@heyputer/puter.js 2.1.2 → 2.1.6

package/index.d.ts

@@ -9,7 +9,7 @@ declare global {
 declare class Puter {
     // Properties
     appID: string;
-    env: 'app' | 'web' | 'gui';
+    env: 'app' | 'web' | 'gui' | 'nodejs' | 'service-worker';
 
     // Utility methods
     print(text: string, options?: { code?: boolean }): void;
@@ -47,15 +47,27 @@ interface AI {
     chat(messages: ChatMessage[], testMode?: boolean, options?: NonStreamingChatOptions): Promise<ChatResponse>;
 
     img2txt(image: string | File | Blob, testMode?: boolean): Promise<string>;
+    img2txt(image: string | File | Blob, options?: Img2TxtOptions): Promise<string>;
+    img2txt(image: string | File | Blob, testMode?: boolean, options?: Img2TxtOptions): Promise<string>;
+    img2txt(options: Img2TxtOptions): Promise<string>;
 
     txt2img(prompt: string, testMode?: boolean): Promise<HTMLImageElement>;
     txt2img(prompt: string, options?: Txt2ImgOptions): Promise<HTMLImageElement>;
 
+    txt2vid(prompt: string, testMode?: boolean): Promise<HTMLVideoElement>;
+    txt2vid(prompt: string, options?: Txt2VidOptions): Promise<HTMLVideoElement>;
+
     txt2speech(text: string): Promise<HTMLAudioElement>;
     txt2speech(text: string, options?: Txt2SpeechOptions): Promise<HTMLAudioElement>;
     txt2speech(text: string, language?: string): Promise<HTMLAudioElement>;
     txt2speech(text: string, language?: string, voice?: string): Promise<HTMLAudioElement>;
     txt2speech(text: string, language?: string, voice?: string, engine?: string): Promise<HTMLAudioElement>;
+
+    speech2txt(source: string | File | Blob): Promise<string | Speech2TxtResult>;
+    speech2txt(source: string | File | Blob, options?: Speech2TxtOptions): Promise<string | Speech2TxtResult>;
+    speech2txt(options: Speech2TxtOptions): Promise<string | Speech2TxtResult>;
+    speech2txt(source: string | File | Blob, testMode?: boolean): Promise<string | Speech2TxtResult>;
+    speech2txt(source: Speech2TxtOptions, testMode?: boolean): Promise<string | Speech2TxtResult>;
 }
 
 type StreamingChatOptions = Omit<ChatOptions, "stream"> & { stream: true };
@@ -66,6 +78,16 @@ interface ChatOptions {
     stream?: boolean;
     max_tokens?: number;
     temperature?: number;
+    reasoning?: {
+        effort?: 'none' | 'low' | 'medium' | 'high' | 'minimal';
+        [key: string]: unknown;
+    };
+    reasoning_effort?: 'none' | 'low' | 'medium' | 'high' | 'minimal';
+    text?: {
+        verbosity?: 'low' | 'medium' | 'high';
+        [key: string]: unknown;
+    };
+    verbosity?: 'low' | 'medium' | 'high';
     tools?: ToolDefinition[];
 }
 
@@ -108,16 +130,83 @@ interface ToolCall {
 }
 
 interface Txt2ImgOptions {
-    model?: 'gpt-image-1' | 'gemini-2.5-flash-image-preview' | 'dall-e-3';
+    model?: 'gpt-image-1' | 'gpt-image-1-mini' | 'gemini-2.5-flash-image-preview' | 'dall-e-3';
     quality?: 'high' | 'medium' | 'low' | 'hd' | 'standard';
     input_image?: string;
     input_image_mime_type?: string;
 }
 
+interface Txt2VidOptions {
+    prompt?: string;
+    model?: string;
+    duration?: number;
+    seconds?: number;
+    size?: string;
+    resolution?: string;
+    width?: number;
+    height?: number;
+    fps?: number;
+    steps?: number;
+    guidance_scale?: number;
+    seed?: number;
+    output_format?: string;
+    output_quality?: number;
+    negative_prompt?: string;
+    reference_images?: string[];
+    frame_images?: Array<Record<string, unknown>>;
+    metadata?: Record<string, unknown>;
+    provider?: string;
+    service?: string;
+    driver?: string;
+    test_mode?: boolean;
+}
+
+interface Img2TxtOptions {
+    source?: string | File | Blob;
+    provider?: 'aws-textract' | 'mistral';
+    model?: string;
+    pages?: number[];
+    includeImageBase64?: boolean;
+    imageLimit?: number;
+    imageMinSize?: number;
+    bboxAnnotationFormat?: Record<string, unknown>;
+    documentAnnotationFormat?: Record<string, unknown>;
+    testMode?: boolean;
+}
+
 interface Txt2SpeechOptions {
     language?: string;
     voice?: string;
-    engine?: 'standard' | 'neural' | 'generative';
+    engine?: 'standard' | 'neural' | 'long-form' | 'generative' | string;
+    provider?: 'aws-polly' | 'openai' | string;
+    model?: 'gpt-4o-mini-tts' | 'tts-1' | 'tts-1-hd' | string;
+    response_format?: 'mp3' | 'opus' | 'aac' | 'flac' | 'wav' | 'pcm' | string;
+    instructions?: string;
+}
+
+interface Speech2TxtOptions {
+    file?: string | File | Blob;
+    audio?: string | File | Blob;
+    model?: 'gpt-4o-mini-transcribe' | 'gpt-4o-transcribe' | 'gpt-4o-transcribe-diarize' | 'whisper-1' | string;
+    response_format?: 'json' | 'text' | 'diarized_json' | 'srt' | 'verbose_json' | 'vtt' | string;
+    language?: string;
+    prompt?: string;
+    temperature?: number;
+    logprobs?: boolean;
+    timestamp_granularities?: string[];
+    translate?: boolean;
+    stream?: boolean;
+    chunking_strategy?: string;
+    known_speaker_names?: string[];
+    known_speaker_references?: string[];
+    extra_body?: Record<string, unknown>;
+}
+
+interface Speech2TxtResult {
+    text?: string;
+    language?: string;
+    segments?: Array<Record<string, unknown>>;
+    [key: string]: any;
 }
 
 interface ChatResponseChunk {
@@ -146,6 +235,7 @@ interface CreateAppOptions {
     icon?: string;
     maximizeOnStart?: boolean;
     filetypeAssociations?: string[];
+    dedupeName?: boolean;
 }
 
 interface GetAppOptions {
@@ -188,6 +278,8 @@ interface Auth {
     signOut(): void;
     isSignedIn(): boolean;
     getUser(): Promise<User>;
+    getMonthlyUsage(): Promise<MonthlyUsage>;
+    getDetailedAppUsage(appId: string): Promise<DetailedAppUsage>;
 }
 
 interface User {
@@ -196,6 +288,33 @@ interface User {
     email_confirmed: boolean;
 }
 
+interface AllowanceInfo {
+    monthUsageAllowance: number;
+    remaining: number;
+}
+
+interface AppUsage {
+    count: number;
+    total: number;
+}
+
+interface APIUsage {
+    cost: number;
+    count: number;
+    units: number;
+}
+
+interface MonthlyUsage {
+    allowanceInfo: AllowanceInfo;
+    appTotals: Record<string, AppUsage>;
+    usage: Record<string, APIUsage>;
+}
+
+interface DetailedAppUsage {
+    total: number;
+    [key: string]: APIUsage;
+}
+
 // Drivers Module
 interface Drivers {
     call(interface: string, driver: string, method: string, args?: object): Promise<any>;
@@ -214,7 +333,7 @@ interface FileSystem {
     rename(path: string, newName: string): Promise<FSItem>;
     space(): Promise<SpaceInfo>;
     stat(path: string): Promise<FSItem>;
-    upload(items: FileList | File[] | Blob[], dirPath?: string, options?: object): Promise<FSItem[]>;
+    upload(items: FileList | File[] | Blob[], dirPath?: string, options?: UploadOptions): Promise<FSItem[]>;
     write(path: string, data?: string | File | Blob, options?: WriteOptions): Promise<FSItem>;
 }
 
@@ -255,6 +374,12 @@ interface WriteOptions {
     createMissingParents?: boolean;
 }
 
+interface UploadOptions {
+    overwrite?: boolean;
+    dedupeName?: boolean;
+    name?: string;
+}
+
 interface SpaceInfo {
     capacity: number;
     used: number;
@@ -511,4 +636,3 @@ export {
     WorkerExecOptions,
     WorkerInfo, Workers, WriteOptions
 };
-

package/package.json

@@ -1,6 +1,6 @@
 {
     "name": "@heyputer/puter.js",
-    "version": "2.1.2",
+    "version": "2.1.6",
     "description": "Puter.js - A JavaScript library for interacting with Puter services.",
     "main": "src/index.js",
     "types": "index.d.ts",

package/src/index.js

@@ -785,6 +785,43 @@ export const puter =  puterInit();
 export default puter;
 globalThis.puter = puter;
 
+puter.tools = [];
+/**
+ * @type {{messageTarget: Window}}
+ */
+const puterParent = puter.ui.parentApp();
+globalThis.puterParent = puterParent;
+if (puterParent) {
+    console.log("I have a parent, registering tools")
+    puterParent.on('message', async (event) => {
+        console.log("Got tool req ", event)
+        if (event.$ === "requestTools") {
+            console.log("Responding with tools")
+            puterParent.postMessage({
+                $: "providedTools",
+                tools: JSON.parse(JSON.stringify(puter.tools))
+            });
+        }
+
+        if (event.$ === "executeTool") {
+            console.log("xecuting tools")
+            /**
+             * Puter tools format
+             * @type {[{exec: Function, function: {description: string, name: string, parameters: {properties: any, required: Array<string>}, type: string}}]}
+             */
+            const [tool] = puter.tools.filter(e=>e.function.name === event.toolName);
+
+            const response = await tool.exec(event.parameters);
+            puterParent.postMessage({
+                $: "toolResponse",
+                response,
+                tag: event.tag,
+            });
+        }
+    });
+    puterParent.postMessage({$: "ready"});
+}
+
 globalThis.addEventListener && globalThis.addEventListener('message', async (event) => {
     // if the message is not from Puter, then ignore it
     if ( event.origin !== puter.defaultGUIOrigin ) return;