@juspay/neurolink 8.2.0 → 8.4.0

package/dist/lib/providers/openAI.js

@@ -8,8 +8,6 @@ import { AuthenticationError, InvalidModelError, NetworkError, ProviderError, Ra
 import { DEFAULT_MAX_STEPS } from "../core/constants.js";
 import { validateApiKey, createOpenAIConfig, getProviderModel, } from "../utils/providerConfig.js";
 import { streamAnalyticsCollector } from "../core/streamAnalytics.js";
-import { buildMessagesArray, buildMultimodalMessagesArray, convertToCoreMessages, } from "../utils/messageBuilder.js";
-import { buildMultimodalOptions } from "../utils/multimodalOptionsBuilder.js";
 import { createProxyFetch } from "../proxy/proxyFetch.js";
 import { isZodSchema } from "../utils/schemaConversion.js";
 // Configuration helpers - now using consolidated utility
@@ -246,32 +244,8 @@ export class OpenAIProvider extends BaseProvider {
                 filteredOutTools: Object.keys(allTools).filter((name) => !tools[name]),
             });
             // Build message array from options with multimodal support
-            const hasMultimodalInput = !!(options.input?.images?.length ||
-                options.input?.content?.length ||
-                options.input?.files?.length ||
-                options.input?.csvFiles?.length ||
-                options.input?.pdfFiles?.length);
-            let messages;
-            if (hasMultimodalInput) {
-                logger.debug(`OpenAI: Detected multimodal input, using multimodal message builder`, {
-                    hasImages: !!options.input?.images?.length,
-                    imageCount: options.input?.images?.length || 0,
-                    hasContent: !!options.input?.content?.length,
-                    contentCount: options.input?.content?.length || 0,
-                    hasFiles: !!options.input?.files?.length,
-                    fileCount: options.input?.files?.length || 0,
-                    hasCSVFiles: !!options.input?.csvFiles?.length,
-                    csvFileCount: options.input?.csvFiles?.length || 0,
-                });
-                const multimodalOptions = buildMultimodalOptions(options, this.providerName, this.modelName);
-                const mm = await buildMultimodalMessagesArray(multimodalOptions, this.providerName, this.modelName);
-                // Convert multimodal messages to Vercel AI SDK format (CoreMessage[])
-                messages = convertToCoreMessages(mm);
-            }
-            else {
-                logger.debug(`OpenAI: Text-only input, using standard message builder`);
-                messages = await buildMessagesArray(options);
-            }
+            // Using protected helper from BaseProvider to eliminate code duplication
+            const messages = await this.buildMessagesForStream(options);
             // Debug the actual request being sent to OpenAI
             logger.debug(`OpenAI: streamText request parameters:`, {
                 modelName: this.modelName,

package/dist/lib/providers/openaiCompatible.js

@@ -7,8 +7,6 @@ import { createTimeoutController, TimeoutError } from "../utils/timeout.js";
 import { streamAnalyticsCollector } from "../core/streamAnalytics.js";
 import { createProxyFetch } from "../proxy/proxyFetch.js";
 import { DEFAULT_MAX_STEPS } from "../core/constants.js";
-import { buildMessagesArray, buildMultimodalMessagesArray, convertToCoreMessages, } from "../utils/messageBuilder.js";
-import { buildMultimodalOptions } from "../utils/multimodalOptionsBuilder.js";
 // Constants
 const FALLBACK_OPENAI_COMPATIBLE_MODEL = "gpt-3.5-turbo";
 // Configuration helpers
@@ -161,35 +159,9 @@ export class OpenAICompatibleProvider extends BaseProvider {
         const timeout = this.getTimeout(options);
         const timeoutController = createTimeoutController(timeout, this.providerName, "stream");
         try {
-            // Check for multimodal input (images, PDFs, CSVs, files)
-            const hasMultimodalInput = !!(options.input?.images?.length ||
-                options.input?.content?.length ||
-                options.input?.files?.length ||
-                options.input?.csvFiles?.length ||
-                options.input?.pdfFiles?.length);
-            let messages;
-            if (hasMultimodalInput) {
-                logger.debug(`OpenAI Compatible: Detected multimodal input, using multimodal message builder`, {
-                    hasImages: !!options.input?.images?.length,
-                    imageCount: options.input?.images?.length || 0,
-                    hasContent: !!options.input?.content?.length,
-                    contentCount: options.input?.content?.length || 0,
-                    hasFiles: !!options.input?.files?.length,
-                    fileCount: options.input?.files?.length || 0,
-                    hasCSVFiles: !!options.input?.csvFiles?.length,
-                    csvFileCount: options.input?.csvFiles?.length || 0,
-                    hasPDFFiles: !!options.input?.pdfFiles?.length,
-                    pdfFileCount: options.input?.pdfFiles?.length || 0,
-                });
-                const multimodalOptions = buildMultimodalOptions(options, this.providerName, this.modelName);
-                const mm = await buildMultimodalMessagesArray(multimodalOptions, this.providerName, this.modelName);
-                // Convert multimodal messages to Vercel AI SDK format (CoreMessage[])
-                messages = convertToCoreMessages(mm);
-            }
-            else {
-                logger.debug(`OpenAI Compatible: Text-only input, using standard message builder`);
-                messages = await buildMessagesArray(options);
-            }
+            // Build message array from options with multimodal support
+            // Using protected helper from BaseProvider to eliminate code duplication
+            const messages = await this.buildMessagesForStream(options);
             const model = await this.getAISDKModelWithMiddleware(options); // This is where network connection happens!
             const result = streamText({
                 model,

package/dist/lib/types/content.d.ts

@@ -1,115 +1,18 @@
 /**
  * Content type definitions for multimodal support
- * Supports text and image content with provider-specific formatting
- */
-/**
- * Text content type for multimodal messages
- */
-export type TextContent = {
-    type: "text";
-    text: string;
-};
-/**
- * Image content type for multimodal messages
- */
-export type ImageContent = {
-    type: "image";
-    data: Buffer | string;
-    mediaType?: "image/jpeg" | "image/png" | "image/gif" | "image/webp" | "image/bmp" | "image/tiff";
-    metadata?: {
-        description?: string;
-        quality?: "low" | "high" | "auto";
-        dimensions?: {
-            width: number;
-            height: number;
-        };
-        filename?: string;
-    };
-};
-/**
- * CSV content type for multimodal messages
- */
-export type CSVContent = {
-    type: "csv";
-    data: Buffer | string;
-    metadata?: {
-        filename?: string;
-        maxRows?: number;
-        formatStyle?: "raw" | "markdown" | "json";
-        description?: string;
-    };
-};
-/**
- * PDF document content type for multimodal messages
- */
-export type PDFContent = {
-    type: "pdf";
-    data: Buffer | string;
-    metadata?: {
-        filename?: string;
-        pages?: number;
-        version?: string;
-        description?: string;
-    };
-};
-/**
- * Union type for all content types
- */
-export type Content = TextContent | ImageContent | CSVContent | PDFContent;
-/**
- * Vision capability information for providers
- */
-export type VisionCapability = {
-    provider: string;
-    supportedModels: string[];
-    maxImageSize?: number;
-    supportedFormats: string[];
-    maxImagesPerRequest?: number;
-};
-/**
- * Provider-specific image format requirements
- */
-export type ProviderImageFormat = {
-    provider: string;
-    format: "data_uri" | "base64" | "inline_data" | "source";
-    requiresPrefix?: boolean;
-    mimeTypeField?: string;
-    dataField?: string;
-};
-/**
- * Image processing result
- */
-export type ProcessedImage = {
-    data: string;
-    mediaType: string;
-    size: number;
-    format: "data_uri" | "base64" | "inline_data" | "source";
-};
-/**
- * Multimodal message structure for provider adapters
- */
-export type MultimodalMessage = {
-    role: "user" | "assistant" | "system";
-    content: Content[];
-};
-/**
- * Multimodal input type for options that may contain images or content arrays
- */
-export type MultimodalInput = {
-    text: string;
-    images?: Array<Buffer | string>;
-    content?: Array<TextContent | ImageContent>;
-    csvFiles?: Array<Buffer | string>;
-    pdfFiles?: Array<Buffer | string>;
-    files?: Array<Buffer | string>;
-};
-/**
- * Provider-specific multimodal payload
- */
-export type ProviderMultimodalPayload = {
-    provider: string;
-    model: string;
-    messages?: MultimodalMessage[];
-    contents?: unknown[];
-    [key: string]: unknown;
-};
+ *
+ * @deprecated This file has been reorganized. All multimodal types are now in './multimodal.js'
+ * These re-exports are maintained for backward compatibility.
+ * Please import from './multimodal.js' in new code.
+ *
+ * Migration guide:
+ * ```typescript
+ * // Old (still works)
+ * import type { MultimodalInput } from './types/content.js';
+ *
+ * // New (preferred)
+ * import type { MultimodalInput } from './types/multimodal.js';
+ * ```
+ */
+export type { TextContent, ImageContent, CSVContent, PDFContent, AudioContent, VideoContent, Content, MultimodalInput, MultimodalMessage, VisionCapability, ProviderImageFormat, ProcessedImage, ProviderMultimodalPayload, } from "./multimodal.js";
+export { isTextContent, isImageContent, isCSVContent, isPDFContent, isAudioContent, isVideoContent, isMultimodalInput, } from "./multimodal.js";

package/dist/lib/types/content.js

@@ -1,6 +1,20 @@
 /**
  * Content type definitions for multimodal support
- * Supports text and image content with provider-specific formatting
+ *
+ * @deprecated This file has been reorganized. All multimodal types are now in './multimodal.js'
+ * These re-exports are maintained for backward compatibility.
+ * Please import from './multimodal.js' in new code.
+ *
+ * Migration guide:
+ * ```typescript
+ * // Old (still works)
+ * import type { MultimodalInput } from './types/content.js';
+ *
+ * // New (preferred)
+ * import type { MultimodalInput } from './types/multimodal.js';
+ * ```
  */
-export {};
+// Runtime function re-exports for type guards
+// These MUST be regular exports (not "export type") because they are actual functions
+export { isTextContent, isImageContent, isCSVContent, isPDFContent, isAudioContent, isVideoContent, isMultimodalInput, } from "./multimodal.js";
 //# sourceMappingURL=content.js.map

package/dist/lib/types/conversation.d.ts

@@ -92,24 +92,10 @@ export type ChatMessage = {
     };
 };
 /**
- * Content format for multimodal messages (used internally)
+ * Multimodal message types - Re-exported from multimodal.ts
+ * @deprecated Import from './multimodal.js' instead for better organization
  */
-export type MessageContent = {
-    type: string;
-    text?: string;
-    image?: string;
-    mimeType?: string;
-    [key: string]: unknown;
-};
-/**
- * Extended chat message for multimodal support (internal use)
- */
-export type MultimodalChatMessage = {
-    /** Role of the message sender */
-    role: "user" | "assistant" | "system";
-    /** Content of the message - can be text or multimodal content array */
-    content: string | MessageContent[];
-};
+export type { MessageContent, MultimodalChatMessage } from "./multimodal.js";
 /**
  * Events emitted by conversation memory system
  */

package/dist/lib/types/generateTypes.d.ts

@@ -6,7 +6,7 @@ import type { EvaluationData } from "./evaluation.js";
 import type { ChatMessage, ConversationMemoryConfig } from "./conversation.js";
 import type { MiddlewareFactoryOptions } from "./middlewareTypes.js";
 import type { JsonValue } from "./common.js";
-import type { TextContent, ImageContent } from "./content.js";
+import type { Content } from "./content.js";
 /**
  * Generate function options type - Primary method for content generation
  * Supports multimodal content while maintaining backward compatibility
@@ -18,7 +18,7 @@ export type GenerateOptions = {
         csvFiles?: Array<Buffer | string>;
         pdfFiles?: Array<Buffer | string>;
         files?: Array<Buffer | string>;
-        content?: Array<TextContent | ImageContent>;
+        content?: Content[];
     };
     output?: {
         format?: "text" | "structured" | "json";

package/dist/lib/types/index.d.ts

@@ -2,6 +2,8 @@
  * Centralized type exports for NeuroLink
  */
 export * from "./common.js";
+export { AIProviderName } from "../constants/enums.js";
+export type { ZodUnknownSchema, ValidationSchema, OptionalValidationSchema, StandardRecord, OptionalStandardRecord, } from "./typeAliases.js";
 export * from "./tools.js";
 export * from "./providers.js";
 export * from "./cli.js";

package/dist/lib/types/index.js

@@ -3,6 +3,8 @@
  */
 // Common utility types
 export * from "./common.js";
+// Constants and enums
+export { AIProviderName } from "../constants/enums.js";
 // Tool system types
 export * from "./tools.js";
 // Provider types

package/dist/lib/types/multimodal.d.ts

@@ -0,0 +1,282 @@
+/**
+ * Multimodal Content Types for NeuroLink
+ *
+ * Central registry for all multimodal input/output types.
+ * This file consolidates types from content.ts and conversation.ts
+ * to provide a single source of truth for multimodal functionality.
+ *
+ * @module types/multimodal
+ *
+ * @example Basic Multimodal Input
+ * ```typescript
+ * import type { MultimodalInput } from './types/multimodal.js';
+ *
+ * const input: MultimodalInput = {
+ *   text: "What's in this image?",
+ *   images: [imageBuffer, "https://example.com/image.jpg"],
+ *   pdfFiles: [pdfBuffer]
+ * };
+ * ```
+ *
+ * @example Audio/Video Input (Future)
+ * ```typescript
+ * const avInput: MultimodalInput = {
+ *   text: "Transcribe this audio and analyze this video",
+ *   audioFiles: [audioBuffer],
+ *   videoFiles: ["path/to/video.mp4"]
+ * };
+ * ```
+ *
+ * @example Advanced Content Array
+ * ```typescript
+ * const advanced: MultimodalInput = {
+ *   text: "irrelevant", // ignored when content[] is provided
+ *   content: [
+ *     { type: "text", text: "Analyze these items:" },
+ *     { type: "image", data: imageBuffer, mediaType: "image/jpeg" },
+ *     { type: "pdf", data: pdfBuffer, metadata: { filename: "report.pdf" } }
+ *   ]
+ * };
+ * ```
+ */
+/**
+ * Text content type for multimodal messages
+ */
+export type TextContent = {
+    type: "text";
+    text: string;
+};
+/**
+ * Image content type for multimodal messages
+ */
+export type ImageContent = {
+    type: "image";
+    data: Buffer | string;
+    mediaType?: "image/jpeg" | "image/png" | "image/gif" | "image/webp" | "image/bmp" | "image/tiff";
+    metadata?: {
+        description?: string;
+        quality?: "low" | "high" | "auto";
+        dimensions?: {
+            width: number;
+            height: number;
+        };
+        filename?: string;
+    };
+};
+/**
+ * CSV content type for multimodal messages
+ */
+export type CSVContent = {
+    type: "csv";
+    data: Buffer | string;
+    metadata?: {
+        filename?: string;
+        maxRows?: number;
+        formatStyle?: "raw" | "markdown" | "json";
+        description?: string;
+    };
+};
+/**
+ * PDF document content type for multimodal messages
+ */
+export type PDFContent = {
+    type: "pdf";
+    data: Buffer | string;
+    metadata?: {
+        filename?: string;
+        pages?: number;
+        version?: string;
+        description?: string;
+    };
+};
+/**
+ * Audio content type for multimodal messages
+ *
+ * NOTE: This is for FILE-BASED audio input (not streaming).
+ * For streaming audio (live transcription), use AudioInputSpec from streamTypes.ts
+ *
+ * @example
+ * ```typescript
+ * const audioContent: AudioContent = {
+ *   type: "audio",
+ *   data: audioBuffer,
+ *   mediaType: "audio/mpeg",
+ *   metadata: {
+ *     filename: "recording.mp3",
+ *     duration: 120.5,
+ *     transcription: "Hello world"
+ *   }
+ * };
+ * ```
+ */
+export type AudioContent = {
+    type: "audio";
+    data: Buffer | string;
+    mediaType?: "audio/mpeg" | "audio/wav" | "audio/ogg" | "audio/webm" | "audio/aac" | "audio/flac" | "audio/mp4";
+    metadata?: {
+        filename?: string;
+        duration?: number;
+        sampleRate?: number;
+        channels?: number;
+        transcription?: string;
+        language?: string;
+    };
+};
+/**
+ * Video content type for multimodal messages
+ *
+ * NOTE: This is for FILE-BASED video input.
+ * For streaming video, this type may be extended in future.
+ *
+ * @example
+ * ```typescript
+ * const videoContent: VideoContent = {
+ *   type: "video",
+ *   data: videoBuffer,
+ *   mediaType: "video/mp4",
+ *   metadata: {
+ *     filename: "demo.mp4",
+ *     duration: 300,
+ *     dimensions: { width: 1920, height: 1080 }
+ *   }
+ * };
+ * ```
+ */
+export type VideoContent = {
+    type: "video";
+    data: Buffer | string;
+    mediaType?: "video/mp4" | "video/webm" | "video/ogg" | "video/quicktime" | "video/x-msvideo" | "video/x-matroska";
+    metadata?: {
+        filename?: string;
+        duration?: number;
+        dimensions?: {
+            width: number;
+            height: number;
+        };
+        frameRate?: number;
+        codec?: string;
+        extractedFrames?: string[];
+        transcription?: string;
+    };
+};
+/**
+ * Union type for all content types
+ * Covers text, images, documents, and multimedia
+ */
+export type Content = TextContent | ImageContent | CSVContent | PDFContent | AudioContent | VideoContent;
+/**
+ * Multimodal input type for options that may contain images or content arrays
+ * This is the primary interface for users to provide multimodal content
+ */
+export type MultimodalInput = {
+    text: string;
+    images?: Array<Buffer | string>;
+    content?: Content[];
+    csvFiles?: Array<Buffer | string>;
+    pdfFiles?: Array<Buffer | string>;
+    files?: Array<Buffer | string>;
+    /** Audio files for file-based audio processing (future) */
+    audioFiles?: Array<Buffer | string>;
+    /** Video files for file-based video processing (future) */
+    videoFiles?: Array<Buffer | string>;
+};
+/**
+ * Content format for multimodal messages (used internally)
+ * Compatible with Vercel AI SDK message format
+ */
+export type MessageContent = {
+    type: string;
+    text?: string;
+    image?: string;
+    mimeType?: string;
+    [key: string]: unknown;
+};
+/**
+ * Extended chat message for multimodal support (internal use)
+ * Used during message processing and transformation
+ */
+export type MultimodalChatMessage = {
+    /** Role of the message sender */
+    role: "user" | "assistant" | "system";
+    /** Content of the message - can be text or multimodal content array */
+    content: string | MessageContent[];
+};
+/**
+ * Multimodal message structure for provider adapters
+ */
+export type MultimodalMessage = {
+    role: "user" | "assistant" | "system";
+    content: Content[];
+};
+/**
+ * Vision capability information for providers
+ */
+export type VisionCapability = {
+    provider: string;
+    supportedModels: string[];
+    maxImageSize?: number;
+    supportedFormats: string[];
+    maxImagesPerRequest?: number;
+};
+/**
+ * Provider-specific image format requirements
+ */
+export type ProviderImageFormat = {
+    provider: string;
+    format: "data_uri" | "base64" | "inline_data" | "source";
+    requiresPrefix?: boolean;
+    mimeTypeField?: string;
+    dataField?: string;
+};
+/**
+ * Image processing result
+ */
+export type ProcessedImage = {
+    data: string;
+    mediaType: string;
+    size: number;
+    format: "data_uri" | "base64" | "inline_data" | "source";
+};
+/**
+ * Provider-specific multimodal payload
+ */
+export type ProviderMultimodalPayload = {
+    provider: string;
+    model: string;
+    messages?: MultimodalMessage[];
+    contents?: unknown[];
+    [key: string]: unknown;
+};
+/**
+ * Type guard to check if content is TextContent
+ */
+export declare function isTextContent(content: Content): content is TextContent;
+/**
+ * Type guard to check if content is ImageContent
+ */
+export declare function isImageContent(content: Content): content is ImageContent;
+/**
+ * Type guard to check if content is CSVContent
+ */
+export declare function isCSVContent(content: Content): content is CSVContent;
+/**
+ * Type guard to check if content is PDFContent
+ */
+export declare function isPDFContent(content: Content): content is PDFContent;
+/**
+ * Type guard to check if content is AudioContent
+ */
+export declare function isAudioContent(content: Content): content is AudioContent;
+/**
+ * Type guard to check if content is VideoContent
+ */
+export declare function isVideoContent(content: Content): content is VideoContent;
+/**
+ * Type guard to check if input contains multimodal content
+ * Now includes audio and video detection
+ */
+export declare function isMultimodalInput(input: unknown): input is MultimodalInput;
+/**
+ * Type guard to check if message content is multimodal (array)
+ */
+export declare function isMultimodalMessageContent(content: string | MessageContent[]): content is MessageContent[];