Fix/types

src/audio/audio.ts

@@ -1,11 +1,14 @@
 import { RequestClient } from "../request";
 import { createFileUploadFormData } from "../utils";
-import { SpeechToTextParams, SpeechToTextResponse } from "./interfaces";
+import { SpeechToTextParams, SpeechToTextSyncResponse, SpeechToTextWebhookResponse } from "./interfaces";
 class Audio {
   constructor(private readonly client: RequestClient) {}
-  speech_to_text(params: SpeechToTextParams): Promise<SpeechToTextResponse>;
-  speech_to_text(file: Blob | Buffer, params?: Omit<SpeechToTextParams, "url" | "file_store_key">): Promise<SpeechToTextResponse>;
-  async speech_to_text(params: SpeechToTextParams | Blob | Buffer, options?: SpeechToTextParams): Promise<SpeechToTextResponse> {
+  speech_to_text(params: SpeechToTextParams): Promise<SpeechToTextSyncResponse>;
+  speech_to_text(file: Blob | Buffer, params?: Omit<SpeechToTextParams, "url" | "file_store_key">): Promise<SpeechToTextSyncResponse>;
+  async speech_to_text(
+    params: SpeechToTextParams | Blob | Buffer,
+    options?: SpeechToTextParams
+  ): Promise<SpeechToTextSyncResponse | SpeechToTextWebhookResponse> {
     if (params instanceof Blob || params instanceof Buffer) {
       const formData = createFileUploadFormData(params, options);
       return await this.client.fetchJSS("/ai/transcribe", "POST", formData);

src/audio/interfaces.ts

@@ -11,16 +11,13 @@ export interface SpeechToTextParams {
   chunk_duration?: number;
 }
 
-export interface SpeechToTextResponse extends BaseResponse {
+export interface SpeechToTextSyncResponse extends BaseResponse {
   text: string;
   chunks: Array<{
     timestamp: number[];
     text: string;
   }>;
-  status?: "processing" | "error";
-  id?: string;
   speakers?: {
-    //  available when by_speaker is true
     speaker: string;
     timestamp: number[];
     text: string;

src/core.ts

@@ -29,7 +29,6 @@ export const JigsawStack = (config?: BaseConfig) => {
   };
 
   return {
-    fetch: client.fetchJSS,
     sentiment: general.sentiment,
     translate: general.translate,
     image_generation: general.image_generation,

src/general/index.ts

@@ -5,11 +5,13 @@ import { createFileUploadFormData } from "../utils";
 import {
   EmbeddingParams,
   EmbeddingResponse,
+  ImageGenerationParams,
   PredictionParams,
   PredictionResponse,
   SentimentResponse,
   SummaryParams,
   SummaryResponse,
+  TextToSQLParams,
   TextToSQLResponse,
   TranslateImageParams,
   TranslateParams,
@@ -50,33 +52,12 @@ class General {
     return await this.client.fetchJSS("/ai/sentiment", "POST", params);
   };
 
-  image_generation = async (params: {
-    prompt: string;
-    aspect_ratio?: "1:1" | "16:9" | "21:9" | "3:2" | "2:3" | "4:5" | "5:4" | "3:4" | "4:3" | "9:16" | "9:21";
-    width?: number;
-    height?: number;
-    steps?: number;
-    output_format?: "png" | "svg";
-    return_type?: "url" | "binary" | "base64";
-    advance_config?: {
-      negative_prompt?: string;
-      guidance?: number;
-      seed?: number;
-    };
-    url?: string;
-    file_store_key?: string;
-  }) => {
+  image_generation = async (params: ImageGenerationParams) => {
     const resp = await this.client.fetchJSS("/ai/image_generation", "POST", params);
-
     return respToFileChoice(resp);
   };
 
-  text_to_sql = async (params: {
-    prompt: string;
-    sql_schema?: string;
-    file_store_key?: string;
-    database?: "postgresql" | "mysql" | "sqlite";
-  }): Promise<TextToSQLResponse> => {
+  text_to_sql = async (params: TextToSQLParams): Promise<TextToSQLResponse> => {
     return await this.client.fetchJSS("/ai/sql", "POST", params);
   };
 

src/general/interfaces.ts

@@ -1,5 +1,33 @@
 import { BaseResponse } from "../../types";
 
+export type TextToSQLParams = {
+  prompt: string;
+  database?: "postgresql" | "mysql" | "sqlite";
+  sql_schema?: string;
+  file_store_key?: string;
+};
+
+export interface ImageGenerationParams {
+  prompt: string;
+  aspect_ratio?: "1:1" | "16:9" | "21:9" | "3:2" | "2:3" | "4:5" | "5:4" | "3:4" | "4:3" | "9:16" | "9:21";
+  width?: number;
+  height?: number;
+  steps?: number;
+  output_format?: "png" | "svg";
+  return_type?: "url" | "binary" | "base64";
+  advance_config?: {
+    negative_prompt?: string;
+    guidance?: number;
+    seed?: number;
+  };
+  url?: string;
+  file_store_key?: string;
+}
+
+export interface ImageGenerationResponse extends BaseResponse {
+  url: string;
+}
+
 export interface SentimentResponse extends BaseResponse {
   sentiment: {
     emotion: string;
@@ -24,12 +52,11 @@ export interface TranslateParams {
   text: string | string[];
 }
 
-export interface TranslateImageParams {
-  target_language: string;
+export type TranslateImageParams = {
   url?: string;
-  file_store_key?: string;
+  target_language: string;
   return_type?: "url" | "binary" | "base64";
-}
+};
 
 export interface SpeechToTextResponse extends BaseResponse {
   text: string;

src/helpers.ts

@@ -1,4 +1,10 @@
-export const respToFileChoice = (resp) => {
+export const respToFileChoice = (
+  resp: Response
+): {
+  blob: () => Promise<Blob>;
+  buffer: () => Promise<Buffer>;
+  file: (filename: string, options?: FilePropertyBag) => Promise<File>;
+} => {
   if (!(resp instanceof Response)) {
     return resp;
   }

src/vercel-ai-toolkit.ts

@@ -89,18 +89,11 @@ export class JigsawStackToolSet {
           target_language: z.string().describe('Target language code (e.g., "en", "es", "fr")'),
         }),
         execute: async ({ url, file_store_key, target_language }) => {
-          await this.jigsawStack.translate.image({
+          return await this.jigsawStack.translate.image({
             url,
-            file_store_key,
             target_language,
+            return_type: "url",
           });
-
-          return {
-            success: true,
-            message: `Successfully translated image to ${target_language}`,
-            target_language,
-            note: "Image has been translated. Use jigsawStack.translate.image() directly to get the binary data.",
-          };
         },
       }),
 
@@ -250,8 +243,8 @@ export class JigsawStackToolSet {
           return await this.jigsawStack.vision.vocr({
             prompt,
             url,
-            file_store_key,
             page_range,
+            file_store_key,
           });
         },
       }),
@@ -264,8 +257,8 @@ export class JigsawStackToolSet {
         }),
         execute: async ({ url, file_store_key }) => {
           return await this.jigsawStack.vision.object_detection({
-            url,
             file_store_key,
+            return_type: "url",
           });
         },
       }),

src/vision/interfaces.ts

@@ -1,9 +1,9 @@
-export interface VOCRParams {
+export type VOCRParams = {
   prompt?: string | string[];
   url?: string;
   file_store_key?: string;
   page_range?: Array<number>;
-}
+};
 
 interface Bounds {
   top_left: {
@@ -46,14 +46,14 @@ export interface VOCRResponse {
   page_range?: Array<number>; // only available when page_range is specified
 }
 
-export interface ObjectDetectionParams {
+export type ObjectDetectionParams = {
   url?: string;
   file_store_key?: string;
   prompts?: string[];
   features?: ("object_detection" | "gui")[];
   annotated_image?: boolean;
   return_type?: "url" | "base64";
-}
+};
 
 export interface ObjectDetectionResponse {
   // Optional annotated image - included only if annotated_image=true and objects/gui_elements exist
@@ -73,6 +73,7 @@ interface GuiElement {
 
 interface DetectedObject {
   bounds: BoundingBox;
+  label: string;
   mask?: string; // URL or base64 string depending on return_type - only present for some objects
 }