autokap 1.0.7 → 1.0.8

package/dist/agent.d.ts

@@ -1,7 +1,11 @@
-import type { ChatCompletionMessageParam } from 'openai/resources/chat/completions';
 import { Browser } from './browser.js';
 import type { BrowserObservation, BrowserVerificationBundle } from './browser.js';
-import type { AgentConfig, AgentResult, ExecutedAction, ActionType, InteractiveElement, PageState, VariantCaptureManifest, VerificationResult } from './types.js';
+import type { AgentConfig, AgentResult, ActionExpectedEffect, ExecutedAction, ActionType, InteractiveElement, OverlayScopeSummary, PageState, VariantCaptureManifest, VerificationResult } from './types.js';
+export declare function buildPromptFingerprint(params: {
+    prompt: string;
+    url: string;
+    pageIdentitySummary?: string | null;
+}): string;
 export declare function inferVariantIdentityFailure(params: {
     pageId?: string | null;
     prompt: string;
@@ -10,6 +14,15 @@ export declare function inferVariantIdentityFailure(params: {
     dialogCount: number;
     pageIdentity?: VariantCaptureManifest["currentPageIdentity"];
 }): string | null;
+/** Build the reasoning body fragment for OpenRouter requests. */
+export declare function resolveProviderReasoningEffort(effort?: AgentConfig['reasoningEffort']): 'xhigh' | 'high' | 'medium' | 'low' | 'none' | null;
+type ProviderBudgetError = {
+    status?: number;
+    message?: string;
+};
+export declare function extractAffordableMaxTokens(message: string): number | null;
+export declare function isInsufficientCreditsError(error: ProviderBudgetError | null | undefined): boolean;
+export declare function deriveRetryMaxTokensFromBudgetError(error: ProviderBudgetError | null | undefined, requestedMaxTokens: number): number | null;
 export declare function inferDeterministicReadyDecision(params: {
     observation: BrowserObservation;
     assessment: string;
@@ -60,20 +73,36 @@ export declare function parseVerificationDecisionText(content: string): {
     verified: boolean;
     reason?: string;
 } | null;
-export declare function inferActionFromAssistantText(content: string): {
+export declare function inferActionFromAssistantText(content: string, options?: {
+    allowHeuristicWait?: boolean;
+}): {
     name: ActionType;
     args: Record<string, unknown>;
 } | null;
-export declare function executeAction(browser: Browser, action: ActionType, args: Record<string, unknown>): Promise<{
+export declare function executeAction(browser: Browser, action: ActionType, args: Record<string, unknown>, options?: {
+    beforeObservation?: BrowserObservation | null;
+    fast?: boolean;
+}): Promise<{
     success: boolean;
     error?: string;
     outcome?: string;
     stateChanged?: boolean;
+    expectedEffect?: ActionExpectedEffect;
+    observedEffects?: string[];
+    effectConfirmed?: boolean;
+    noEffectReason?: string | null;
+    afterObservation?: BrowserObservation | null;
 }>;
 export declare function verifyCaptureReadiness(browser: Browser, config: AgentConfig, apiKey: string, options?: {
     assessment?: string;
     stepNumber?: number;
 }): Promise<VerificationResult>;
+export declare function buildFreshNodeCaptureVerification(params: {
+    browser: Browser;
+    config: AgentConfig;
+    nodeId: string;
+}): Promise<VerificationResult | undefined>;
+export declare function urlsStrictlyMatch(expectedUrl: string | undefined, currentUrl: string | undefined): boolean;
 export declare function getLivePreviewScreenshot(pageState: Pick<PageState, 'cleanScreenshot' | 'screenshot'>): Buffer;
 export declare function analyzeReplayCandidate(recordedActions: ExecutedAction[], params?: {
     currentUrl?: string;
@@ -85,10 +114,57 @@ export declare function analyzeReplayCandidate(recordedActions: ExecutedAction[]
     isAuthenticated?: boolean;
     currentDialogCount?: number | null;
     pageIdentity?: VariantCaptureManifest['currentPageIdentity'] | null;
+    ignoreRecordedViewport?: boolean;
 }): {
     replayableActions: ExecutedAction[];
     skipReason: string | null;
+    trimmedActionCount: number;
+};
+export declare function findReplayResumeIndex(replayableActions: ExecutedAction[], currentUrl: string | undefined, startIndex?: number, observation?: BrowserObservation | null, pageIdentity?: VariantCaptureManifest['currentPageIdentity'] | null): number | null;
+export declare function sanitizeSemanticContextHint(value: string | null | undefined): string | null;
+export declare function deriveOverlayScope(params: {
+    pageState: Pick<PageState, 'akTree'>;
+    pageIdentity?: VariantCaptureManifest['currentPageIdentity'] | null;
+    prompt?: string;
+    currentSubgoal?: string | null;
+}): OverlayScopeSummary | null;
+export declare function guardTapTargetSelection(params: {
+    pageState: Pick<PageState, 'akTree'>;
+    action: ActionType;
+    args: Record<string, unknown>;
+    pageIdentity?: VariantCaptureManifest['currentPageIdentity'] | null;
+    currentSubgoal?: string | null;
+    overlayScope?: OverlayScopeSummary | null;
+}): {
+    blocked: boolean;
+    reason?: string;
+    betterCandidate?: {
+        nodeId: string;
+        label: string;
+    };
 };
+export declare function shouldBlockFocusTool(params: {
+    pageState: Pick<PageState, 'akTree'>;
+    pageIdentity?: VariantCaptureManifest['currentPageIdentity'] | null;
+    currentSubgoal?: string | null;
+    overlayScope?: OverlayScopeSummary | null;
+}): string | null;
+export declare function shouldBlockAnalyzeTool(params: {
+    overlayScope?: OverlayScopeSummary | null;
+    overlayChangedRecently: boolean;
+    fingerprint: string;
+    lastAnalyzeFingerprint?: string | null;
+    lastAnalyzeIteration?: number;
+    iteration: number;
+    lastExecutionFailed: boolean;
+}): string | null;
+export declare function buildCompactAKTreeSummary(params: {
+    pageState: Pick<PageState, 'akTree'>;
+    pageIdentity?: VariantCaptureManifest['currentPageIdentity'] | null;
+    currentSubgoal?: string | null;
+    overlayScope?: OverlayScopeSummary | null;
+    lastVerificationFailure?: string;
+}): string;
 export declare function matchRecordedActionToElement(action: Pick<ExecutedAction, 'params'>, interactiveElements: InteractiveElement[]): InteractiveElement | null;
 export declare function resolveReplayActionArgs(action: ExecutedAction, interactiveElements: InteractiveElement[]): {
     args: Record<string, unknown> | null;
@@ -103,6 +179,8 @@ export declare function compactReplayActions(recordedActions: ExecutedAction[],
     } | null;
     /** When true, filter out login/auth actions (session already active). */
     isAuthenticated?: boolean;
+    /** When true, drop recorded resize actions and let the orchestrator own the viewport. */
+    ignoreRecordedViewport?: boolean;
 }): ExecutedAction[];
 export declare function countRecentNoEffectActions(actionHistory: ExecutedAction[]): number;
 export declare function shouldTriggerRecovery(actionHistory: ExecutedAction[]): boolean;
@@ -113,6 +191,11 @@ export declare function inferPrematureGiveUpCorrection(params: {
     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;
@@ -126,11 +209,11 @@ export interface WorkflowCacheEntry {
 }
 export declare function findReusableWorkflow(workflowCache: WorkflowCacheEntry[], subgoalName: string): WorkflowCacheEntry | null;
 /**
- * Trim the conversation thread to avoid context window overflow.
- * Always preserves the first `preservedPrefixMessages` entries, then the last maxMessages messages.
- * Also compresses old screenshots and DOM blocks to save tokens.
+ * Replace base64 image content in older conversation messages with a text placeholder.
+ * Keeps only the most recent `keepRecentImages` user messages that contain screenshots.
+ * This dramatically reduces token cost on long agent runs (40-60% savings).
  */
-export declare function trimConversationHistory(messages: ChatCompletionMessageParam[], maxMessages?: number, preservedPrefixMessages?: number): void;
+export { trimConversationHistory } from './agent-message-utils.js';
 export declare function runAgent(browser: Browser, config: AgentConfig, apiKey: string): Promise<AgentResult>;
 /**
  * Replay a previous successful capture's recorded actions instead of running the full AI agent.
@@ -139,4 +222,5 @@ export declare function runAgent(browser: Browser, config: AgentConfig, apiKey:
  */
 export declare function replayAgent(browser: Browser, config: AgentConfig, apiKey: string, recordedActions: ExecutedAction[], options?: {
     allowFullAgentFallback?: boolean;
+    ignoreRecordedViewport?: boolean;
 }): Promise<AgentResult>;