explorbot 0.3.1 → 0.3.4

package/bin/explorbot-cli.ts

@@ -7,6 +7,7 @@ import { Command } from 'commander';
 import figureSet from 'figures';
 import { render } from 'ink';
 import React from 'react';
+import { flushTelemetry } from '../src/ai/provider.js';
 import { App } from '../src/components/App.js';
 import { StatusPane } from '../src/components/StatusPane.js';
 import { ConfigParser, EXPLORBOT_ENV_VARS, PROVIDERS } from '../src/config.js';
@@ -15,7 +16,7 @@ import { remote } from '../src/remote.js';
 import { Stats } from '../src/stats.js';
 import { Plan } from '../src/test-plan.js';
 import { getCliName } from '../src/utils/cli-name.ts';
-import { isVerboseMode, log, setPreserveConsoleLogs, setQuietMode } from '../src/utils/logger.js';
+import { isVerboseMode, log, setPreserveConsoleLogs, setQuietMode, tag } from '../src/utils/logger.js';
 import { jsonToTable } from '../src/utils/markdown-parser.js';
 import { parseMarkdownToTerminal } from '../src/utils/markdown-terminal.js';
 import { type NextStepSection, printNextSteps, relativeToCwd } from '../src/utils/next-steps.ts';
@@ -29,6 +30,16 @@ const pkgVersion = JSON.parse(fs.readFileSync(pkgPath, 'utf-8')).version as stri
 program.name(cli).description('AI-powered web exploration tool').version(pkgVersion, '-V, --version');
 remote.registerOption(program);
 
+process.on('uncaughtException', async (error) => {
+  tag('error').log(`Uncaught exception: ${error instanceof Error ? `${error.message}\n${error.stack}` : String(error)}`);
+  await flushTelemetry();
+  process.exit(1);
+});
+
+process.on('unhandledRejection', (reason) => {
+  tag('error').log(`Unhandled rejection: ${reason instanceof Error ? `${reason.message}\n${reason.stack}` : String(reason)}`);
+});
+
 if (!process.env.EXPLORBOT_NO_BANNER && !process.argv.includes('prima')) {
   console.log(`⛵ ${chalk.yellow.bold(`Explorbot v${pkgVersion}`)} ${chalk.dim('Autonomous Testing Agent')}`);
 }
@@ -103,9 +114,7 @@ async function startTUI(explorBot: ExplorBot): Promise<void> {
 async function showStatsAndExit(code: number): Promise<never> {
   if (remote.isAttached()) {
     await remote.close(code);
-    process.exit(code);
-  }
-  if (Stats.hasActivity()) {
+  } else if (Stats.hasActivity()) {
     await new Promise<void>((resolve) => {
       const { unmount } = render(
         React.createElement(StatusPane, {
@@ -121,6 +130,7 @@ async function showStatsAndExit(code: number): Promise<never> {
       );
     });
   }
+  await flushTelemetry();
   process.exit(code);
 }
 
@@ -546,6 +556,7 @@ program
   .command('learn [url] [description]')
   .description('Add knowledge for URLs')
   .option('-p, --path <path>', 'Working directory path')
+  .option('--replace', 'Replace existing knowledge for this URL instead of appending')
   .action(async (url, description, options) => {
     try {
       await ConfigParser.getInstance().loadConfig({
@@ -556,7 +567,7 @@ program
       const tracker = new KnowledgeTracker();
 
       if (url && description) {
-        const result = tracker.addKnowledge(url, description);
+        const result = tracker.addKnowledge(url, description, { replace: options.replace });
         const action = result.isNewFile ? 'Created' : 'Updated';
         console.log(`Knowledge ${action} in: ${result.filename}`);
         return;

package/dist/bin/explorbot-cli.js

@@ -7,6 +7,7 @@ import { Command } from 'commander';
 import figureSet from 'figures';
 import { render } from 'ink';
 import React from 'react';
+import { flushTelemetry } from '../src/ai/provider.js';
 import { App } from '../src/components/App.js';
 import { StatusPane } from '../src/components/StatusPane.js';
 import { ConfigParser, EXPLORBOT_ENV_VARS, PROVIDERS } from '../src/config.js';
@@ -15,7 +16,7 @@ import { remote } from '../src/remote.js';
 import { Stats } from '../src/stats.js';
 import { Plan } from '../src/test-plan.js';
 import { getCliName } from "../src/utils/cli-name.js";
-import { isVerboseMode, log, setPreserveConsoleLogs, setQuietMode } from '../src/utils/logger.js';
+import { isVerboseMode, log, setPreserveConsoleLogs, setQuietMode, tag } from '../src/utils/logger.js';
 import { jsonToTable } from '../src/utils/markdown-parser.js';
 import { parseMarkdownToTerminal } from '../src/utils/markdown-terminal.js';
 import { printNextSteps, relativeToCwd } from "../src/utils/next-steps.js";
@@ -25,6 +26,14 @@ const pkgPath = path.resolve(path.dirname(fileURLToPath(import.meta.url)), '../p
 const pkgVersion = JSON.parse(fs.readFileSync(pkgPath, 'utf-8')).version;
 program.name(cli).description('AI-powered web exploration tool').version(pkgVersion, '-V, --version');
 remote.registerOption(program);
+process.on('uncaughtException', async (error) => {
+    tag('error').log(`Uncaught exception: ${error instanceof Error ? `${error.message}\n${error.stack}` : String(error)}`);
+    await flushTelemetry();
+    process.exit(1);
+});
+process.on('unhandledRejection', (reason) => {
+    tag('error').log(`Unhandled rejection: ${reason instanceof Error ? `${reason.message}\n${reason.stack}` : String(reason)}`);
+});
 if (!process.env.EXPLORBOT_NO_BANNER && !process.argv.includes('prima')) {
     console.log(`⛵ ${chalk.yellow.bold(`Explorbot v${pkgVersion}`)} ${chalk.dim('Autonomous Testing Agent')}`);
 }
@@ -78,9 +87,8 @@ async function startTUI(explorBot) {
 async function showStatsAndExit(code) {
     if (remote.isAttached()) {
         await remote.close(code);
-        process.exit(code);
     }
-    if (Stats.hasActivity()) {
+    else if (Stats.hasActivity()) {
         await new Promise((resolve) => {
             const { unmount } = render(React.createElement(StatusPane, {
                 onComplete: () => {
@@ -93,6 +101,7 @@ async function showStatsAndExit(code) {
             });
         });
     }
+    await flushTelemetry();
     process.exit(code);
 }
 addCommonOptions(program.command('start [path]').description('Start web exploration')).action(async (startPath, options) => {
@@ -492,6 +501,7 @@ program
     .command('learn [url] [description]')
     .description('Add knowledge for URLs')
     .option('-p, --path <path>', 'Working directory path')
+    .option('--replace', 'Replace existing knowledge for this URL instead of appending')
     .action(async (url, description, options) => {
     try {
         await ConfigParser.getInstance().loadConfig({
@@ -500,7 +510,7 @@ program
         const { KnowledgeTracker } = await import('../src/knowledge-tracker.js');
         const tracker = new KnowledgeTracker();
         if (url && description) {
-            const result = tracker.addKnowledge(url, description);
+            const result = tracker.addKnowledge(url, description, { replace: options.replace });
             const action = result.isNewFile ? 'Created' : 'Updated';
             console.log(`Knowledge ${action} in: ${result.filename}`);
             return;

package/dist/models.json

@@ -18,6 +18,8 @@
     "agenticModel": "gpt-5.6-luna"
   },
   "anthropic": {
+    "model": "claude-haiku-4-5-20251001",
+    "visionModel": "claude-haiku-4-5-20251001",
     "agenticModel": "claude-haiku-4-5-20251001"
   },
   "mistral": {

package/dist/package.json

@@ -1,6 +1,6 @@
 {
   "name": "explorbot",
-  "version": "0.3.1",
+  "version": "0.3.4",
   "description": "CLI app built with React Ink, CodeceptJS, and Playwright",
   "license": "Elastic-2.0",
   "type": "module",

package/dist/src/action-result.d.ts

@@ -2,6 +2,7 @@ import { type HtmlConfig } from './config.js';
 import type { Link, WebPageState } from './state-manager.js';
 import { TTLCache } from './utils/cache.js';
 import { type HtmlDiffPart, type HtmlDiffResult } from './utils/html-diff.js';
+import { Overlay } from './utils/overlay.js';
 interface ActionResultData extends WebPageState {
     html?: string;
     fullUrl?: string | undefined;
@@ -28,6 +29,7 @@ interface ActionResultData extends WebPageState {
     focusedElement?: FocusedElement | null;
     iframeURL?: string;
     links?: Link[];
+    overlayHtml?: string;
 }
 export interface PageDiff {
     urlChanged: boolean;
@@ -81,6 +83,7 @@ export declare class ActionResult implements ActionResultData {
     notes: string[];
     links: Link[];
     verifications?: Record<string, boolean>;
+    overlay: Overlay;
     constructor(data: ActionResultData);
     get hash(): string;
     get html(): string;
@@ -132,6 +135,7 @@ export declare class Diff {
     isSameUrl(): boolean;
     urlHasChanged(): boolean;
     get htmlParts(): HtmlDiffPart[];
+    cleanedHtmlParts(): Promise<HtmlDiffPart[]>;
     get ariaChanged(): string | null;
     get ariaChangeCount(): number;
     get htmlDiff(): HtmlDiffResult | null;

package/dist/src/action-result.js

@@ -5,6 +5,7 @@ import { TTLCache } from "./utils/cache.js";
 import { htmlDiff, liveRegionMessages } from "./utils/html-diff.js";
 import { extractHeadings, extractLinks, extractTargetedHtml, htmlCombinedSnapshot, htmlMinimalUISnapshot, htmlTextSnapshot, minifyHtml } from "./utils/html.js";
 import { createDebug } from "./utils/logger.js";
+import { Overlay } from "./utils/overlay.js";
 import { slugify } from "./utils/strings.js";
 import { extractStatePath, matchesUrl } from "./utils/url-matcher.js";
 const debugLog = createDebug('explorbot:state');
@@ -37,6 +38,7 @@ export class ActionResult {
     notes = [];
     links = [];
     verifications;
+    overlay = new Overlay();
     constructor(data) {
         this.id = data.id;
         this.timestamp = data.timestamp ?? new Date();
@@ -77,6 +79,7 @@ export class ActionResult {
         if (data.ariaSnapshot !== undefined) {
             this._ariaSnapshot = data.ariaSnapshot;
         }
+        this.overlay = Overlay.resolve(data);
         if (!this.fullUrl && this.url) {
             this.fullUrl = this.url;
         }
@@ -441,17 +444,9 @@ export class ActionResult {
             pageDiff.ariaChangeCount = diff.ariaChangeCount;
         }
         if (diff.htmlParts.length > 0) {
-            const htmlConfig = this.normalizeHtmlConfig();
-            const processedParts = [];
-            for (const part of diff.htmlParts) {
-                const filteredHtml = htmlCombinedSnapshot(part.subtree, htmlConfig?.combined);
-                const minified = await minifyHtml(filteredHtml);
-                if (minified) {
-                    processedParts.push({ ...part, subtree: minified });
-                }
-            }
-            if (processedParts.length > 0) {
-                pageDiff.htmlParts = collapseHtmlParts(processedParts);
+            const collapsed = collapseHtmlParts(await diff.cleanedHtmlParts());
+            if (collapsed.length > 0) {
+                pageDiff.htmlParts = collapsed;
             }
         }
         if (pageDiff.ariaChanges && this.iframeSnapshots.length > 0) {
@@ -487,7 +482,9 @@ function collapseHtmlParts(parts) {
     const total = parts.reduce((sum, p) => sum + p.subtree.length, 0);
     const fullPageReRender = total > HTML_PARTS_TOTAL_BUDGET || parts.length > HTML_PARTS_COUNT_LIMIT;
     if (fullPageReRender) {
-        return parts.map((part) => ({
+        return parts
+            .filter((part) => part.added.length > 0 || part.removed.length > 0)
+            .map((part) => ({
             ...part,
             subtree: `<html><head></head><body>...collapsed (${part.subtree.length} chars, ${part.added.length} added, ${part.removed.length} removed)...</body></html>`,
         }));
@@ -540,6 +537,17 @@ export class Diff {
             return [];
         return this._htmlDiffResult.parts;
     }
+    async cleanedHtmlParts() {
+        const htmlConfig = ConfigParser.getInstance().getConfig().html;
+        const cleaned = [];
+        for (const part of this.htmlParts) {
+            const minified = await minifyHtml(htmlCombinedSnapshot(part.subtree, htmlConfig?.combined));
+            if (!minified)
+                continue;
+            cleaned.push({ ...part, subtree: minified });
+        }
+        return cleaned;
+    }
     get ariaChanged() {
         return this._ariaDiffResult;
     }

package/dist/src/action.d.ts

@@ -15,6 +15,7 @@ declare class Action {
         name: string;
         args: any[];
     }>;
+    executedSteps: ExecutedStep[];
     lastValue: unknown;
     recorder?: PlaywrightRecorder;
     recovery: RecoveryRunner;
@@ -32,6 +33,7 @@ declare class Action {
         includeScreenshot?: boolean;
         codeBlock?: string;
     }): Promise<ActionResult>;
+    captureOverlayHtml(): Promise<string>;
     captureMainDocumentStatus(): Promise<number | undefined>;
     captureResponses(): () => void;
     recordNetworkCall(request: any, status: number): void;
@@ -57,3 +59,12 @@ export type RecoveryRunner = <T>(fn: () => Promise<T>) => Promise<T>;
 export interface ExecuteOptions {
     verbatim?: boolean;
 }
+export declare const attachStepLogger: (target: ExecutedStep[], assertionsTarget?: Array<{
+    name: string;
+    args: any[];
+}>) => (() => void);
+export interface ExecutedStep {
+    command: string;
+    success: boolean;
+    error?: string;
+}

package/dist/src/action.js

@@ -8,8 +8,9 @@ import { clearActivity, setActivity } from "./activity.js";
 import { ConfigParser, outputPath } from './config.js';
 import { Observability } from "./observability.js";
 import { browserErrorMessage, isFatalBrowserError, isNavigationTransitionError } from "./utils/browser-errors.js";
-import { captureHtmlForSnapshot, htmlCombinedSnapshot, minifyHtml } from './utils/html.js';
+import { captureHtmlForSnapshot, getVisibleOverlayHtmlExtractorSource, htmlCombinedSnapshot, minifyHtml } from './utils/html.js';
 import { createDebug, setStepSpanParent, tag } from './utils/logger.js';
+import { Overlay } from './utils/overlay.js';
 import { sleep, waitForPageReadiness } from "./utils/page-readiness.js";
 import { safeFilename } from "./utils/strings.js";
 import { codeceptJSSandbox, hasPlaywrightCommands, playwrightSandbox, sanitizeCodeBlock } from "./utils/web-sandbox.js";
@@ -30,6 +31,7 @@ class Action {
     playwrightHelper;
     playwrightGroupId = null;
     assertionSteps = [];
+    executedSteps = [];
     lastValue;
     recorder;
     recovery;
@@ -131,10 +133,13 @@ class Action {
             let ariaSnapshot = null;
             let ariaSnapshotFile = undefined;
             let focusedElement = null;
+            let overlayHtml = '';
             try {
                 const page = this.playwrightHelper.page;
                 ariaSnapshot = await page.locator('body').ariaSnapshot();
                 focusedElement = await page.evaluate(readFocusedElement);
+                if (!frame)
+                    overlayHtml = await this.captureOverlayHtml();
             }
             catch (err) {
                 debugLog('ARIA snapshot failed:', err instanceof Error ? `${err.message}\n${err.stack}` : err);
@@ -161,6 +166,7 @@ class Action {
                 ariaSnapshot,
                 ariaSnapshotFile,
                 focusedElement,
+                overlayHtml: overlayHtml || undefined,
                 iframeURL: frame ? frame.url?.() || 'iframe' : undefined,
             });
             this.stateManager.updateState(result, codeBlock);
@@ -175,6 +181,12 @@ class Action {
             return new ActionResult({ url, error: msg });
         }
     }
+    async captureOverlayHtml() {
+        return this.playwrightHelper.page.evaluate(({ extractorSource, config }) => {
+            const extract = new Function(`return ${extractorSource}`)();
+            return extract(config);
+        }, { extractorSource: getVisibleOverlayHtmlExtractorSource(), config: Overlay.captureConfig() });
+    }
     async captureMainDocumentStatus() {
         if (this.mainDocumentStatus)
             return this.mainDocumentStatus;
@@ -293,7 +305,7 @@ class Action {
         let codeString = code.replace(/^\(I\) => /, '').trim();
         const executedSteps = [];
         const assertionSteps = [];
-        const stepListener = attachStepLogger(executedSteps, assertionSteps);
+        const detachSteps = attachStepLogger(executedSteps, assertionSteps);
         const groupId = this.recorder ? await this.recorder.beginAction(codeString) : null;
         this.playwrightGroupId = groupId;
         const detachResponses = this.captureResponses();
@@ -319,10 +331,12 @@ class Action {
                 await recorder.add(() => sleep(this.config.action?.delay || 500));
                 await recorder.promise();
                 this.lastValue = await returned;
+                if (!recorder.isRunning())
+                    throw new Error('CodeceptJS recorder is stopped, commands were skipped and never reached the browser');
             }
             this.restorePageTimeout();
             if (executedSteps.length > 0) {
-                codeString = executedSteps.join('\n');
+                codeString = executedSteps.map((step) => step.command).join('\n');
             }
             const pageState = await this.captureOnce({ codeBlock: codeString });
             this.actionResult = pageState;
@@ -339,11 +353,12 @@ class Action {
             throw err;
         }
         finally {
+            this.executedSteps = executedSteps;
             this.restorePageTimeout();
             detachResponses();
             if (groupId)
                 await this.recorder.endAction();
-            detachStepLogger(stepListener);
+            detachSteps();
             if (stepSpan) {
                 stepSpan.end();
             }
@@ -432,13 +447,16 @@ async function captureTitle(page, actor) {
     return '';
 }
 const ASSERTION_STEP_NAMES = new Set(['see', 'dontSee', 'seeElement', 'dontSeeElement', 'seeInField', 'dontSeeInField', 'seeInCurrentUrl', 'dontSeeInCurrentUrl']);
-const attachStepLogger = (target, assertionsTarget) => {
+export const attachStepLogger = (target, assertionsTarget) => {
     const listener = (step, error) => {
         if (!step?.toCode)
             return;
         if (step.name?.startsWith('grab'))
             return;
-        target.push(step.toCode());
+        const executed = { command: step.toCode(), success: !error };
+        if (error)
+            executed.error = errorToString(error);
+        target.push(executed);
         if (assertionsTarget && ASSERTION_STEP_NAMES.has(step.name)) {
             assertionsTarget.push({ name: step.name, args: step.args || [] });
         }
@@ -450,11 +468,10 @@ const attachStepLogger = (target, assertionsTarget) => {
     };
     codeceptjs.event.dispatcher.on(codeceptjs.event.step.passed, listener);
     codeceptjs.event.dispatcher.on(codeceptjs.event.step.failed, listener);
-    return listener;
-};
-const detachStepLogger = (listener) => {
-    codeceptjs.event.dispatcher.off(codeceptjs.event.step.passed, listener);
-    codeceptjs.event.dispatcher.off(codeceptjs.event.step.failed, listener);
+    return () => {
+        codeceptjs.event.dispatcher.off(codeceptjs.event.step.passed, listener);
+        codeceptjs.event.dispatcher.off(codeceptjs.event.step.failed, listener);
+    };
 };
 const readFocusedElement = () => {
     const el = document.activeElement;

package/dist/src/ai/conversation.d.ts

@@ -4,8 +4,9 @@ export interface ToolExecution {
     input: any;
     output: any;
     wasSuccessful: boolean;
+    reasoning?: string;
 }
-export declare function toToolExecution(toolName: string, input: any, rawOutput: any): ToolExecution;
+export declare function toToolExecution(toolName: string, input: any, rawOutput: any, reasoning?: string): ToolExecution;
 export declare function toolExecutionLabel(input: Record<string, any> | undefined): string;
 export declare const NARRATION_TOOL = "commentary";
 export declare class Conversation {

package/dist/src/ai/conversation.js

@@ -1,8 +1,8 @@
-export function toToolExecution(toolName, input, rawOutput) {
+export function toToolExecution(toolName, input, rawOutput, reasoning) {
     let output = rawOutput;
     if (rawOutput?.type === 'json' && rawOutput?.value)
         output = rawOutput.value;
-    return { toolName, input, output, wasSuccessful: output?.success !== false };
+    return { toolName, input, output, wasSuccessful: output?.success !== false, reasoning };
 }
 export function toolExecutionLabel(input) {
     return input?.explanation || input?.assertion || input?.reason || input?.request || '';
@@ -189,10 +189,14 @@ export class Conversation {
                 continue;
             if (!Array.isArray(message.content))
                 continue;
+            const reasoning = message.content
+                .filter((part) => part.type === 'reasoning' && part.text?.trim())
+                .map((part) => part.text.trim())
+                .join('\n');
             for (const part of message.content) {
                 if (part.type !== 'tool-call')
                     continue;
-                toolCalls.set(part.toolCallId, part.input);
+                toolCalls.set(part.toolCallId, { input: part.input, reasoning });
             }
         }
         const executions = [];
@@ -206,7 +210,8 @@ export class Conversation {
                     continue;
                 if (part.toolName === NARRATION_TOOL)
                     continue;
-                executions.push(toToolExecution(part.toolName, toolCalls.get(part.toolCallId) || {}, part.output));
+                const call = toolCalls.get(part.toolCallId);
+                executions.push(toToolExecution(part.toolName, call?.input || {}, part.output, call?.reasoning));
             }
         }
         return executions;

package/dist/src/ai/driller.js

@@ -9,6 +9,7 @@ import { collectInteractiveNodes } from "../utils/aria.js";
 import { EXPLORBOT_ATTRS, HTML_COMPOSITE_AREA_HINTS, HTML_COMPOSITE_TARGET_ROLES, HTML_EXTRACTION_LIMITS, HTML_FORM_CONTROL_ROLES, HTML_FORM_CONTROL_TAGS, HTML_INTERACTIVE_ROLES, HTML_SELECTORS, HTML_VISIBILITY_LIMITS, getComponentScopeHtmlExtractorSource, getVisibleOverlayHtmlExtractorSource, inferHtmlRole, } from "../utils/html.js";
 import { createDebug, tag } from "../utils/logger.js";
 import { loop, pause } from "../utils/loop.js";
+import { OVERLAY_SELECTORS } from "../utils/overlay.js";
 import { annotatePageElements } from "../utils/web-annotate.js";
 import { eidxInContainer } from "../utils/web-eidx.js";
 import { WebElement } from "../utils/web-element.js";
@@ -578,7 +579,8 @@ export class Driller extends TaskAgent {
             config: {
                 interactiveContentSelector: HTML_SELECTORS.interactiveContent,
                 limits: HTML_EXTRACTION_LIMITS,
-                overlaySelectors: HTML_SELECTORS.semanticOverlays,
+                overlaySelectors: OVERLAY_SELECTORS.semanticOverlays,
+                overlaySemanticSelector: OVERLAY_SELECTORS.overlaySemanticSelector,
                 visibilityLimits: HTML_VISIBILITY_LIMITS,
             },
         }));

package/dist/src/ai/navigator.d.ts

@@ -1,5 +1,6 @@
 import { ActionResult } from '../action-result.js';
 import type Action from '../action.js';
+import type { ExecutedStep } from '../action.js';
 import type { ExplorbotConfig } from '../config.js';
 import type { ExperienceTracker } from '../experience-tracker.js';
 import Explorer from '../explorer.js';
@@ -16,6 +17,7 @@ declare class Navigator implements Agent {
     hooksRunner: HooksRunner;
     MAX_ATTEMPTS: number;
     lastFailureReason: string | null;
+    executedSteps: ExecutedStep[];
     systemPrompt: string;
     freeSailSystemPrompt: string;
     explorer: Explorer;
@@ -55,6 +57,7 @@ declare class Navigator implements Agent {
         freshState: ActionResult;
         urlMatches: boolean;
     }>;
+    targetUrlReached(action: Action, expectedUrl: string, state: ActionResult): boolean;
     ariaDiff(freshState: ActionResult, previous: ActionResult): Promise<string | null>;
     saveFlow(message: string, expectedUrl: string | undefined, actionResult: ActionResult, progressBlocks: string[]): void;
     rescueDelayedRedirect(action: Action, expectedUrl: string): Promise<boolean>;

package/dist/src/ai/navigator.js

@@ -2,8 +2,8 @@ import { tool } from 'ai';
 import dedent from 'dedent';
 import { z } from 'zod';
 import { ActionResult } from '../action-result.js';
-import { normalizeUrl } from '../state-manager.js';
 import { renderAssertion } from "../playwright-recorder.js";
+import { normalizeUrl } from '../state-manager.js';
 import { isFatalBrowserError } from "../utils/browser-errors.js";
 import { getCliName } from "../utils/cli-name.js";
 import { extractCodeBlocks } from '../utils/code-extractor.js';
@@ -26,6 +26,7 @@ class Navigator {
     hooksRunner;
     MAX_ATTEMPTS = Number.parseInt(process.env.MAX_ATTEMPTS || '5');
     lastFailureReason = null;
+    executedSteps = [];
     systemPrompt = dedent `
   <role>
     You are senior test automation engineer with master QA skills.
@@ -194,10 +195,15 @@ class Navigator {
         if (!this.provider)
             throw new Error('AI-assisted recovery is unavailable: no AI model is configured.');
         this.lastFailureReason = null;
+        this.executedSteps = [];
         tag('info').log('AI Navigator resolving state at', actionResult.url);
         debugLog('Resolution message:', message);
         const action = opts?.action ?? this.explorer.action();
         const expectedUrl = opts?.expectedUrl;
+        if (expectedUrl && this.targetUrlReached(action, expectedUrl, actionResult)) {
+            tag('success').log(`Already at ${expectedUrl} — navigation resolved`);
+            return true;
+        }
         const knowledge = this.knowledgeTracker.renderRelevantContext(actionResult);
         const conversation = this.provider.startConversation(this.systemPrompt, 'navigator');
         conversation.addUserText(await this.buildResolutionPrompt(message, actionResult, opts?.experience));
@@ -278,14 +284,20 @@ class Navigator {
                 const freshHash = check.freshState.getStateHash();
                 resolved = check.urlMatches && freshHash !== actionResult.getStateHash();
                 if (!resolved && attempt.ok) {
-                    lastFailure = `URL did not change (still ${check.freshState.url})`;
+                    if (check.urlMatches) {
+                        lastFailure = `Reached ${check.freshState.url} but the page state did not change`;
+                        tag('warning').log(`Page state did not change at ${check.freshState.url}`);
+                    }
+                    else {
+                        lastFailure = `Reached ${check.freshState.url}, expected ${expectedUrl}`;
+                        tag('warning').log(`URL verification failed: expected ${expectedUrl}, got ${check.freshState.url}`);
+                    }
                     batchFailures.push({
                         code: codeBlock,
                         error: lastFailure,
                         ariaChanges: await this.ariaDiff(check.freshState, prevActionResult),
                         urlAfter: check.freshState.url,
                     });
-                    tag('warning').log(`URL verification failed: expected ${expectedUrl}, got ${check.freshState.url}`);
                 }
                 if (freshHash !== prevHash && (attempt.ok || check.urlMatches)) {
                     progressBlocks.push(codeBlock);
@@ -411,6 +423,7 @@ class Navigator {
         await action.exitIframe();
         debugLog(`Attempting resolution: ${codeBlock}`);
         const ok = await action.attempt(codeBlock, message);
+        this.executedSteps.push(...action.executedSteps);
         const page = action.playwrightHelper?.page;
         if (page) {
             try {
@@ -437,9 +450,12 @@ class Navigator {
             }
         }
         const freshState = await this.explorer.capture();
-        const urlMatches = this.isSameExpectedOrigin(expectedUrl, action.stateManager) && matchesNavigationUrl(expectedUrl, this.comparableUrl(freshState, expectedUrl));
+        const urlMatches = this.targetUrlReached(action, expectedUrl, freshState);
         return { freshState, urlMatches };
     }
+    targetUrlReached(action, expectedUrl, state) {
+        return this.isSameExpectedOrigin(expectedUrl, action.stateManager) && matchesNavigationUrl(expectedUrl, this.comparableUrl(state, expectedUrl));
+    }
     async ariaDiff(freshState, previous) {
         if (freshState.getStateHash() === previous.getStateHash())
             return null;

package/dist/src/ai/pilot.js

@@ -5,7 +5,7 @@ import { ActionResult } from "../action-result.js";
 import { ConfigParser } from "../config.js";
 import { Stats } from "../stats.js";
 import { TestResult } from "../test-plan.js";
-import { collectInteractiveNodes, detectFocusArea } from "../utils/aria.js";
+import { collectInteractiveNodes } from "../utils/aria.js";
 import { ErrorPageError } from "../utils/error-page.js";
 import { createDebug, tag } from "../utils/logger.js";
 const debugLog = createDebug('explorbot:pilot');
@@ -16,6 +16,7 @@ import { withdrawVisionTools } from "./tools.js";
 const CHECK_TOOLS = ['verify', 'see', 'research'];
 const EVIDENCE_TOOLS = ['verify', 'see'];
 const META_TOOLS = ['record', 'reset', 'stop', 'finish'];
+const PILOT_REASONING_LIMIT = 500;
 const PILOT_MESSAGE_LIMIT = 2;
 const PILOT_MESSAGE_MAX_LENGTH = 160;
 export class Pilot {
@@ -738,7 +739,7 @@ export class Pilot {
         lines.push(`h2: ${state.h2 || ''}`);
         lines.push(`h3: ${state.h3 || ''}`);
         lines.push(`h4: ${state.h4 || ''}`);
-        const focusArea = detectFocusArea(state.ariaSnapshot);
+        const focusArea = state.overlay;
         if (focusArea.detected) {
             lines.push(`modal: ${focusArea.name || focusArea.type}`);
         }
@@ -941,6 +942,12 @@ export class Pilot {
                 line += `\n   result: ${resultMessage}`;
             if (errorDetail && errorDetail !== resultMessage)
                 line += `\n   error: ${errorDetail}`;
+            if (!t.wasSuccessful && t.reasoning) {
+                let rationale = t.reasoning;
+                if (rationale.length > PILOT_REASONING_LIMIT)
+                    rationale = `...${rationale.slice(-PILOT_REASONING_LIMIT)}`;
+                line += `\n   tester reasoned: ${rationale.replace(/\n+/g, ' ')}`;
+            }
             const attempts = t.output?.attempts;
             if (attempts && attempts.length > 1 && t.wasSuccessful) {
                 const failedBefore = attempts.filter((a) => !a.success);

package/dist/src/ai/provider.d.ts

@@ -7,6 +7,7 @@ declare class AiError extends Error {
 }
 export declare class ContextLengthError extends Error {
 }
+export declare function flushTelemetry(): Promise<void>;
 export declare class Provider {
     config: AIConfig;
     telemetryEnabled: boolean;
@@ -17,6 +18,7 @@ export declare class Provider {
     modelCallWaiters: (() => void)[];
     constructor(config: AIConfig);
     validateConnection(): Promise<void>;
+    stop(): Promise<void>;
     getModelForAgent(agentName?: string): any;
     getAgenticModel(agentName?: string): any;
     getVisionModel(): any;