windows-use 0.1.0 → 0.2.0

package/dist/index.d.ts

@@ -7,7 +7,7 @@ declare const ConfigSchema: z.ZodObject<{
     baseURL: z.ZodString;
     model: z.ZodString;
     maxSteps: z.ZodDefault<z.ZodNumber>;
-    contextWindowSize: z.ZodDefault<z.ZodNumber>;
+    maxRounds: z.ZodDefault<z.ZodNumber>;
     cdpUrl: z.ZodDefault<z.ZodString>;
     timeoutMs: z.ZodDefault<z.ZodNumber>;
 }, "strip", z.ZodTypeAny, {
@@ -15,7 +15,7 @@ declare const ConfigSchema: z.ZodObject<{
     baseURL: string;
     model: string;
     maxSteps: number;
-    contextWindowSize: number;
+    maxRounds: number;
     cdpUrl: string;
     timeoutMs: number;
 }, {
@@ -23,25 +23,31 @@ declare const ConfigSchema: z.ZodObject<{
     baseURL: string;
     model: string;
     maxSteps?: number | undefined;
-    contextWindowSize?: number | undefined;
+    maxRounds?: number | undefined;
     cdpUrl?: string | undefined;
     timeoutMs?: number | undefined;
 }>;
 type Config = z.infer<typeof ConfigSchema>;
 
+/**
+ * Load config with priority: overrides > env vars > config file > defaults
+ */
 declare function loadConfig(overrides?: Partial<Config>): Config;
 
 /**
  * Manages a Playwright CDP connection to the user's Chrome.
- * Lazy-initialized: only connects when first browser tool is called.
+ * Auto-launches Chrome with --remote-debugging-port if not already running.
+ * Syncs user's Chrome profile to preserve cookies/login state.
  */
 declare class BrowserClient {
     private browser;
     private context;
     private _page;
     private cdpUrl;
+    private chromeProcess;
     constructor(cdpUrl: string);
     connect(): Promise<void>;
+    private launchChrome;
     getPage(): Promise<Page>;
     /** Create a new tab and switch to it. */
     newPage(): Promise<Page>;
@@ -49,11 +55,49 @@ declare class BrowserClient {
     get connected(): boolean;
 }
 
+interface StoredScreenshot {
+    id: string;
+    base64: string;
+    mimeType: 'image/png' | 'image/jpeg';
+    label: string;
+}
+/**
+ * Simple in-memory screenshot store.
+ * Screenshot tools save images here with auto-incrementing IDs.
+ * Report content references them via [Image:img_1] markers.
+ */
+declare class ScreenshotStore {
+    private counter;
+    private store;
+    save(base64: string, mimeType: 'image/png' | 'image/jpeg', label: string): string;
+    get(id: string): StoredScreenshot | undefined;
+    listIds(): string[];
+}
+/** A block in parsed report content */
+type ContentBlock = {
+    type: 'text';
+    text: string;
+} | {
+    type: 'image';
+    id: string;
+    base64: string;
+    mimeType: 'image/png' | 'image/jpeg';
+    label: string;
+};
+/**
+ * Parse report content string, expanding [Image:img_X] markers into image blocks.
+ * Returns an array of text and image content blocks.
+ */
+declare function parseReportContent(content: string, store: ScreenshotStore): ContentBlock[];
+/** Strip [Image:...] markers, returning text-only content */
+declare function stripImageMarkers(content: string): string;
 interface ToolContext {
     sessionId: string;
     cdpUrl: string;
     /** Lazy browser client getter — only connects on first call */
     getBrowser: () => Promise<BrowserClient>;
+    /** Screenshot store — tools save screenshots here, report references by [Image:id] */
+    screenshots: ScreenshotStore;
 }
 type ToolResult = {
     type: 'text';
@@ -62,11 +106,11 @@ type ToolResult = {
     type: 'image';
     base64: string;
     mimeType: 'image/png' | 'image/jpeg';
+    screenshotId: string;
 } | {
     type: 'report';
     status: 'completed' | 'blocked' | 'need_guidance';
-    summary: string;
-    screenshot?: string;
+    content: string;
     data?: unknown;
 };
 interface ToolDefinition {
@@ -86,17 +130,15 @@ declare class ToolRegistry {
 
 type Message = OpenAI.Chat.Completions.ChatCompletionMessageParam;
 /**
- * Sliding window message history.
- * Always keeps: system prompt (index 0) + most recent N messages.
+ * Simple message history — stores all messages without windowing.
+ * Small models are cheap, no need to truncate context.
  */
 declare class ContextManager {
     private messages;
-    private readonly maxMessages;
-    constructor(maxMessages: number);
     append(message: Message): void;
-    /** Returns the system prompt + the most recent messages within the window. */
-    getWindow(): Message[];
-    /** Total messages stored (before windowing). */
+    /** Returns all messages. */
+    getMessages(): Message[];
+    /** Total messages stored. */
     get length(): number;
 }
 
@@ -109,11 +151,31 @@ declare class LLMClient {
 
 interface RunResult {
     status: 'completed' | 'blocked' | 'need_guidance';
-    summary: string;
-    screenshot?: string;
+    /** Rich content with [Image:img_X] markers. Use parseReportContent() to expand. */
+    content: string;
     data?: unknown;
     stepsUsed: number;
 }
+type StepEvent = {
+    type: 'thinking';
+    step: number;
+    content: string;
+} | {
+    type: 'tool_call';
+    step: number;
+    name: string;
+    args: unknown;
+} | {
+    type: 'tool_result';
+    step: number;
+    name: string;
+    result: string;
+} | {
+    type: 'error';
+    step: number;
+    message: string;
+};
+type OnStepCallback = (event: StepEvent) => void;
 declare class AgentRunner {
     private llmClient;
     private contextManager;
@@ -121,7 +183,16 @@ declare class AgentRunner {
     private config;
     private toolContext;
     private initialized;
+    private onStep;
+    private roundsUsed;
     constructor(llmClient: LLMClient, contextManager: ContextManager, toolRegistry: ToolRegistry, config: Config, toolContext: ToolContext);
+    /** Register a callback to receive step-by-step progress events */
+    setOnStep(cb: OnStepCallback): void;
+    private emit;
+    /** How many instruction rounds have been used in this session */
+    get currentRound(): number;
+    /** Whether this session has exhausted its max rounds */
+    get roundsExhausted(): boolean;
     run(instruction: string): Promise<RunResult>;
 }
 
@@ -132,6 +203,7 @@ interface Session {
     config: Config;
     runner: AgentRunner;
     browserClient: BrowserClient;
+    screenshots: ScreenshotStore;
     timeoutHandle: ReturnType<typeof setTimeout>;
 }
 declare class SessionRegistry {
@@ -145,4 +217,4 @@ declare class SessionRegistry {
 
 declare function createToolRegistry(): ToolRegistry;
 
-export { AgentRunner, BrowserClient, type Config, ContextManager, LLMClient, type RunResult, type Session, SessionRegistry, type ToolContext, type ToolDefinition, ToolRegistry, type ToolResult, createToolRegistry, loadConfig };
+export { AgentRunner, BrowserClient, type Config, type ContentBlock, ContextManager, LLMClient, type OnStepCallback, type RunResult, ScreenshotStore, type Session, SessionRegistry, type StepEvent, type StoredScreenshot, type ToolContext, type ToolDefinition, ToolRegistry, type ToolResult, createToolRegistry, loadConfig, parseReportContent, stripImageMarkers };