@oagi/oagi 0.1.4 → 0.2.0

package/dist/index.d.cts

@@ -22,6 +22,7 @@ declare const ActionSchema: z.ZodObject<{
         type: "type";
         scroll: "scroll";
         finish: "finish";
+        fail: "fail";
         wait: "wait";
         call_user: "call_user";
     }>;
@@ -72,7 +73,7 @@ interface Todo {
     /**
      * Current status of the todo
      */
-    status: 'pending' | 'in_progress' | 'completed' | 'blocked';
+    status: 'pending' | 'in_progress' | 'completed' | 'skipped';
     /**
      * Summary of execution for this todo
      */
@@ -266,11 +267,11 @@ declare const ActionEventSchema: z.ZodObject<{
     type: z.ZodLiteral<"action">;
     step_num: z.ZodNumber;
     actions: z.ZodArray<z.ZodCustom<{
-        type: "click" | "left_double" | "left_triple" | "right_single" | "drag" | "hotkey" | "type" | "scroll" | "finish" | "wait" | "call_user";
+        type: "click" | "left_double" | "left_triple" | "right_single" | "drag" | "hotkey" | "type" | "scroll" | "finish" | "fail" | "wait" | "call_user";
         argument: string;
         count: number;
     }, {
-        type: "click" | "left_double" | "left_triple" | "right_single" | "drag" | "hotkey" | "type" | "scroll" | "finish" | "wait" | "call_user";
+        type: "click" | "left_double" | "left_triple" | "right_single" | "drag" | "hotkey" | "type" | "scroll" | "finish" | "fail" | "wait" | "call_user";
         argument: string;
         count: number;
     }>>;
@@ -416,23 +417,6 @@ declare class DefaultAgent implements Agent {
     execute(instruction: string, action_handler: ActionHandler, image_provider: ImageProvider): Promise<boolean>;
 }
 
-/**
- * -----------------------------------------------------------------------------
- *  Copyright (c) OpenAGI Foundation
- *  All rights reserved.
- *
- *  This file is part of the official API project.
- *  Licensed under the MIT License.
- * -----------------------------------------------------------------------------
- */
-
-interface Agent {
-    /**
-     * Protocol for synchronous task execution agents.
-     */
-    execute(instruction: string, action_handler: ActionHandler, image_provider: ImageProvider): Promise<boolean>;
-}
-
 /**
  * -----------------------------------------------------------------------------
  *  Copyright (c) OpenAGI Foundation
@@ -515,6 +499,170 @@ declare class Client {
     callWorker({ workerId, overallTodo, taskDescription, todos, history, currentTodoIndex, taskExecutionSummary, currentScreenshot, currentSubtaskInstruction, windowSteps, windowScreenshots, resultScreenshot, priorNotes, latestTodoSummary, apiVersion, }: GenerateOption): Promise<GenerateResponse>;
 }
 
+/**
+ * -----------------------------------------------------------------------------
+ *  Copyright (c) OpenAGI Foundation
+ *  All rights reserved.
+ *
+ *  This file is part of the official API project.
+ *  Licensed under the MIT License.
+ * -----------------------------------------------------------------------------
+ */
+
+type ImageLike = ArrayBuffer | string;
+type TodoStatus = Todo['status'];
+interface TodoItem {
+    description: string;
+    status: TodoStatus;
+}
+interface TaskerAction {
+    timestamp: string;
+    action_type: string;
+    target?: string | null;
+    details?: Record<string, unknown>;
+    reasoning?: string | null;
+    result?: string | null;
+    screenshot_uuid?: string | null;
+}
+interface TodoHistory {
+    todo_index: number;
+    todo: string;
+    actions: TaskerAction[];
+    summary?: string;
+    completed: boolean;
+}
+interface PlannerOutput {
+    instruction: string;
+    reasoning: string;
+    subtodos: string[];
+}
+interface ReflectionOutput {
+    continue_current: boolean;
+    new_instruction?: string | null;
+    reasoning: string;
+    success_assessment: boolean;
+}
+type PlannerContext = Record<string, unknown>;
+declare class PlannerMemory {
+    taskDescription: string;
+    todos: TodoItem[];
+    history: TodoHistory[];
+    taskExecutionSummary: string;
+    todoExecutionSummaries: Record<number, string>;
+    setTask(taskDescription: string, todos: string[] | TodoItem[]): void;
+    getCurrentTodo(): {
+        todo: TodoItem;
+        index: number;
+    } | null;
+    updateTodo(index: number, status: TodoStatus, summary?: string): void;
+    addHistory(todoIndex: number, actions: TaskerAction[], summary?: string, completed?: boolean): void;
+    getContext(): PlannerContext;
+    getTodoStatusSummary(): Record<string, number>;
+    appendTodo(description: string): void;
+}
+declare class Planner {
+    private apiKey?;
+    private baseUrl?;
+    private client?;
+    private ownsClient;
+    constructor(client?: Client, apiKey?: string | undefined, baseUrl?: string | undefined);
+    private ensureClient;
+    getClient(): Client;
+    close(): Promise<void>;
+    private extractMemoryData;
+    private extractJsonString;
+    private parsePlannerOutput;
+    private parseReflectionOutput;
+    private formatExecutionNotes;
+    private ensureScreenshotUuid;
+    initialPlan(todo: string, context: PlannerContext, screenshot?: ImageLike, memory?: PlannerMemory, todoIndex?: number): Promise<{
+        output: PlannerOutput;
+        requestId?: string | null;
+    }>;
+    reflect(actions: TaskerAction[], context: PlannerContext, screenshot?: ImageLike, memory?: PlannerMemory, todoIndex?: number, currentInstruction?: string, reflectionInterval?: number): Promise<{
+        output: ReflectionOutput;
+        requestId?: string | null;
+    }>;
+    summarize(_executionHistory: TaskerAction[], context: PlannerContext, memory?: PlannerMemory, todoIndex?: number): Promise<{
+        summary: string;
+        requestId?: string | null;
+    }>;
+}
+declare class TaskerAgent implements Agent {
+    /** Hierarchical agent that manages multi-todo workflows. */
+    private apiKey?;
+    private baseUrl?;
+    private model;
+    private maxSteps;
+    private temperature?;
+    private reflectionInterval;
+    private planner;
+    private stepObserver?;
+    private stepDelay;
+    private memory;
+    private currentTaskeeAgent?;
+    constructor(apiKey?: string, baseUrl?: string, model?: string, maxSteps?: number, temperature?: number | undefined, reflectionInterval?: number, planner?: Planner, stepObserver?: StepObserver, stepDelay?: number);
+    setTask(task: string, todos: string[]): void;
+    set_task(task: string, todos: string[]): void;
+    execute(_instruction: string, actionHandler: ActionHandler, imageProvider: ImageProvider): Promise<boolean>;
+    private prepare;
+    private executeTodo;
+    private updateMemoryFromExecution;
+    private updateTaskSummary;
+    getMemory(): PlannerMemory;
+    appendTodo(description: string): void;
+}
+
+/**
+ * -----------------------------------------------------------------------------
+ *  Copyright (c) OpenAGI Foundation
+ *  All rights reserved.
+ *
+ *  This file is part of the official API project.
+ *  Licensed under the MIT License.
+ * -----------------------------------------------------------------------------
+ */
+
+declare enum ExportFormat {
+    /** Supported export formats. */
+    MARKDOWN = "markdown",
+    HTML = "html",
+    JSON = "json"
+}
+declare class AsyncAgentObserver extends StepObserver {
+    /**
+     * Records agent execution events and exports to various formats.
+     *
+     * This class implements the AsyncObserver protocol and provides
+     * functionality for recording events during agent execution and
+     * exporting them to Markdown or HTML formats.
+     */
+    events: ObserverEvent[];
+    onEvent(event: ObserverEvent): Promise<void>;
+    addLog(message: string): void;
+    addSplit(label?: string): void;
+    clear(): void;
+    getEventsByStep(step_num: number): ObserverEvent[];
+    export(format: ExportFormat | string, path: string, images_dir?: string | null): void;
+}
+
+/**
+ * -----------------------------------------------------------------------------
+ *  Copyright (c) OpenAGI Foundation
+ *  All rights reserved.
+ *
+ *  This file is part of the official API project.
+ *  Licensed under the MIT License.
+ * -----------------------------------------------------------------------------
+ */
+
+interface Agent {
+    /**
+     * Protocol for synchronous task execution agents.
+     */
+    execute(instruction: string, action_handler: ActionHandler, image_provider: ImageProvider): Promise<boolean>;
+}
+
 /**
  * -----------------------------------------------------------------------------
  *  Copyright (c) OpenAGI Foundation
@@ -571,4 +719,4 @@ declare class DefaultActionHandler implements ActionHandler {
     handle(actions: Action[]): Promise<void>;
 }
 
-export { APIError, type Action, Actor, type Agent, AuthenticationError, Client, type ClientOptions, ConfigurationError, DefaultActionHandler, DefaultAgent, type ErrorDetail, type ErrorResponse, type GenerateResponse, NetworkError, NotFoundError, OAGIError, RateLimitError, RequestTimeoutError, ScreenshotMaker, ServerError, type Step, type UploadFileResponse, ValidationError };
+export { APIError, type Action, Actor, type Agent, AsyncAgentObserver, AuthenticationError, Client, type ClientOptions, ConfigurationError, DefaultActionHandler, DefaultAgent, type ErrorDetail, type ErrorResponse, ExportFormat, type GenerateResponse, NetworkError, NotFoundError, OAGIError, RateLimitError, RequestTimeoutError, ScreenshotMaker, ServerError, type Step, TaskerAgent, type UploadFileResponse, ValidationError };

package/dist/index.d.ts

@@ -22,6 +22,7 @@ declare const ActionSchema: z.ZodObject<{
         type: "type";
         scroll: "scroll";
         finish: "finish";
+        fail: "fail";
         wait: "wait";
         call_user: "call_user";
     }>;
@@ -72,7 +73,7 @@ interface Todo {
     /**
      * Current status of the todo
      */
-    status: 'pending' | 'in_progress' | 'completed' | 'blocked';
+    status: 'pending' | 'in_progress' | 'completed' | 'skipped';
     /**
      * Summary of execution for this todo
      */
@@ -266,11 +267,11 @@ declare const ActionEventSchema: z.ZodObject<{
     type: z.ZodLiteral<"action">;
     step_num: z.ZodNumber;
     actions: z.ZodArray<z.ZodCustom<{
-        type: "click" | "left_double" | "left_triple" | "right_single" | "drag" | "hotkey" | "type" | "scroll" | "finish" | "wait" | "call_user";
+        type: "click" | "left_double" | "left_triple" | "right_single" | "drag" | "hotkey" | "type" | "scroll" | "finish" | "fail" | "wait" | "call_user";
         argument: string;
         count: number;
     }, {
-        type: "click" | "left_double" | "left_triple" | "right_single" | "drag" | "hotkey" | "type" | "scroll" | "finish" | "wait" | "call_user";
+        type: "click" | "left_double" | "left_triple" | "right_single" | "drag" | "hotkey" | "type" | "scroll" | "finish" | "fail" | "wait" | "call_user";
         argument: string;
         count: number;
     }>>;
@@ -416,23 +417,6 @@ declare class DefaultAgent implements Agent {
     execute(instruction: string, action_handler: ActionHandler, image_provider: ImageProvider): Promise<boolean>;
 }
 
-/**
- * -----------------------------------------------------------------------------
- *  Copyright (c) OpenAGI Foundation
- *  All rights reserved.
- *
- *  This file is part of the official API project.
- *  Licensed under the MIT License.
- * -----------------------------------------------------------------------------
- */
-
-interface Agent {
-    /**
-     * Protocol for synchronous task execution agents.
-     */
-    execute(instruction: string, action_handler: ActionHandler, image_provider: ImageProvider): Promise<boolean>;
-}
-
 /**
  * -----------------------------------------------------------------------------
  *  Copyright (c) OpenAGI Foundation
@@ -515,6 +499,170 @@ declare class Client {
     callWorker({ workerId, overallTodo, taskDescription, todos, history, currentTodoIndex, taskExecutionSummary, currentScreenshot, currentSubtaskInstruction, windowSteps, windowScreenshots, resultScreenshot, priorNotes, latestTodoSummary, apiVersion, }: GenerateOption): Promise<GenerateResponse>;
 }
 
+/**
+ * -----------------------------------------------------------------------------
+ *  Copyright (c) OpenAGI Foundation
+ *  All rights reserved.
+ *
+ *  This file is part of the official API project.
+ *  Licensed under the MIT License.
+ * -----------------------------------------------------------------------------
+ */
+
+type ImageLike = ArrayBuffer | string;
+type TodoStatus = Todo['status'];
+interface TodoItem {
+    description: string;
+    status: TodoStatus;
+}
+interface TaskerAction {
+    timestamp: string;
+    action_type: string;
+    target?: string | null;
+    details?: Record<string, unknown>;
+    reasoning?: string | null;
+    result?: string | null;
+    screenshot_uuid?: string | null;
+}
+interface TodoHistory {
+    todo_index: number;
+    todo: string;
+    actions: TaskerAction[];
+    summary?: string;
+    completed: boolean;
+}
+interface PlannerOutput {
+    instruction: string;
+    reasoning: string;
+    subtodos: string[];
+}
+interface ReflectionOutput {
+    continue_current: boolean;
+    new_instruction?: string | null;
+    reasoning: string;
+    success_assessment: boolean;
+}
+type PlannerContext = Record<string, unknown>;
+declare class PlannerMemory {
+    taskDescription: string;
+    todos: TodoItem[];
+    history: TodoHistory[];
+    taskExecutionSummary: string;
+    todoExecutionSummaries: Record<number, string>;
+    setTask(taskDescription: string, todos: string[] | TodoItem[]): void;
+    getCurrentTodo(): {
+        todo: TodoItem;
+        index: number;
+    } | null;
+    updateTodo(index: number, status: TodoStatus, summary?: string): void;
+    addHistory(todoIndex: number, actions: TaskerAction[], summary?: string, completed?: boolean): void;
+    getContext(): PlannerContext;
+    getTodoStatusSummary(): Record<string, number>;
+    appendTodo(description: string): void;
+}
+declare class Planner {
+    private apiKey?;
+    private baseUrl?;
+    private client?;
+    private ownsClient;
+    constructor(client?: Client, apiKey?: string | undefined, baseUrl?: string | undefined);
+    private ensureClient;
+    getClient(): Client;
+    close(): Promise<void>;
+    private extractMemoryData;
+    private extractJsonString;
+    private parsePlannerOutput;
+    private parseReflectionOutput;
+    private formatExecutionNotes;
+    private ensureScreenshotUuid;
+    initialPlan(todo: string, context: PlannerContext, screenshot?: ImageLike, memory?: PlannerMemory, todoIndex?: number): Promise<{
+        output: PlannerOutput;
+        requestId?: string | null;
+    }>;
+    reflect(actions: TaskerAction[], context: PlannerContext, screenshot?: ImageLike, memory?: PlannerMemory, todoIndex?: number, currentInstruction?: string, reflectionInterval?: number): Promise<{
+        output: ReflectionOutput;
+        requestId?: string | null;
+    }>;
+    summarize(_executionHistory: TaskerAction[], context: PlannerContext, memory?: PlannerMemory, todoIndex?: number): Promise<{
+        summary: string;
+        requestId?: string | null;
+    }>;
+}
+declare class TaskerAgent implements Agent {
+    /** Hierarchical agent that manages multi-todo workflows. */
+    private apiKey?;
+    private baseUrl?;
+    private model;
+    private maxSteps;
+    private temperature?;
+    private reflectionInterval;
+    private planner;
+    private stepObserver?;
+    private stepDelay;
+    private memory;
+    private currentTaskeeAgent?;
+    constructor(apiKey?: string, baseUrl?: string, model?: string, maxSteps?: number, temperature?: number | undefined, reflectionInterval?: number, planner?: Planner, stepObserver?: StepObserver, stepDelay?: number);
+    setTask(task: string, todos: string[]): void;
+    set_task(task: string, todos: string[]): void;
+    execute(_instruction: string, actionHandler: ActionHandler, imageProvider: ImageProvider): Promise<boolean>;
+    private prepare;
+    private executeTodo;
+    private updateMemoryFromExecution;
+    private updateTaskSummary;
+    getMemory(): PlannerMemory;
+    appendTodo(description: string): void;
+}
+
+/**
+ * -----------------------------------------------------------------------------
+ *  Copyright (c) OpenAGI Foundation
+ *  All rights reserved.
+ *
+ *  This file is part of the official API project.
+ *  Licensed under the MIT License.
+ * -----------------------------------------------------------------------------
+ */
+
+declare enum ExportFormat {
+    /** Supported export formats. */
+    MARKDOWN = "markdown",
+    HTML = "html",
+    JSON = "json"
+}
+declare class AsyncAgentObserver extends StepObserver {
+    /**
+     * Records agent execution events and exports to various formats.
+     *
+     * This class implements the AsyncObserver protocol and provides
+     * functionality for recording events during agent execution and
+     * exporting them to Markdown or HTML formats.
+     */
+    events: ObserverEvent[];
+    onEvent(event: ObserverEvent): Promise<void>;
+    addLog(message: string): void;
+    addSplit(label?: string): void;
+    clear(): void;
+    getEventsByStep(step_num: number): ObserverEvent[];
+    export(format: ExportFormat | string, path: string, images_dir?: string | null): void;
+}
+
+/**
+ * -----------------------------------------------------------------------------
+ *  Copyright (c) OpenAGI Foundation
+ *  All rights reserved.
+ *
+ *  This file is part of the official API project.
+ *  Licensed under the MIT License.
+ * -----------------------------------------------------------------------------
+ */
+
+interface Agent {
+    /**
+     * Protocol for synchronous task execution agents.
+     */
+    execute(instruction: string, action_handler: ActionHandler, image_provider: ImageProvider): Promise<boolean>;
+}
+
 /**
  * -----------------------------------------------------------------------------
  *  Copyright (c) OpenAGI Foundation
@@ -571,4 +719,4 @@ declare class DefaultActionHandler implements ActionHandler {
     handle(actions: Action[]): Promise<void>;
 }
 
-export { APIError, type Action, Actor, type Agent, AuthenticationError, Client, type ClientOptions, ConfigurationError, DefaultActionHandler, DefaultAgent, type ErrorDetail, type ErrorResponse, type GenerateResponse, NetworkError, NotFoundError, OAGIError, RateLimitError, RequestTimeoutError, ScreenshotMaker, ServerError, type Step, type UploadFileResponse, ValidationError };
+export { APIError, type Action, Actor, type Agent, AsyncAgentObserver, AuthenticationError, Client, type ClientOptions, ConfigurationError, DefaultActionHandler, DefaultAgent, type ErrorDetail, type ErrorResponse, ExportFormat, type GenerateResponse, NetworkError, NotFoundError, OAGIError, RateLimitError, RequestTimeoutError, ScreenshotMaker, ServerError, type Step, TaskerAgent, type UploadFileResponse, ValidationError };

package/dist/index.js

@@ -2,11 +2,13 @@
 import {
   APIError,
   Actor,
+  AsyncAgentObserver,
   AuthenticationError,
   Client,
   ConfigurationError,
   DefaultActionHandler,
   DefaultAgent,
+  ExportFormat,
   NetworkError,
   NotFoundError,
   OAGIError,
@@ -14,16 +16,19 @@ import {
   RequestTimeoutError,
   ScreenshotMaker,
   ServerError,
+  TaskerAgent,
   ValidationError
-} from "./chunk-JVNTVY6W.js";
+} from "./chunk-LA6CBP44.js";
 export {
   APIError,
   Actor,
+  AsyncAgentObserver,
   AuthenticationError,
   Client,
   ConfigurationError,
   DefaultActionHandler,
   DefaultAgent,
+  ExportFormat,
   NetworkError,
   NotFoundError,
   OAGIError,
@@ -31,6 +36,7 @@ export {
   RequestTimeoutError,
   ScreenshotMaker,
   ServerError,
+  TaskerAgent,
   ValidationError
 };
 //# sourceMappingURL=index.js.map

package/dist/index.ts

@@ -0,0 +1,23 @@
+/**
+ * -----------------------------------------------------------------------------
+ *  Copyright (c) OpenAGI Foundation
+ *  All rights reserved.
+ *
+ *  This file is part of the official API project.
+ *  Licensed under the MIT License.
+ * -----------------------------------------------------------------------------
+ */
+
+export type {
+  ActionEvent,
+  BaseEvent,
+  ImageEvent,
+  LogEvent,
+  ObserverEvent,
+  PlanEvent,
+  SplitEvent,
+  StepEvent,
+} from '../../types/index.js';
+
+export { AsyncAgentObserver, ExportFormat } from './agent_observer.js';
+export { exportToHtml, exportToJson, exportToMarkdown } from './exporters.js';

package/dist/protocol.ts

@@ -0,0 +1,14 @@
+/**
+ * -----------------------------------------------------------------------------
+ *  Copyright (c) OpenAGI Foundation
+ *  All rights reserved.
+ *
+ *  This file is part of the official API project.
+ *  Licensed under the MIT License.
+ * -----------------------------------------------------------------------------
+ */
+
+// Re-export from types for convenience
+import type { AsyncObserver } from '../../types/index.js';
+
+export type { AsyncObserver };