@wix/pathgrade 1.0.30 → 1.0.32

package/dist/types.d.ts

@@ -8,10 +8,10 @@ export interface CommandResult {
     timedOut?: boolean;
     killed?: boolean;
 }
-export interface TurnCommand extends CommandResult {
+interface TurnCommand extends CommandResult {
     command: string;
 }
-export type ConversationReplySource = 'opener' | 'reaction' | 'persona_llm';
+type ConversationReplySource = 'opener' | 'reaction' | 'persona_llm';
 export interface ConversationTurn {
     turn_number: number;
     user_message: string;
@@ -24,7 +24,7 @@ export interface ConversationTurn {
     step_scorer_results?: ScorerResult[];
     tool_events?: import('./tool-events.js').ToolEvent[];
 }
-export type ConversationCompletionReason = 'max_turns' | 'signal' | 'done_phrase' | 'done_when' | 'timeout' | 'no_replies' | 'error';
+type ConversationCompletionReason = 'max_turns' | 'signal' | 'done_phrase' | 'done_when' | 'timeout' | 'no_replies' | 'error';
 export interface ScorerResult {
     scorer_type: string;
     score: number;
@@ -32,25 +32,7 @@ export interface ScorerResult {
     details: string;
     status?: import('./sdk/types.js').ScorerStatus;
 }
-/**
- * JSON contract for deterministic scorer stdout.
- * The scorer script must print a JSON object matching this shape to stdout.
- */
-export interface ScorerOutput {
-    /** Score between 0.0 and 1.0 (required). Clamped by pathgrade. */
-    score: number;
-    /** Human-readable summary (optional). */
-    details?: string;
-    /** Per-check breakdown, rendered as checkmarks in reports (optional). */
-    checks?: ScorerCheck[];
-}
-/** Individual check result within a ScorerOutput. */
-export interface ScorerCheck {
-    name: string;
-    passed: boolean;
-    message?: string;
-}
-export interface JudgeToolCallLogData {
+interface JudgeToolCallLogData {
     /** Tool name (readFile | listDir | grep | getToolEvents). */
     name: string;
     /** Arguments the LLM passed to the tool. */
@@ -219,6 +201,16 @@ export interface EvalReport {
     trials: TrialResult[];
     skills_used: string[];
 }
+export type ComparisonUnavailableReason = 'definition_metadata_missing' | 'scorer_metadata_missing' | 'runtime_metadata_missing' | 'sampling_incomplete';
+export interface ComparisonContract {
+    version: 1;
+    definition_revision?: string;
+    scorer_revision?: string;
+    runtime_revision?: string;
+    sampling_revision?: string;
+    report_schema_revision: 'pathgrade-results-v2';
+    unavailable_reasons?: ComparisonUnavailableReason[];
+}
 /**
  * TrialResult with `session_log` and `conversation` stripped. These fields
  * live only in the per-group trace files; the consolidated results.json keeps
@@ -232,6 +224,8 @@ export type StrippedTrialResult = Omit<TrialResult, 'session_log' | 'conversatio
 export type PathgradeGroupReport = Omit<EvalReport, 'trials' | 'status'> & {
     status?: 'pass' | 'fail';
     trials: StrippedTrialResult[];
+    source_file?: string;
+    comparison_contract?: ComparisonContract;
     /** Relative path from `.pathgrade/` to the trace file for this group. */
     trace_file: string;
 };
@@ -277,6 +271,8 @@ export interface PathgradeReport {
      * in every group passed its vitest test.
      */
     status: 'pass' | 'fail';
+    /** Agent Evals compatibility manifest when changed selection finds no runnable evals. */
+    run_kind?: 'evaluation' | 'no-affected';
     groups: PathgradeGroupReport[];
     /**
      * Present when `pathgrade run --changed` produced the run. Absent on
@@ -286,7 +282,7 @@ export interface PathgradeReport {
     selection?: PathgradeSelectionReport;
 }
 export type PassAtKUnavailableReason = 'non_binary_reward' | 'incomplete_attempts' | 'case_identity_mismatch';
-export interface TrialPaths {
+interface TrialPaths {
     root: string;
     workspace: string;
     home: string;
@@ -303,11 +299,11 @@ export interface TrialRuntime {
 }
 export type EnvironmentHandle = string | TrialRuntime;
 export type AgentCommandRunner = (cmd: string) => Promise<CommandResult>;
-export interface AgentTurnInput {
+interface AgentTurnInput {
     message: string;
     continueSession?: boolean;
 }
-export type VisibleAssistantMessageSource = 'assistant_message';
+type VisibleAssistantMessageSource = 'assistant_message';
 export interface AgentTurnResult {
     rawOutput: string;
     assistantMessage: string;
@@ -381,7 +377,6 @@ export interface AgentSession {
     dispose?(): Promise<void>;
 }
 export declare function getWorkspacePath(handle: EnvironmentHandle): string;
-export declare function getRuntimeHandle(handle: EnvironmentHandle): string;
 export declare function getRuntimeEnv(handle: EnvironmentHandle): Record<string, string>;
 export type McpConfigOrigin = 'generated_mock' | 'caller_config';
 export interface AgentSessionOptions {
@@ -435,3 +430,4 @@ export declare abstract class BaseAgent {
     run(_instruction: string, _workspacePath: string, _runCommand: AgentCommandRunner): Promise<string>;
 }
 export declare function createAgentSession(agent: BaseAgent, runtime: EnvironmentHandle, runCommand: AgentCommandRunner, options?: AgentSessionOptions): Promise<AgentSession>;
+export {};

package/dist/types.js

@@ -1,7 +1,7 @@
 export function getWorkspacePath(handle) {
     return typeof handle === 'string' ? handle : handle.workspacePath;
 }
-export function getRuntimeHandle(handle) {
+function getRuntimeHandle(handle) {
     return typeof handle === 'string' ? handle : handle.handle;
 }
 export function getRuntimeEnv(handle) {

package/dist/utils/cli.d.ts

@@ -15,21 +15,6 @@ export declare const fmt: {
 };
 /** Print a section header with a rule line */
 export declare function header(title: string, width?: number): void;
-/** Print a labeled key-value pair */
-export declare function kv(label: string, value: string, indent?: number): void;
-/** Print a trial result row */
-export declare function trialRow(trialId: number, total: number, reward: number, duration: string, commands: number, scorers: {
-    type: string;
-    score: number;
-}[]): void;
-/** Print the results summary block */
-export declare function resultsSummary(passRate: number, passAtK: number, passPowK: number, trials: number, preset?: string): void;
-/** Print a validation result */
-export declare function validationResult(passed: boolean, reward: number, scorers: {
-    type: string;
-    score: number;
-    details: string;
-}[]): void;
 /**
  * In-place spinner that shows current phase + elapsed time on one line.
  *

package/dist/utils/cli.js

@@ -27,13 +27,13 @@ export function header(title, width = 60) {
     console.log(`\n${fmt.bold(`── ${title} `)}${fmt.dim(rule)}`);
 }
 /** Print a labeled key-value pair */
-export function kv(label, value, indent = 2) {
+function kv(label, value, indent = 2) {
     const pad = ' '.repeat(indent);
     const labelPad = label.padEnd(12);
     console.log(`${pad}${fmt.dim(labelPad)}${value}`);
 }
 /** Print a trial result row */
-export function trialRow(trialId, total, reward, duration, commands, scorers) {
+function trialRow(trialId, total, reward, duration, commands, scorers) {
     const pad = '  ';
     const status = reward >= 0.5 ? fmt.pass('PASS') : fmt.fail('FAIL');
     const rewardStr = reward.toFixed(2);
@@ -47,7 +47,7 @@ export function trialRow(trialId, total, reward, duration, commands, scorers) {
     console.log(`${pad}  ${fmt.dim(trialLabel)} ${status}  ${fmt.bold(rewardStr)}  ${fmt.dim(duration.padEnd(7))} ${fmt.dim(commands + ' cmds')}  ${scorerStr}`);
 }
 /** Print the results summary block */
-export function resultsSummary(passRate, passAtK, passPowK, trials, preset) {
+function resultsSummary(passRate, passAtK, passPowK, trials, preset) {
     const presetLabel = preset === 'smoke' ? ' (smoke test)'
         : preset === 'reliable' ? ' (reliable)'
             : preset === 'regression' ? ' (regression)'
@@ -61,7 +61,7 @@ export function resultsSummary(passRate, passAtK, passPowK, trials, preset) {
     console.log();
 }
 /** Print a validation result */
-export function validationResult(passed, reward, scorers) {
+function validationResult(passed, reward, scorers) {
     for (const g of scorers) {
         const scoreStr = g.score.toFixed(2);
         const color = g.score >= 0.5 ? fmt.green(scoreStr) : fmt.red(scoreStr);

package/dist/utils/llm-mocks.d.ts

@@ -26,11 +26,11 @@ export interface CreateMockLLMOptions {
     /** When set, replaces the queue for `.call()`. Useful for dynamic / side-effecting tests. */
     respond?: (prompt: string, opts?: LLMCallOptions) => MockResponse | Promise<MockResponse>;
 }
-export interface MockCallRecord {
+interface MockCallRecord {
     prompt: string;
     opts?: LLMCallOptions;
 }
-export interface MockToolCallRecord {
+interface MockToolCallRecord {
     messages: ToolUseMessage[];
     opts: CallWithToolsOptions;
 }
@@ -41,3 +41,4 @@ export interface MockLLM extends LLMPort {
     readonly toolCalls: readonly MockToolCallRecord[];
 }
 export declare function createMockLLM(opts?: CreateMockLLMOptions): MockLLM;
+export {};

package/dist/utils/llm-providers/cli.d.ts

@@ -3,7 +3,6 @@ import type { LLMProvider } from '../llm-types.js';
 export declare function resetCliCache(): void;
 /** Check if claude CLI is installed and authenticated. */
 export declare function isClaudeCliAvailable(): Promise<boolean>;
-export declare function isCodexCliAvailable(): Promise<boolean>;
 interface CliEnvelope {
     result?: string;
     structured_output?: unknown;

package/dist/utils/llm-providers/cli.js

@@ -27,7 +27,7 @@ export function resetCliCache() {
 export async function isClaudeCliAvailable() {
     return getCliAvailability('claude', ['auth', 'status'], 5);
 }
-export async function isCodexCliAvailable() {
+async function isCodexCliAvailable() {
     return getCliAvailability('codex', ['login', 'status'], 5);
 }
 export function parseCliEnvelope(raw) {

package/dist/utils/llm-types.d.ts

@@ -73,12 +73,12 @@ export interface ToolResultBlock {
     content: string;
     is_error?: boolean;
 }
-export interface TextBlock {
+interface TextBlock {
     type: 'text';
     text: string;
 }
-export type AssistantContentBlock = TextBlock | ToolUseBlock;
-export type UserContentBlock = TextBlock | ToolResultBlock;
+type AssistantContentBlock = TextBlock | ToolUseBlock;
+type UserContentBlock = TextBlock | ToolResultBlock;
 export interface ToolUseMessage {
     role: 'user' | 'assistant';
     content: string | Array<AssistantContentBlock | UserContentBlock>;
@@ -105,3 +105,4 @@ export interface LLMProviderAdapter {
 }
 /** @deprecated Use `LLMProviderAdapter`. Alias retained only during the RFC 010a migration. */
 export type LLMProvider = LLMProviderAdapter;
+export {};

package/dist/utils/llm.d.ts

@@ -1,5 +1,5 @@
 import type { LLMProviderAdapter, LLMCallOptions, LLMCallResult, LLMPort } from './llm-types.js';
-export type { LLMCallOptions, LLMCallResult, LLMPort, LLMProviderAdapter, LLMProvider, TokenUsage } from './llm-types.js';
+export type { LLMCallOptions, LLMCallResult, LLMPort, LLMProviderAdapter, TokenUsage } from './llm-types.js';
 export interface CreateLLMClientOptions {
     adapters?: LLMProviderAdapter[];
     agent?: string;
@@ -11,7 +11,7 @@ export declare class ProviderNotSupportedError extends Error {
     readonly adapterName: string;
     constructor(adapterName: string);
 }
-export { isClaudeCliAvailable, isCodexCliAvailable, resetCliCache, extractStructuredOutput, parseCliEnvelope } from './llm-providers/cli.js';
+export { isClaudeCliAvailable, } from './llm-providers/cli.js';
 /**
  * Create an LLM client that tries providers in order.
  *

package/dist/utils/llm.js

@@ -14,7 +14,7 @@ export class ProviderNotSupportedError extends Error {
     }
 }
 // Re-export provider utilities that callsites depend on
-export { isClaudeCliAvailable, isCodexCliAvailable, resetCliCache, extractStructuredOutput, parseCliEnvelope } from './llm-providers/cli.js';
+export { isClaudeCliAvailable, } from './llm-providers/cli.js';
 function inferProviderFromModel(model) {
     const normalized = model?.trim().toLowerCase();
     if (!normalized)

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@wix/pathgrade",
-  "version": "1.0.30",
+  "version": "1.0.32",
   "packageManager": "yarn@4.12.0",
   "description": "Evaluate whether AI agents discover and use your skills correctly",
   "exports": {
@@ -140,5 +140,5 @@
     "typescript": "^5.9.3",
     "zod": "4.3.6"
   },
-  "falconPackageHash": "c85431c64d766806f41ca2adad9d390e73dc8d09a2feea2ae9285a1c"
+  "falconPackageHash": "c7a69bbfe306a62f3d89b21d92a09f18996516cd374f1772d1181d3f"
 }