelStreamingObserver, OtelTraceExporter, OtlpJsonFileExporter, type OutputMessage, type PassAtKAggregation, type PiAgentSdkResolvedConfig, type PiCodingAgentResolvedConfig, type PoolSlot, type ProgressEvent, type PromptInputs, type PromptScriptConfig, type Provider, type ProviderFactoryFn, type ProviderKind, ProviderRegistry, type ProviderRequest, type ProviderResponse, type ProviderStreamCallbacks, type ProviderTokenUsage, type RegexEvaluatorConfig, type RepoCheckout, type RepoClone, type RepoConfig, RepoManager, type RepoSource, type ResolvedTarget, type ResolvedWorkspaceTemplate, ResponseCache, type RubricItem, type RubricsEvaluatorConfig, type RunEvalCaseOptions, type RunEvaluationOptions, type ScoreRange, type ScriptExecutionContext, SimpleTraceFileExporter, SkillTriggerEvaluator, type SkillTriggerEvaluatorConfig, type StartsWithEvaluatorConfig, type SystemTestMessage, TEST_MESSAGE_ROLES, type TargetAccessConfig, type TargetDefinition, TemplateNotDirectoryError, TemplateNotFoundError, type TestMessage, type TestMessageContent, type TestMessageRole, type TokenUsage, TokenUsageEvaluator, type TokenUsageEvaluatorConfig, type TokenUsageEvaluatorOptions, type ToolCall, type ToolTestMessage, ToolTrajectoryEvaluator, type ToolTrajectoryEvaluatorConfig, type ToolTrajectoryEvaluatorOptions, type ToolTrajectoryExpectedItem, type TraceComputeResult, type TraceSummary, type TranspileResult, type TrialAggregation, type TrialResult, type TrialStrategy, type TrialsConfig, type UserTestMessage, type VSCodeResolvedConfig, type WorkspaceConfig, WorkspaceCreationError, type WorkspaceHookConfig, type WorkspaceHooksConfig, WorkspacePoolManager, type WorkspaceScriptConfig, assembleLlmGraderPrompt, assembleLlmGraderPrompt as assembleLlmJudgePrompt, avgToolDurationMs, buildDirectoryChain, buildOutputSchema, buildPromptInputs, buildRubricOutputSchema, buildScoreRangeOutputSchema, buildSearchRoots, calculateRubricScore, captureFileChanges, clampScore, cleanupEvalWorkspaces, cleanupWorkspace, computeTraceSummary, computeWorkspaceFingerprint, consumeClaudeLogEntries, consumeCodexLogEntries, consumeCopilotCliLogEntries, consumeCopilotSdkLogEntries, consumePiLogEntries, createAgentKernel, createBuiltinProviderRegistry, createBuiltinRegistry, createProvider, createTempWorkspace, deepEqual, defineConfig, detectFormat, discoverAssertions, discoverGraders, discoverGraders as discoverJudges, discoverProviders, ensureVSCodeSubagents, evaluate, executeScript, executeWorkspaceScript, explorationRatio, extractCacheConfig, extractFailOnError, extractJsonBlob, extractTargetFromSuite, extractTargetsFromSuite, extractTargetsFromTestCase, extractTrialsConfig, fileExists, findGitRoot, freeformEvaluationSchema, generateRubrics, getAgentvHome, getOutputFilenames, getSubagentsRoot, getTraceStateRoot, getWorkspacePath, getWorkspacePoolRoot, getWorkspacesRoot, initializeBaseline, isAgentSkillsFormat, isEvaluatorKind, isJsonObject, isJsonValue, isNonEmptyString, isTestMessage, isTestMessageRole, listTargetNames, loadConfig, loadEvalCaseById, loadEvalCases, loadEvalSuite, loadTestById, loadTestSuite, loadTests, loadTsConfig, mergeExecutionMetrics, negateScore, normalizeLineEndings, parseAgentSkillsEvals, parseJsonFromText, parseJsonSafe, readJsonFile, readTargetDefinitions, readTestSuiteMetadata, readTextFile, resolveAndCreateProvider, resolveFileReference, resolveTargetDefinition, resolveWorkspaceTemplate, rubricEvaluationSchema, runContainsAllAssertion, runContainsAnyAssertion, runContainsAssertion, runEndsWithAssertion, runEqualsAssertion, runEvalCase, runEvaluation, runIcontainsAllAssertion, runIcontainsAnyAssertion, runIcontainsAssertion, runIsJsonAssertion, runRegexAssertion, runStartsWithAssertion, scoreToVerdict, shouldEnableCache, shouldSkipCacheForTemperature, subscribeToClaudeLogEntries, subscribeToCodexLogEntries, subscribeToCopilotCliLogEntries, subscribeToCopilotSdkLogEntries, subscribeToPiLogEntries, substituteVariables, toCamelCaseDeep, toSnakeCaseDeep, tokensPerTool, transpileEvalYaml, transpileEvalYamlFile, trimBaselineResult };

package/dist/index.d.ts

@@ -23,8 +23,6 @@ interface ProviderStreamCallbacks {
 interface ProviderRequest {
     readonly question: string;
     readonly systemPrompt?: string;
-    readonly guidelines?: string;
-    readonly guideline_patterns?: readonly string[];
     readonly chatPrompt?: ChatPrompt;
     readonly inputFiles?: readonly string[];
     readonly evalCaseId?: string;
@@ -1016,15 +1014,13 @@ type EvaluatorConfig = CodeEvaluatorConfig | LlmGraderEvaluatorConfig | Composit
  */
 interface EvalTest {
     readonly id: string;
-    readonly dataset?: string;
+    readonly eval_set?: string;
     readonly conversation_id?: string;
     readonly question: string;
     readonly input: readonly TestMessage[];
     readonly input_segments: readonly JsonObject[];
     readonly expected_output: readonly JsonObject[];
     readonly reference_answer?: string;
-    readonly guideline_paths: readonly string[];
-    readonly guideline_patterns?: readonly string[];
     readonly file_paths: readonly string[];
     readonly criteria: string;
     readonly evaluator?: EvaluatorKind;
@@ -1128,11 +1124,10 @@ type FailOnError = boolean;
 interface EvaluationResult {
     readonly timestamp: string;
     readonly testId: string;
-    readonly dataset?: string;
+    readonly eval_set?: string;
     readonly conversationId?: string;
     readonly score: number;
     readonly assertions: readonly AssertionEntry[];
-    readonly outputText: string;
     readonly target: string;
     /** Token usage metrics from provider (optional) */
     readonly tokenUsage?: TokenUsage;
@@ -1155,10 +1150,10 @@ interface EvaluationResult {
     readonly trace?: TraceSummary;
     /** Path to the temporary workspace directory (included on failure for debugging) */
     readonly workspacePath?: string;
-    /** Input messages or prompt string sent to the agent */
-    readonly input?: readonly Message[] | string;
-    /** Full output messages from agent execution (only included when --trace flag is set) */
-    readonly output?: readonly Message[];
+    /** Input messages sent to the agent. Always Message[] for consistent shape with output. */
+    readonly input?: readonly Message[];
+    /** Output messages from agent execution. Always present — at minimum contains the final assistant message. */
+    readonly output: readonly Message[];
     /** Captured output from workspace before_all script */
     readonly beforeAllOutput?: string;
     /** Captured output from workspace before_each script */
@@ -1200,7 +1195,7 @@ interface EvaluatorResult {
     readonly verdict?: EvaluationVerdict;
     readonly assertions: readonly AssertionEntry[];
     readonly rawRequest?: JsonObject;
-    readonly evaluatorProviderRequest?: JsonObject;
+    readonly input?: JsonObject;
     readonly scores?: readonly EvaluatorResult[];
     /** Optional structured details from code graders (e.g., TP/TN/FP/FN counts). */
     readonly details?: JsonObject;
@@ -1262,7 +1257,6 @@ type ExecutionDefaults = {
 };
 type AgentVConfig$1 = {
     readonly required_version?: string;
-    readonly guideline_patterns?: readonly string[];
     readonly eval_patterns?: readonly string[];
     readonly execution?: ExecutionDefaults;
 };
@@ -1271,10 +1265,6 @@ type AgentVConfig$1 = {
  * Searches from eval file directory up to repo root.
  */
 declare function loadConfig(evalFilePath: string, repoRoot: string): Promise<AgentVConfig$1 | null>;
-/**
- * Determine whether a path references guideline content (instructions or prompts).
- */
-declare function isGuidelineFile(filePath: string, patterns?: readonly string[]): boolean;
 /**
  * Extract target name from parsed eval suite (checks execution.target then falls back to root-level target).
  */
@@ -1321,16 +1311,15 @@ declare function extractFailOnError(suite: JsonObject): FailOnError | undefined;
 type FormattingMode = 'agent' | 'lm';
 
 /**
- * Build prompt inputs by consolidating user request context and guideline content.
+ * Build prompt inputs by consolidating user request context.
  */
 interface PromptInputs {
     readonly question: string;
-    readonly guidelines: string;
     readonly chatPrompt?: ChatPrompt;
     readonly systemMessage?: string;
 }
 /**
- * Build prompt inputs by consolidating user request context and guideline content.
+ * Build prompt inputs by consolidating user request context.
  *
  * @param testCase - The evaluation test case
  * @param mode - Formatting mode: 'agent' for file references, 'lm' for embedded content (default: 'lm')
@@ -2019,7 +2008,6 @@ interface EvaluationContext {
     readonly attempt: number;
     readonly promptInputs: {
         readonly question: string;
-        readonly guidelines: string;
         readonly systemMessage?: string;
         readonly chatPrompt?: ChatPrompt;
     };
@@ -3500,7 +3488,7 @@ declare class OtelStreamingObserver {
     private rootCtx;
     constructor(tracer: Tracer, api: OtelApi, captureContent: boolean, parentCtx?: any | undefined);
     /** Create root eval span immediately (visible in backend right away) */
-    startEvalCase(testId: string, target: string, dataset?: string): void;
+    startEvalCase(testId: string, target: string, evalSet?: string): void;
     /** Create and immediately export a tool span */
     onToolCall(name: string, input: unknown, output: unknown, _durationMs: number, toolCallId?: string): void;
     /** Create and immediately export an LLM span */
@@ -3613,4 +3601,4 @@ type AgentKernel = {
 };
 declare function createAgentKernel(): AgentKernel;
 
-export { type AcquireWorkspaceOptions, type AgentKernel, type AgentVConfig$1 as AgentVConfig, type AgentVResolvedConfig, type AgentVConfig as AgentVTsConfig, type AnthropicResolvedConfig, type ArgsMatchMode, type AssertContext, type AssertEntry, type AssertFn, type AssertResult, type AssertionEntry, type AssertionResult, type AssistantTestMessage, type AzureResolvedConfig, type CacheConfig, type ChildEvaluatorResult, type ClaudeResolvedConfig, type CliResolvedConfig, CodeEvaluator, type CodeEvaluatorConfig, type CodeEvaluatorOptions, type CompositeAggregatorConfig, CompositeEvaluator, type CompositeEvaluatorConfig, type CompositeEvaluatorOptions, type ConfidenceIntervalAggregation, type ContainsAllEvaluatorConfig, type ContainsAnyEvaluatorConfig, type ContainsEvaluatorConfig, type CopilotCliResolvedConfig, type CopilotSdkResolvedConfig, CostEvaluator, type CostEvaluatorConfig, type CostEvaluatorOptions, DEFAULT_EVALUATOR_TEMPLATE, DEFAULT_EVAL_PATTERNS, DEFAULT_EXPLORATION_TOOLS, DeterministicAssertionEvaluator, type EndsWithEvaluatorConfig, type EnsureSubagentsOptions, type EnsureSubagentsResult, type EnvLookup, type EqualsEvaluatorConfig, type EvalAssertionInput, type EvalCase, type EvalConfig, type EvalMetadata, type EvalRunResult, type EvalSuiteResult, type EvalSummary, type EvalTest, type EvalTestInput, type EvalsJsonCase, type EvalsJsonFile, type EvaluationCache, type EvaluationContext, type EvaluationResult, type EvaluationScore, type EvaluationVerdict, type Evaluator, type EvaluatorConfig, type EvaluatorDispatchContext, type EvaluatorFactory, type EvaluatorFactoryFn, type EvaluatorKind, EvaluatorRegistry, type EvaluatorResult, type ExecutionDefaults, type ExecutionError, type ExecutionMetrics, ExecutionMetricsEvaluator, type ExecutionMetricsEvaluatorConfig, type ExecutionMetricsEvaluatorOptions, type ExecutionStatus, type FailOnError, type FailureStage, FieldAccuracyEvaluator, type FieldAccuracyEvaluatorConfig, type FieldAccuracyEvaluatorOptions, type FieldAggregationType, type FieldConfig, type FieldMatchType, type GeminiResolvedConfig, type GenerateRubricsOptions, type IcontainsAllEvaluatorConfig, type IcontainsAnyEvaluatorConfig, type IcontainsEvaluatorConfig, type InlineAssertEvaluatorConfig, type IsJsonEvaluatorConfig, type JsonObject, type JsonPrimitive, type JsonValue, LatencyEvaluator, type LatencyEvaluatorConfig, type LatencyEvaluatorOptions, LlmGraderEvaluator, type LlmGraderEvaluatorConfig, type LlmGraderEvaluatorOptions, type LlmGraderPromptAssembly, LlmGraderEvaluator as LlmJudgeEvaluator, type LlmJudgeEvaluatorConfig, type LlmGraderEvaluatorOptions as LlmJudgeEvaluatorOptions, type LlmGraderPromptAssembly as LlmJudgePromptAssembly, type LocalPathValidationError, type MeanAggregation, type Message, type MockResolvedConfig, OTEL_BACKEND_PRESETS, type OpenAIResolvedConfig, type OpenRouterResolvedConfig, type OtelBackendPreset, type OtelExportOptions, OtelStreamingObserver, OtelTraceExporter, OtlpJsonFileExporter, type OutputMessage, type PassAtKAggregation, type PiAgentSdkResolvedConfig, type PiCodingAgentResolvedConfig, type PoolSlot, type ProgressEvent, type PromptInputs, type PromptScriptConfig, type Provider, type ProviderFactoryFn, type ProviderKind, ProviderRegistry, type ProviderRequest, type ProviderResponse, type ProviderStreamCallbacks, type ProviderTokenUsage, type RegexEvaluatorConfig, type RepoCheckout, type RepoClone, type RepoConfig, RepoManager, type RepoSource, type ResolvedTarget, type ResolvedWorkspaceTemplate, ResponseCache, type RubricItem, type RubricsEvaluatorConfig, type RunEvalCaseOptions, type RunEvaluationOptions, type ScoreRange, type ScriptExecutionContext, SimpleTraceFileExporter, SkillTriggerEvaluator, type SkillTriggerEvaluatorConfig, type StartsWithEvaluatorConfig, type SystemTestMessage, TEST_MESSAGE_ROLES, type TargetAccessConfig, type TargetDefinition, TemplateNotDirectoryError, TemplateNotFoundError, type TestMessage, type TestMessageContent, type TestMessageRole, type TokenUsage, TokenUsageEvaluator, type TokenUsageEvaluatorConfig, type TokenUsageEvaluatorOptions, type ToolCall, type ToolTestMessage, ToolTrajectoryEvaluator, type ToolTrajectoryEvaluatorConfig, type ToolTrajectoryEvaluatorOptions, type ToolTrajectoryExpectedItem, type TraceComputeResult, type TraceSummary, type TranspileResult, type TrialAggregation, type TrialResult, type TrialStrategy, type TrialsConfig, type UserTestMessage, type VSCodeResolvedConfig, type WorkspaceConfig, WorkspaceCreationError, type WorkspaceHookConfig, type WorkspaceHooksConfig, WorkspacePoolManager, type WorkspaceScriptConfig, assembleLlmGraderPrompt, assembleLlmGraderPrompt as assembleLlmJudgePrompt, avgToolDurationMs, buildDirectoryChain, buildOutputSchema, buildPromptInputs, buildRubricOutputSchema, buildScoreRangeOutputSchema, buildSearchRoots, calculateRubricScore, captureFileChanges, clampScore, cleanupEvalWorkspaces, cleanupWorkspace, computeTraceSummary, computeWorkspaceFingerprint, consumeClaudeLogEntries, consumeCodexLogEntries, consumeCopilotCliLogEntries, consumeCopilotSdkLogEntries, consumePiLogEntries, createAgentKernel, createBuiltinProviderRegistry, createBuiltinRegistry, createProvider, createTempWorkspace, deepEqual, defineConfig, detectFormat, discoverAssertions, discoverGraders, discoverGraders as discoverJudges, discoverProviders, ensureVSCodeSubagents, evaluate, executeScript, executeWorkspaceScript, explorationRatio, extractCacheConfig, extractFailOnError, extractJsonBlob, extractTargetFromSuite, extractTargetsFromSuite, extractTargetsFromTestCase, extractTrialsConfig, fileExists, findGitRoot, freeformEvaluationSchema, generateRubrics, getAgentvHome, getOutputFilenames, getSubagentsRoot, getTraceStateRoot, getWorkspacePath, getWorkspacePoolRoot, getWorkspacesRoot, initializeBaseline, isAgentSkillsFormat, isEvaluatorKind, isGuidelineFile, isJsonObject, isJsonValue, isNonEmptyString, isTestMessage, isTestMessageRole, listTargetNames, loadConfig, loadEvalCaseById, loadEvalCases, loadEvalSuite, loadTestById, loadTestSuite, loadTests, loadTsConfig, mergeExecutionMetrics, negateScore, normalizeLineEndings, parseAgentSkillsEvals, parseJsonFromText, parseJsonSafe, readJsonFile, readTargetDefinitions, readTestSuiteMetadata, readTextFile, resolveAndCreateProvider, resolveFileReference, resolveTargetDefinition, resolveWorkspaceTemplate, rubricEvaluationSchema, runContainsAllAssertion, runContainsAnyAssertion, runContainsAssertion, runEndsWithAssertion, runEqualsAssertion, runEvalCase, runEvaluation, runIcontainsAllAssertion, runIcontainsAnyAssertion, runIcontainsAssertion, runIsJsonAssertion, runRegexAssertion, runStartsWithAssertion, scoreToVerdict, shouldEnableCache, shouldSkipCacheForTemperature, subscribeToClaudeLogEntries, subscribeToCodexLogEntries, subscribeToCopilotCliLogEntries, subscribeToCopilotSdkLogEntries, subscribeToPiLogEntries, substituteVariables, toCamelCaseDeep, toSnakeCaseDeep, tokensPerTool, transpileEvalYaml, transpileEvalYamlFile, trimBaselineResult };
+export { type AcquireWorkspaceOptions, type AgentKernel, type AgentVConfig$1 as AgentVConfig, type AgentVResolvedConfig, type AgentVConfig as AgentVTsConfig, type AnthropicResolvedConfig, type ArgsMatchMode, type AssertContext, type AssertEntry, type AssertFn, type AssertResult, type AssertionEntry, type AssertionResult, type AssistantTestMessage, type AzureResolvedConfig, type CacheConfig, type ChildEvaluatorResult, type ClaudeResolvedConfig, type CliResolvedConfig, CodeEvaluator, type CodeEvaluatorConfig, type CodeEvaluatorOptions, type CompositeAggregatorConfig, CompositeEvaluator, type CompositeEvaluatorConfig, type CompositeEvaluatorOptions, type ConfidenceIntervalAggregation, type ContainsAllEvaluatorConfig, type ContainsAnyEvaluatorConfig, type ContainsEvaluatorConfig, type CopilotCliResolvedConfig, type CopilotSdkResolvedConfig, CostEvaluator, type CostEvaluatorConfig, type CostEvaluatorOptions, DEFAULT_EVALUATOR_TEMPLATE, DEFAULT_EVAL_PATTERNS, DEFAULT_EXPLORATION_TOOLS, DeterministicAssertionEvaluator, type EndsWithEvaluatorConfig, type EnsureSubagentsOptions, type EnsureSubagentsResult, type EnvLookup, type EqualsEvaluatorConfig, type EvalAssertionInput, type EvalCase, type EvalConfig, type EvalMetadata, type EvalRunResult, type EvalSuiteResult, type EvalSummary, type EvalTest, type EvalTestInput, type EvalsJsonCase, type EvalsJsonFile, type EvaluationCache, type EvaluationContext, type EvaluationResult, type EvaluationScore, type EvaluationVerdict, type Evaluator, type EvaluatorConfig, type EvaluatorDispatchContext, type EvaluatorFactory, type EvaluatorFactoryFn, type EvaluatorKind, EvaluatorRegistry, type EvaluatorResult, type ExecutionDefaults, type ExecutionError, type ExecutionMetrics, ExecutionMetricsEvaluator, type ExecutionMetricsEvaluatorConfig, type ExecutionMetricsEvaluatorOptions, type ExecutionStatus, type FailOnError, type FailureStage, FieldAccuracyEvaluator, type FieldAccuracyEvaluatorConfig, type FieldAccuracyEvaluatorOptions, type FieldAggregationType, type FieldConfig, type FieldMatchType, type GeminiResolvedConfig, type GenerateRubricsOptions, type IcontainsAllEvaluatorConfig, type IcontainsAnyEvaluatorConfig, type IcontainsEvaluatorConfig, type InlineAssertEvaluatorConfig, type IsJsonEvaluatorConfig, type JsonObject, type JsonPrimitive, type JsonValue, LatencyEvaluator, type LatencyEvaluatorConfig, type LatencyEvaluatorOptions, LlmGraderEvaluator, type LlmGraderEvaluatorConfig, type LlmGraderEvaluatorOptions, type LlmGraderPromptAssembly, LlmGraderEvaluator as LlmJudgeEvaluator, type LlmJudgeEvaluatorConfig, type LlmGraderEvaluatorOptions as LlmJudgeEvaluatorOptions, type LlmGraderPromptAssembly as LlmJudgePromptAssembly, type LocalPathValidationError, type MeanAggregation, type Message, type MockResolvedConfig, OTEL_BACKEND_PRESETS, type OpenAIResolvedConfig, type OpenRouterResolvedConfig, type OtelBackendPreset, type OtelExportOptions, OtelStreamingObserver, OtelTraceExporter, OtlpJsonFileExporter, type OutputMessage, type PassAtKAggregation, type PiAgentSdkResolvedConfig, type PiCodingAgentResolvedConfig, type PoolSlot, type ProgressEvent, type PromptInputs, type PromptScriptConfig, type Provider, type ProviderFactoryFn, type ProviderKind, ProviderRegistry, type ProviderRequest, type ProviderResponse, type ProviderStreamCallbacks, type ProviderTokenUsage, type RegexEvaluatorConfig, type RepoCheckout, type RepoClone, type RepoConfig, RepoManager, type RepoSource, type ResolvedTarget, type ResolvedWorkspaceTemplate, ResponseCache, type RubricItem, type RubricsEvaluatorConfig, type RunEvalCaseOptions, type RunEvaluationOptions, type ScoreRange, type ScriptExecutionContext, SimpleTraceFileExporter, SkillTriggerEvaluator, type SkillTriggerEvaluatorConfig, type StartsWithEvaluatorConfig, type SystemTestMessage, TEST_MESSAGE_ROLES, type TargetAccessConfig, type TargetDefinition, TemplateNotDirectoryError, TemplateNotFoundError, type TestMessage, type TestMessageContent, type TestMessageRole, type TokenUsage, TokenUsageEvaluator, type TokenUsageEvaluatorConfig, type TokenUsageEvaluatorOptions, type ToolCall, type ToolTestMessage, ToolTrajectoryEvaluator, type ToolTrajectoryEvaluatorConfig, type ToolTrajectoryEvaluatorOptions, type ToolTrajectoryExpectedItem, type TraceComputeResult, type TraceSummary, type TranspileResult, type TrialAggregation, type TrialResult, type TrialStrategy, type TrialsConfig, type UserTestMessage, type VSCodeResolvedConfig, type WorkspaceConfig, WorkspaceCreationError, type WorkspaceHookConfig, type WorkspaceHooksConfig, WorkspacePoolManager, type WorkspaceScriptConfig, assembleLlmGraderPrompt, assembleLlmGraderPrompt as assembleLlmJudgePrompt, avgToolDurationMs, buildDirectoryChain, buildOutputSchema, buildPromptInputs, buildRubricOutputSchema, buildScoreRangeOutputSchema, buildSearchRoots, calculateRubricScore, captureFileChanges, clampScore, cleanupEvalWorkspaces, cleanupWorkspace, computeTraceSummary, computeWorkspaceFingerprint, consumeClaudeLogEntries, consumeCodexLogEntries, consumeCopilotCliLogEntries, consumeCopilotSdkLogEntries, consumePiLogEntries, createAgentKernel, createBuiltinProviderRegistry, createBuiltinRegistry, createProvider, createTempWorkspace, deepEqual, defineConfig, detectFormat, discoverAssertions, discoverGraders, discoverGraders as discoverJudges, discoverProviders, ensureVSCodeSubagents, evaluate, executeScript, executeWorkspaceScript, explorationRatio, extractCacheConfig, extractFailOnError, extractJsonBlob, extractTargetFromSuite, extractTargetsFromSuite, extractTargetsFromTestCase, extractTrialsConfig, fileExists, findGitRoot, freeformEvaluationSchema, generateRubrics, getAgentvHome, getOutputFilenames, getSubagentsRoot, getTraceStateRoot, getWorkspacePath, getWorkspacePoolRoot, getWorkspacesRoot, initializeBaseline, isAgentSkillsFormat, isEvaluatorKind, isJsonObject, isJsonValue, isNonEmptyString, isTestMessage, isTestMessageRole, listTargetNames, loadConfig, loadEvalCaseById, loadEvalCases, loadEvalSuite, loadTestById, loadTestSuite, loadTests, loadTsConfig, mergeExecutionMetrics, negateScore, normalizeLineEndings, parseAgentSkillsEvals, parseJsonFromText, parseJsonSafe, readJsonFile, readTargetDefinitions, readTestSuiteMetadata, readTextFile, resolveAndCreateProvider, resolveFileReference, resolveTargetDefinition, resolveWorkspaceTemplate, rubricEvaluationSchema, runContainsAllAssertion, runContainsAnyAssertion, runContainsAssertion, runEndsWithAssertion, runEqualsAssertion, runEvalCase, runEvaluation, runIcontainsAllAssertion, runIcontainsAnyAssertion, runIcontainsAssertion, runIsJsonAssertion, runRegexAssertion, runStartsWithAssertion, scoreToVerdict, shouldEnableCache, shouldSkipCacheForTemperature, subscribeToClaudeLogEntries, subscribeToCodexLogEntries, subscribeToCopilotCliLogEntries, subscribeToCopilotSdkLogEntries, subscribeToPiLogEntries, substituteVariables, toCamelCaseDeep, toSnakeCaseDeep, tokensPerTool, transpileEvalYaml, transpileEvalYamlFile, trimBaselineResult };