autokap 1.0.7 → 1.0.8

package/dist/execution-types.js

@@ -0,0 +1,65 @@
+/**
+ * Capture Agent — IR Types
+ *
+ * All types for the compiled execution model:
+ * preset (natural language) -> ExecutionProgram (typed IR) -> deterministic runtime
+ */
+/** Sentinel value that resolves to the current variant's locale or theme at runtime */
+export const VARIANT_PLACEHOLDER = '$variant';
+// ── Opcode kinds ────────────────────────────────────────────────────
+export const OPCODE_KINDS = [
+    'NAVIGATE',
+    'DISMISS_OVERLAYS',
+    'ASSERT_ROUTE',
+    'ASSERT_SURFACE',
+    'CLICK',
+    'TYPE',
+    'PRESS_KEY',
+    'WAIT_FOR',
+    'SET_LOCALE',
+    'SET_THEME',
+    'SCROLL',
+    'CAPTURE_SCREENSHOT',
+    'CAPTURE_DOM',
+    'CAPTURE_FRAGMENT',
+    'BEGIN_CLIP',
+    'END_CLIP',
+    'HOVER',
+    'SELECT_OPTION',
+    'CHECK',
+    'DOUBLE_CLICK',
+    'CLONE_ELEMENT',
+    'INJECT_MOCK_DATA',
+    'REMOVE_ELEMENT',
+    'SET_ATTRIBUTE',
+];
+/**
+ * Soft opcodes are non-blocking — if their action or postcondition fails at
+ * runtime, the runner returns `status: 'skipped'` instead of aborting the
+ * variant. The circuit breaker is not ticked. Used by mock data injection so
+ * a missing template selector logs a warning and the capture proceeds.
+ */
+export const SOFT_OPCODE_KINDS = new Set([
+    'CLONE_ELEMENT',
+    'INJECT_MOCK_DATA',
+    'REMOVE_ELEMENT',
+    'SET_ATTRIBUTE',
+]);
+export function isSoftOpcodeKind(kind) {
+    return SOFT_OPCODE_KINDS.has(kind);
+}
+export const DEFAULT_RECOVERY_POLICY = {
+    retries: 2,
+    useSelectorMemory: true,
+    useAltInteraction: true,
+    allowReload: false,
+    allowHealer: true,
+};
+// ── Artifact spec ───────────────────────────────────────────────────
+export const MEDIA_MODES = ['screenshot', 'clip', 'dom'];
+export const DEFAULT_CIRCUIT_BREAKER = {
+    maxPerOpcode: 3,
+    maxPerPage: 5,
+    maxPerRun: 10,
+};
+//# sourceMappingURL=execution-types.js.map

package/dist/fonts-loader.d.ts

@@ -0,0 +1,14 @@
+/**
+ * SF Pro font loader for native SVG mockup rendering.
+ *
+ * Parses the woff2 base64 data URIs from sf-pro-fonts.ts / sf-pro-symbols.ts
+ * with fontkit and caches the parsed Font instances at module level. Used by
+ * src/svg-text.ts and the SVG status bar / browser bar generators to render
+ * text as native `<path>` elements (no Playwright required).
+ */
+import type { Font } from 'fontkit';
+export type SfFontId = 'text-regular' | 'text-semibold' | 'text-bold' | 'display-regular' | 'display-semibold' | 'display-bold' | 'symbols';
+/** Get a parsed Font instance for the requested SF Pro variant. Cached. */
+export declare function getFont(id: SfFontId): Font;
+/** Resolve an SF font id from a CSS-like family + weight pair. */
+export declare function resolveFontId(family: 'text' | 'display', weight?: number | 'bold' | 'semibold' | 'regular' | 'medium' | 'normal'): SfFontId;

package/dist/fonts-loader.js

@@ -0,0 +1,55 @@
+/**
+ * SF Pro font loader for native SVG mockup rendering.
+ *
+ * Parses the woff2 base64 data URIs from sf-pro-fonts.ts / sf-pro-symbols.ts
+ * with fontkit and caches the parsed Font instances at module level. Used by
+ * src/svg-text.ts and the SVG status bar / browser bar generators to render
+ * text as native `<path>` elements (no Playwright required).
+ */
+import * as fontkit from 'fontkit';
+import { SF_PRO_TEXT_REGULAR, SF_PRO_TEXT_SEMIBOLD, SF_PRO_DISPLAY_REGULAR, SF_PRO_DISPLAY_SEMIBOLD, } from './sf-pro-fonts.js';
+import { SF_PRO_SYMBOLS } from './sf-pro-symbols.js';
+const FONT_DATA = {
+    // SF Pro "Bold" maps to "Semibold" in our subset (we don't ship a separate
+    // bold weight). Native macOS uses semibold for menu bar app names anyway.
+    'text-regular': SF_PRO_TEXT_REGULAR,
+    'text-semibold': SF_PRO_TEXT_SEMIBOLD,
+    'text-bold': SF_PRO_TEXT_SEMIBOLD,
+    'display-regular': SF_PRO_DISPLAY_REGULAR,
+    'display-semibold': SF_PRO_DISPLAY_SEMIBOLD,
+    'display-bold': SF_PRO_DISPLAY_SEMIBOLD,
+    symbols: SF_PRO_SYMBOLS,
+};
+const cache = new Map();
+function dataUriToBuffer(dataUri) {
+    const comma = dataUri.indexOf(',');
+    if (comma === -1) {
+        throw new Error('invalid font data URI');
+    }
+    const base64 = dataUri.slice(comma + 1);
+    return Buffer.from(base64, 'base64');
+}
+/** Get a parsed Font instance for the requested SF Pro variant. Cached. */
+export function getFont(id) {
+    const cached = cache.get(id);
+    if (cached)
+        return cached;
+    const buf = dataUriToBuffer(FONT_DATA[id]);
+    const parsed = fontkit.create(buf);
+    // fontkit.create can return a FontCollection, but our subsets are all
+    // single fonts.
+    if ('fonts' in parsed) {
+        throw new Error(`unexpected font collection for ${id}`);
+    }
+    cache.set(id, parsed);
+    return parsed;
+}
+/** Resolve an SF font id from a CSS-like family + weight pair. */
+export function resolveFontId(family, weight = 400) {
+    const isBoldish = weight === 'bold' ||
+        weight === 'semibold' ||
+        (typeof weight === 'number' && weight >= 600);
+    const suffix = isBoldish ? 'semibold' : 'regular';
+    return `${family}-${suffix}`;
+}
+//# sourceMappingURL=fonts-loader.js.map

package/dist/hybrid-navigator.js

@@ -390,9 +390,9 @@ export function buildNavigationPrompt(resolvedUrl, config) {
         'Your tasks:',
         '1. Load the page at the URL above',
         '2. Handle login if credentials are provided',
-        '3. Dismiss ALL cookie banners, consent walls, analytics prompts, overlays, and popups — do this BEFORE calling ready_to_capture',
+        '3. Dismiss ALL cookie banners, consent walls, analytics prompts, overlays, and popups — do this BEFORE calling capture',
         '4. Navigate to the specific page/state described below',
-        '5. Once the page is clean (no overlays, no spinners, correct page), call ready_to_capture',
+        '5. Once the page is clean (no overlays, no spinners, correct page), call capture',
     ];
     if (hasTwoPartPrompt) {
         // Two-part mode: clear separation between navigation goal and what will be recorded.
@@ -402,13 +402,13 @@ export function buildNavigationPrompt(resolvedUrl, config) {
         parts.push('</navigation_goal>');
         parts.push('');
         parts.push('<recording_preview>');
-        parts.push(`After you call ready_to_capture, the recording system will perform: ${config.recordingScript}`);
+        parts.push(`After you call capture, the recording system will perform: ${config.recordingScript}`);
         parts.push('You must NOT perform these recording actions yourself. Your job is ONLY to reach the state where the recording can begin.');
         parts.push('</recording_preview>');
         parts.push('');
         parts.push('Follow the <navigation_goal> instructions exactly. Navigate to the described page/state. ' +
             'Dismiss all overlays and popups along the way. ' +
-            'Once you are on the correct page with no obstructions, call ready_to_capture.');
+            'Once you are on the correct page with no obstructions, call capture.');
     }
     else if (config.navigationPrompt) {
         // Legacy mode: single script for both navigation and recording.
@@ -423,15 +423,15 @@ export function buildNavigationPrompt(resolvedUrl, config) {
     }
     else {
         parts.push('');
-        parts.push('No specific navigation required. Just load the page, dismiss overlays, and call ready_to_capture once the exact pre-recording start state is visible.');
+        parts.push('No specific navigation required. Just load the page, dismiss overlays, and call capture once the exact pre-recording start state is visible.');
     }
     parts.push('');
-    parts.push('IMPORTANT — Before calling ready_to_capture:\n' +
-        '- Dismiss ALL overlays: cookie banners, analytics prompts, consent walls, feedback widgets, onboarding tooltips. Use dismiss_overlays or click dismiss/close buttons.\n' +
+    parts.push('IMPORTANT — Before calling capture:\n' +
+        '- Dismiss ALL overlays: cookie banners, analytics prompts, consent walls, feedback widgets, onboarding tooltips. Tap dismiss/close buttons or press Escape.\n' +
         '- If the URL is /home, /dashboard, /app, or / you are on a GENERIC dashboard, NOT inside a specific project/workspace. ' +
-        'Do NOT call ready_to_capture from a generic dashboard unless explicitly targeting the homepage.\n' +
+        'Do NOT call capture from a generic dashboard unless explicitly targeting the homepage.\n' +
         '- Check the URL path: it must match the specific entity/page required.\n' +
-        '- If you changed language or theme via settings, navigate BACK to the target page before calling ready_to_capture.');
+        '- If you changed language or theme via settings, navigate BACK to the target page before calling capture.');
     if (config.credentials) {
         parts.push('');
         parts.push('If the page requires authentication, use the provided credentials to log in.');
@@ -439,15 +439,15 @@ export function buildNavigationPrompt(resolvedUrl, config) {
     if (config.lang) {
         parts.push('');
         parts.push(`<variant_requirement type="language" value="${config.lang}">\n` +
-            `The app\'s fixed UI chrome (menus, buttons, labels, navigation) MUST be in "${config.lang}" when you call ready_to_capture.\n` +
-            `If the fixed chrome is in a different language, you MUST switch it before calling ready_to_capture.\n` +
+            `The app\'s fixed UI chrome (menus, buttons, labels, navigation) MUST be in "${config.lang}" when you call capture.\n` +
+            `If the fixed chrome is in a different language, you MUST switch it before calling capture.\n` +
             `Ignore user-generated content (project names, preset names, imported data) that may remain in another language.\n` +
             `</variant_requirement>`);
     }
     if (config.theme) {
         parts.push('');
         parts.push(`<variant_requirement type="theme" value="${config.theme}">\n` +
-            `The app MUST be in "${config.theme}" theme when you call ready_to_capture.\n` +
+            `The app MUST be in "${config.theme}" theme when you call capture.\n` +
             `If the app has an in-app theme toggle/selector, use it to switch to "${config.theme}" mode.\n` +
             `</variant_requirement>`);
     }

package/dist/index.d.ts

@@ -1,15 +1,18 @@
-export { Browser } from './browser.js';
-export { runAgent } from './agent.js';
-export { captureIsolatedElement } from './element-capture.js';
-export { dismissCookiesAndWidgets } from './cookie-dismiss.js';
+export type { ExecutionProgram, ExecutionOpcode, OpcodeKind, SemanticTarget, PostconditionSpec, PostconditionType, RecoveryPolicy, VariantSpec, PreconditionSpec, ArtifactSpec, MediaMode, CircuitBreakerConfig, HealerPatch, OpcodeResult, OpcodeResultStatus, VariantResult, ArtifactResult, RunResult, RunTelemetry, LLMStepUsage, LLMStepType, RuntimeAdapter, ClickOptions, WaitCondition, RecordingOptions, RecordingResult, HoverOpcode, SelectOptionOpcode, CheckOpcode, DoubleClickOpcode, CloneElementOpcode, InjectMockDataOpcode, RemoveElementOpcode, SetAttributeOpcode, MockDataSlot, MockDataRow, MockDataGroup, } from './execution-types.js';
+export { DEFAULT_RECOVERY_POLICY, DEFAULT_CIRCUIT_BREAKER, VARIANT_PLACEHOLDER, SOFT_OPCODE_KINDS, isSoftOpcodeKind, } from './execution-types.js';
+export { ExecutionProgramSchema, ExecutionOpcodeSchema, parseProgram, parseOpcode, safeParseProgramResult, MockDataSlotSchema, MockDataRowSchema, MockDataGroupSchema, } from './execution-schema.js';
+export { callLLM } from './llm-provider.js';
+export type { LLMProviderConfig, LLMCallResult } from './llm-provider.js';
+export { generateAltText } from './alt-text.js';
+export type { AltTextResult, AltTextContext } from './alt-text.js';
 export { logger, setOnLog, setOnScreenshot, emitScreenshot } from './logger.js';
 export type { LogEntry, LogLevel, OnLogCallback, OnScreenshotCallback } from './logger.js';
 export { applyDeviceFrame, getDeviceFrames, getDeviceFrame, invalidateDeviceConfigCache, resolveDeviceFrameDescriptor, rasterizeDeviceFrame, } from './mockup.js';
 export type { DeviceFrameId, DeviceFrameDefinition, DeviceCategory, MockupOptions, MockupOrientation, ResolvedDeviceFrameDescriptor, } from './mockup.js';
 export { generateStatusBarHtml } from './status-bar.js';
 export { generateBrowserBarHtml } from './browser-bar.js';
-export type { BrowserBarConfig, BrowserBarRenderOptions } from './browser-bar.js';
+export type { BrowserBarConfig, BrowserBarRenderOptions, BrowserBarStyle } from './browser-bar.js';
 export { convertToGif, convertToMp4, ensureFfmpegAvailable, extractThumbnail, getMediaDurationMs, } from './clip-postprocess.js';
 export { evaluateActionSecurity, describeSecurityTarget } from './security.js';
 export type { StatusBarConfig, StatusBarLayout, StatusBarDeviceType, StatusBarRenderOptions } from './status-bar.js';
-export type * from './types.js';
+export type { BrowserOptions, BrowserStorageState, BrowserSessionStorageState, AKTree, AKNode, AKType, AKBounds, AKPageState, InteractiveElement, PageState, VideoPageSignals, VideoStep, VideoStepType, VideoPlan, VideoAgentConfig, VideoAgentResult, ClipDefinition, ClipOptions, StepUsage, IsolatedElement, OutscaleConfig, CapturePageIdentity, ValidatedSessionProfile, FocusQuery, } from './types.js';

package/dist/index.js

@@ -1,7 +1,13 @@
-export { Browser } from './browser.js';
-export { runAgent } from './agent.js';
-export { captureIsolatedElement } from './element-capture.js';
-export { dismissCookiesAndWidgets } from './cookie-dismiss.js';
+// Playwright-free root entry. The CLI owns Playwright exclusively — import
+// Browser/WebPlaywrightLocal/opcode runner symbols directly via their files
+// (or the `screenshot-agent/browser` subpath for external consumers).
+// Server-side callers must use specific subpaths (ESLint enforces this).
+export { DEFAULT_RECOVERY_POLICY, DEFAULT_CIRCUIT_BREAKER, VARIANT_PLACEHOLDER, SOFT_OPCODE_KINDS, isSoftOpcodeKind, } from './execution-types.js';
+export { ExecutionProgramSchema, ExecutionOpcodeSchema, parseProgram, parseOpcode, safeParseProgramResult, MockDataSlotSchema, MockDataRowSchema, MockDataGroupSchema, } from './execution-schema.js';
+// ── LLM-backed helpers (OpenAI, no browser) ─────────────────────────
+export { callLLM } from './llm-provider.js';
+export { generateAltText } from './alt-text.js';
+// ── Shared utilities ────────────────────────────────────────────────
 export { logger, setOnLog, setOnScreenshot, emitScreenshot } from './logger.js';
 export { applyDeviceFrame, getDeviceFrames, getDeviceFrame, invalidateDeviceConfigCache, resolveDeviceFrameDescriptor, rasterizeDeviceFrame, } from './mockup.js';
 export { generateStatusBarHtml } from './status-bar.js';

package/dist/legacy/agent-action-recovery.d.ts

@@ -0,0 +1,45 @@
+/**
+ * Action recovery, guard inference, and login detection utilities for the capture agent.
+ * Extracted from agent.ts — these are pure functions with no closure dependencies.
+ */
+import type { ExecutedAction, ActionType } from '../types.js';
+export declare const META_ACTIONS: Set<ActionType>;
+export declare const BOOTSTRAP_ACTIONS: Set<ActionType>;
+export declare const REPLAYABLE_ACTIONS: ReadonlyArray<ActionType>;
+export declare function isNoEffectAction(action: ExecutedAction): boolean;
+export declare function isBootstrapStabilizationAction(action: ExecutedAction): boolean;
+export declare function hasMeaningfulBrowserAction(actionHistory: ExecutedAction[]): boolean;
+export declare function buildRecoveryActionSignature(action: ExecutedAction): string;
+export declare function isExplicitLoginAction(action: ExecutedAction): boolean;
+export declare function hasRecentExplicitLoginAction(previousActions: ExecutedAction[]): boolean;
+export declare function isLoginAction(action: ExecutedAction, previousActions?: ExecutedAction[]): boolean;
+export declare function compactReplayActions(recordedActions: ExecutedAction[], params?: {
+    currentUrl?: string;
+    targetUrl?: string;
+    currentViewport?: {
+        width: number;
+        height: number;
+    } | null;
+    isAuthenticated?: boolean;
+    ignoreRecordedViewport?: boolean;
+}): ExecutedAction[];
+export declare function countRecentNoEffectActions(actionHistory: ExecutedAction[]): number;
+export declare function shouldTriggerRecovery(actionHistory: ExecutedAction[]): boolean;
+export declare function inferPrematureGiveUpCorrection(params: {
+    reason: string;
+    actionHistory: ExecutedAction[];
+    lastVerificationFailure?: string;
+    iteration: number;
+    maxIterations: number;
+}): string | null;
+export declare function inferSearchScrollLoopGuard(params: {
+    actionHistory: ExecutedAction[];
+    action: ActionType;
+    args: Record<string, unknown>;
+}): string | null;
+export declare function inferRepeatedActionGuard(params: {
+    actionHistory: ExecutedAction[];
+    action: ActionType;
+    args: Record<string, unknown>;
+    currentUrl?: string;
+}): string | null;