@juspay/neurolink 9.2.0 → 9.4.0

package/dist/lib/providers/anthropic.js

@@ -1,13 +1,13 @@
 import { createAnthropic } from "@ai-sdk/anthropic";
 import { streamText } from "ai";
-import { AIProviderName, AnthropicModels } from "../constants/enums.js";
+import { AnthropicModels } from "../constants/enums.js";
 import { BaseProvider } from "../core/baseProvider.js";
-import { logger } from "../utils/logger.js";
-import { createTimeoutController, TimeoutError } from "../utils/timeout.js";
-import { AuthenticationError, NetworkError, ProviderError, RateLimitError, } from "../types/errors.js";
 import { DEFAULT_MAX_STEPS } from "../core/constants.js";
-import { validateApiKey, createAnthropicConfig, getProviderModel, } from "../utils/providerConfig.js";
 import { createProxyFetch } from "../proxy/proxyFetch.js";
+import { AuthenticationError, NetworkError, ProviderError, RateLimitError, } from "../types/errors.js";
+import { logger } from "../utils/logger.js";
+import { createAnthropicConfig, getProviderModel, validateApiKey, } from "../utils/providerConfig.js";
+import { createTimeoutController, TimeoutError } from "../utils/timeout.js";
 // Configuration helpers - now using consolidated utility
 const getAnthropicApiKey = () => {
     return validateApiKey(createAnthropicConfig());
@@ -88,9 +88,12 @@ export class AnthropicProvider extends BaseProvider {
         const timeout = this.getTimeout(options);
         const timeoutController = createTimeoutController(timeout, this.providerName, "stream");
         try {
-            // ✅ Get tools for streaming (same as generate method)
+            // Get tools - options.tools is pre-merged by BaseProvider.stream() with
+            // base tools (MCP/built-in) + user-provided tools (RAG, etc.)
             const shouldUseTools = !options.disableTools && this.supportsTools();
-            const tools = shouldUseTools ? await this.getAllTools() : {};
+            const tools = shouldUseTools
+                ? options.tools || (await this.getAllTools())
+                : {};
             // Build message array from options with multimodal support
             // Using protected helper from BaseProvider to eliminate code duplication
             const messages = await this.buildMessagesForStream(options);

package/dist/lib/providers/googleAiStudio.d.ts

@@ -1,9 +1,9 @@
-import { type Schema, type LanguageModelV1 } from "ai";
-import type { ZodUnknownSchema } from "../types/typeAliases.js";
-import { AIProviderName } from "../constants/enums.js";
-import type { StreamOptions, StreamResult } from "../types/streamTypes.js";
-import type { TextGenerationOptions, EnhancedGenerateResult } from "../types/generateTypes.js";
+import { type LanguageModelV1, type Schema } from "ai";
+import { type AIProviderName } from "../constants/enums.js";
 import { BaseProvider } from "../core/baseProvider.js";
+import type { EnhancedGenerateResult, TextGenerationOptions } from "../types/generateTypes.js";
+import type { StreamOptions, StreamResult } from "../types/streamTypes.js";
+import type { ZodUnknownSchema } from "../types/typeAliases.js";
 /**
  * Google AI Studio provider implementation using BaseProvider
  * Migrated from original GoogleAIStudio class to new factory pattern

package/dist/lib/providers/googleAiStudio.js

@@ -1,16 +1,16 @@
 import { createGoogleGenerativeAI } from "@ai-sdk/google";
 import { streamText } from "ai";
-import { AIProviderName, GoogleAIModels, ErrorCategory, ErrorSeverity, } from "../constants/enums.js";
-import { NeuroLinkError, ERROR_CODES } from "../utils/errorHandling.js";
+import { ErrorCategory, ErrorSeverity, GoogleAIModels, } from "../constants/enums.js";
 import { BaseProvider } from "../core/baseProvider.js";
-import { logger } from "../utils/logger.js";
-import { createTimeoutController, TimeoutError } from "../utils/timeout.js";
-import { AuthenticationError, NetworkError, ProviderError, RateLimitError, } from "../types/errors.js";
 import { DEFAULT_MAX_STEPS, DEFAULT_TOOL_MAX_RETRIES, } from "../core/constants.js";
 import { streamAnalyticsCollector } from "../core/streamAnalytics.js";
+import { AuthenticationError, NetworkError, ProviderError, RateLimitError, } from "../types/errors.js";
+import { ERROR_CODES, NeuroLinkError } from "../utils/errorHandling.js";
+import { logger } from "../utils/logger.js";
 import { isGemini3Model } from "../utils/modelDetection.js";
 import { convertZodToJsonSchema, inlineJsonSchema, isZodSchema, } from "../utils/schemaConversion.js";
 import { createNativeThinkingConfig } from "../utils/thinkingConfig.js";
+import { createTimeoutController, TimeoutError } from "../utils/timeout.js";
 // Google AI Live API types now imported from ../types/providerSpecific.js
 // Import proper types for multimodal message handling
 // Create Google GenAI client
@@ -424,9 +424,12 @@ export class GoogleAIStudioProvider extends BaseProvider {
         const timeout = this.getTimeout(options);
         const timeoutController = createTimeoutController(timeout, this.providerName, "stream");
         try {
-            // Get tools consistently with generate method
+            // Get tools consistently with generate method (include user-provided RAG tools)
             const shouldUseTools = !options.disableTools && this.supportsTools();
-            const tools = shouldUseTools ? await this.getAllTools() : {};
+            const baseTools = shouldUseTools ? await this.getAllTools() : {};
+            const tools = shouldUseTools
+                ? { ...baseTools, ...(options.tools || {}) }
+                : {};
             // Build message array from options with multimodal support
             // Using protected helper from BaseProvider to eliminate code duplication
             const messages = await this.buildMessagesForStream(options);

package/dist/lib/providers/googleVertex.d.ts

@@ -1,9 +1,9 @@
+import { type LanguageModel, type LanguageModelV1, type Schema } from "ai";
 import type { ZodType, ZodTypeDef } from "zod";
-import { type Schema, type LanguageModelV1, type LanguageModel } from "ai";
-import { AIProviderName } from "../constants/enums.js";
-import type { StreamOptions, StreamResult } from "../types/streamTypes.js";
-import type { TextGenerationOptions, EnhancedGenerateResult } from "../types/generateTypes.js";
+import { type AIProviderName } from "../constants/enums.js";
 import { BaseProvider } from "../core/baseProvider.js";
+import type { EnhancedGenerateResult, TextGenerationOptions } from "../types/generateTypes.js";
+import type { StreamOptions, StreamResult } from "../types/streamTypes.js";
 /**
  * Google Vertex AI Provider v2 - BaseProvider Implementation
  *
@@ -64,6 +64,11 @@ export declare class GoogleVertexProvider extends BaseProvider {
     constructor(modelName?: string, _providerName?: string, sdk?: unknown, region?: string);
     protected getProviderName(): AIProviderName;
     protected getDefaultModel(): string;
+    /**
+     * Get the default embedding model for Google Vertex
+     * @returns The default Vertex AI embedding model name
+     */
+    protected getDefaultEmbeddingModel(): string;
     /**
      * Returns the Vercel AI SDK model instance for Google Vertex
      * Creates fresh model instances for each request
@@ -237,6 +242,13 @@ export declare class GoogleVertexProvider extends BaseProvider {
      * @returns A promise that resolves to the generation result, including the image data.
      */
     protected executeImageGeneration(options: TextGenerationOptions): Promise<EnhancedGenerateResult>;
+    /**
+     * Generate embeddings for text using Google Vertex AI text-embedding models
+     * @param text - The text to embed
+     * @param modelName - The embedding model to use (default: text-embedding-004)
+     * @returns Promise resolving to the embedding vector
+     */
+    embed(text: string, modelName?: string): Promise<number[]>;
     /**
      * Get model suggestions when a model is not found
      */

package/dist/lib/providers/googleVertex.js

@@ -1,24 +1,24 @@
 import { createVertex, } from "@ai-sdk/google-vertex";
 import { createVertexAnthropic, } from "@ai-sdk/google-vertex/anthropic";
-import { streamText, Output, } from "ai";
-import { AIProviderName, ErrorCategory, ErrorSeverity, } from "../constants/enums.js";
-import { NeuroLinkError, ERROR_CODES } from "../utils/errorHandling.js";
+import { Output, streamText, } from "ai";
+import dns from "dns";
+import fs from "fs";
+import os from "os";
+import path from "path";
+import { ErrorCategory, ErrorSeverity, } from "../constants/enums.js";
 import { BaseProvider } from "../core/baseProvider.js";
-import { logger } from "../utils/logger.js";
-import { createTimeoutController, TimeoutError } from "../utils/timeout.js";
-import { AuthenticationError, ProviderError } from "../types/errors.js";
-import { DEFAULT_MAX_STEPS, GLOBAL_LOCATION_MODELS, DEFAULT_TOOL_MAX_RETRIES, } from "../core/constants.js";
+import { DEFAULT_MAX_STEPS, DEFAULT_TOOL_MAX_RETRIES, GLOBAL_LOCATION_MODELS, } from "../core/constants.js";
 import { ModelConfigurationManager } from "../core/modelConfiguration.js";
-import { validateApiKey, createVertexProjectConfig, createGoogleAuthConfig, } from "../utils/providerConfig.js";
+import { createProxyFetch } from "../proxy/proxyFetch.js";
+import { AuthenticationError, ProviderError } from "../types/errors.js";
+import { ERROR_CODES, NeuroLinkError } from "../utils/errorHandling.js";
+import { FileDetector } from "../utils/fileDetector.js";
+import { logger } from "../utils/logger.js";
 import { isGemini3Model } from "../utils/modelDetection.js";
+import { createGoogleAuthConfig, createVertexProjectConfig, validateApiKey, } from "../utils/providerConfig.js";
 import { convertZodToJsonSchema, inlineJsonSchema, } from "../utils/schemaConversion.js";
 import { createNativeThinkingConfig } from "../utils/thinkingConfig.js";
-import fs from "fs";
-import path from "path";
-import os from "os";
-import dns from "dns";
-import { createProxyFetch } from "../proxy/proxyFetch.js";
-import { FileDetector } from "../utils/fileDetector.js";
+import { createTimeoutController, TimeoutError } from "../utils/timeout.js";
 // Import proper types for multimodal message handling
 // Enhanced Anthropic support with direct imports
 // Using the dual provider architecture from Vercel AI SDK
@@ -339,6 +339,15 @@ export class GoogleVertexProvider extends BaseProvider {
     getDefaultModel() {
         return getDefaultVertexModel();
     }
+    /**
+     * Get the default embedding model for Google Vertex
+     * @returns The default Vertex AI embedding model name
+     */
+    getDefaultEmbeddingModel() {
+        return (process.env.VERTEX_EMBEDDING_MODEL ||
+            process.env.GOOGLE_EMBEDDING_MODEL ||
+            "text-embedding-004");
+    }
     /**
      * Returns the Vercel AI SDK model instance for Google Vertex
      * Creates fresh model instances for each request
@@ -702,11 +711,16 @@ export class GoogleVertexProvider extends BaseProvider {
             // 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!
-            // Get all available tools (direct + MCP + external) for streaming
+            // Get all available tools (direct + MCP + external + user-provided RAG tools) for streaming
             const shouldUseTools = !options.disableTools && this.supportsTools();
-            const tools = shouldUseTools ? await this.getAllTools() : {};
+            const baseStreamTools = shouldUseTools ? await this.getAllTools() : {};
+            const tools = shouldUseTools
+                ? { ...baseStreamTools, ...(options.tools || {}) }
+                : {};
             logger.debug(`${functionTag}: Tools for streaming`, {
                 shouldUseTools,
+                baseToolCount: Object.keys(baseStreamTools).length,
+                externalToolCount: Object.keys(options.tools || {}).length,
                 toolCount: Object.keys(tools).length,
                 toolNames: Object.keys(tools),
             });
@@ -2747,6 +2761,48 @@ export class GoogleVertexProvider extends BaseProvider {
             throw this.handleProviderError(error);
         }
     }
+    /**
+     * Generate embeddings for text using Google Vertex AI text-embedding models
+     * @param text - The text to embed
+     * @param modelName - The embedding model to use (default: text-embedding-004)
+     * @returns Promise resolving to the embedding vector
+     */
+    async embed(text, modelName) {
+        const embeddingModelName = modelName || "text-embedding-004";
+        logger.debug("Generating embedding", {
+            provider: this.providerName,
+            model: embeddingModelName,
+            textLength: text.length,
+        });
+        try {
+            // Create embedding model using the AI SDK
+            const { embed } = await import("ai");
+            // Create the Vertex provider with current settings
+            const vertexSettings = await createVertexSettings(this.location);
+            const vertex = createVertex(vertexSettings);
+            // Get the text embedding model
+            const embeddingModel = vertex.textEmbeddingModel(embeddingModelName);
+            // Generate the embedding
+            const result = await embed({
+                model: embeddingModel,
+                value: text,
+            });
+            logger.debug("Embedding generated successfully", {
+                provider: this.providerName,
+                model: embeddingModelName,
+                embeddingDimension: result.embedding.length,
+            });
+            return result.embedding;
+        }
+        catch (error) {
+            logger.error("Embedding generation failed", {
+                error: error instanceof Error ? error.message : String(error),
+                model: embeddingModelName,
+                textLength: text.length,
+            });
+            throw this.handleProviderError(error);
+        }
+    }
     /**
      * Get model suggestions when a model is not found
      */

package/dist/lib/providers/litellm.d.ts

@@ -1,8 +1,8 @@
+import { type LanguageModelV1, type Schema } from "ai";
 import type { ZodType, ZodTypeDef } from "zod";
-import { type Schema, type LanguageModelV1 } from "ai";
-import { AIProviderName } from "../constants/enums.js";
-import type { StreamOptions, StreamResult } from "../types/streamTypes.js";
+import type { AIProviderName } from "../constants/enums.js";
 import { BaseProvider } from "../core/baseProvider.js";
+import type { StreamOptions, StreamResult } from "../types/streamTypes.js";
 /**
  * LiteLLM Provider - BaseProvider Implementation
  * Provides access to 100+ models via LiteLLM proxy server

package/dist/lib/providers/litellm.js

@@ -1,13 +1,12 @@
 import { createOpenAI } from "@ai-sdk/openai";
-import { streamText, Output } from "ai";
-import { AIProviderName } from "../constants/enums.js";
+import { Output, streamText, } from "ai";
 import { BaseProvider } from "../core/baseProvider.js";
-import { logger } from "../utils/logger.js";
-import { createTimeoutController, TimeoutError } from "../utils/timeout.js";
-import { getProviderModel } from "../utils/providerConfig.js";
-import { streamAnalyticsCollector } from "../core/streamAnalytics.js";
 import { DEFAULT_MAX_STEPS } from "../core/constants.js";
+import { streamAnalyticsCollector } from "../core/streamAnalytics.js";
 import { createProxyFetch } from "../proxy/proxyFetch.js";
+import { logger } from "../utils/logger.js";
+import { getProviderModel } from "../utils/providerConfig.js";
+import { createTimeoutController, TimeoutError } from "../utils/timeout.js";
 // Configuration helpers
 const getLiteLLMConfig = () => {
     return {
@@ -126,9 +125,11 @@ export class LiteLLMProvider extends BaseProvider {
             // 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!
-            // Get all available tools (direct + MCP + external) for streaming - matching Vertex pattern
+            // Get tools - options.tools is pre-merged by BaseProvider.stream()
             const shouldUseTools = !options.disableTools && this.supportsTools();
-            const tools = shouldUseTools ? await this.getAllTools() : {};
+            const tools = shouldUseTools
+                ? options.tools || (await this.getAllTools())
+                : {};
             logger.debug(`LiteLLM: Tools for streaming`, {
                 shouldUseTools,
                 toolCount: Object.keys(tools).length,
@@ -219,8 +220,7 @@ export class LiteLLMProvider extends BaseProvider {
                                 errorType: errorChunk.type,
                                 errorDetails: errorChunk.error,
                             });
-                            throw new Error(`LiteLLM streaming error: ${errorChunk.error?.message ||
-                                "Unknown error"}`);
+                            throw new Error(`LiteLLM streaming error: ${errorChunk.error?.message || "Unknown error"}`);
                         }
                         if ("textDelta" in chunk) {
                             // Text delta from fullStream

package/dist/lib/providers/mistral.d.ts

@@ -1,8 +1,8 @@
 import { type LanguageModelV1 } from "ai";
-import type { ValidationSchema } from "../types/typeAliases.js";
-import { AIProviderName } from "../constants/enums.js";
-import type { StreamOptions, StreamResult } from "../types/streamTypes.js";
+import type { AIProviderName } from "../constants/enums.js";
 import { BaseProvider } from "../core/baseProvider.js";
+import type { StreamOptions, StreamResult } from "../types/streamTypes.js";
+import type { ValidationSchema } from "../types/typeAliases.js";
 /**
  * Mistral AI Provider v2 - BaseProvider Implementation
  * Supports official AI-SDK integration with all Mistral models

package/dist/lib/providers/mistral.js

@@ -1,13 +1,12 @@
 import { createMistral } from "@ai-sdk/mistral";
 import { streamText } from "ai";
-import { AIProviderName } from "../constants/enums.js";
 import { BaseProvider } from "../core/baseProvider.js";
-import { logger } from "../utils/logger.js";
-import { createTimeoutController, TimeoutError } from "../utils/timeout.js";
 import { DEFAULT_MAX_STEPS } from "../core/constants.js";
-import { validateApiKey, createMistralConfig, getProviderModel, } from "../utils/providerConfig.js";
 import { streamAnalyticsCollector } from "../core/streamAnalytics.js";
 import { createProxyFetch } from "../proxy/proxyFetch.js";
+import { logger } from "../utils/logger.js";
+import { createMistralConfig, getProviderModel, validateApiKey, } from "../utils/providerConfig.js";
+import { createTimeoutController, TimeoutError } from "../utils/timeout.js";
 // Configuration helpers - now using consolidated utility
 const getMistralApiKey = () => {
     return validateApiKey(createMistralConfig());
@@ -47,9 +46,11 @@ export class MistralProvider extends BaseProvider {
         const timeout = this.getTimeout(options);
         const timeoutController = createTimeoutController(timeout, this.providerName, "stream");
         try {
-            // Get tools consistently with generate method
+            // Get tools - options.tools is pre-merged by BaseProvider.stream()
             const shouldUseTools = !options.disableTools && this.supportsTools();
-            const tools = shouldUseTools ? await this.getAllTools() : {};
+            const tools = shouldUseTools
+                ? options.tools || (await this.getAllTools())
+                : {};
             // Build message array from options with multimodal support
             // Using protected helper from BaseProvider to eliminate code duplication
             const messages = await this.buildMessagesForStream(options);

package/dist/lib/providers/ollama.d.ts

@@ -1,9 +1,8 @@
-import { AIProviderName } from "../constants/enums.js";
-import type { LanguageModelV1 } from "ai";
+import type { LanguageModelV1, Schema } from "ai";
+import type { AIProviderName } from "../constants/enums.js";
+import { BaseProvider } from "../core/baseProvider.js";
 import type { StreamOptions, StreamResult } from "../types/streamTypes.js";
 import type { ZodUnknownSchema } from "../types/typeAliases.js";
-import type { Schema } from "ai";
-import { BaseProvider } from "../core/baseProvider.js";
 /**
  * Ollama Provider v2 - BaseProvider Implementation
  *

package/dist/lib/providers/ollama.js

@@ -1,13 +1,12 @@
-import { AIProviderName } from "../constants/enums.js";
+import { createAnalytics } from "../core/analytics.js";
 import { BaseProvider } from "../core/baseProvider.js";
-import { logger } from "../utils/logger.js";
+import { DEFAULT_MAX_STEPS } from "../core/constants.js";
 import { modelConfig } from "../core/modelConfiguration.js";
 import { createProxyFetch } from "../proxy/proxyFetch.js";
-import { TimeoutError } from "../utils/timeout.js";
+import { logger } from "../utils/logger.js";
 import { buildMultimodalMessagesArray } from "../utils/messageBuilder.js";
 import { buildMultimodalOptions } from "../utils/multimodalOptionsBuilder.js";
-import { DEFAULT_MAX_STEPS } from "../core/constants.js";
-import { createAnalytics } from "../core/analytics.js";
+import { TimeoutError } from "../utils/timeout.js";
 // Model version constants (configurable via environment)
 const DEFAULT_OLLAMA_MODEL = "llama3.1:8b";
 const FALLBACK_OLLAMA_MODEL = "llama3.2:latest"; // Used when primary model fails
@@ -608,7 +607,9 @@ export class OllamaProvider extends BaseProvider {
         const maxIterations = options.maxSteps || DEFAULT_MAX_STEPS;
         let iteration = 0;
         // Get all available tools (direct + MCP + external)
-        const allTools = await this.getAllTools();
+        // BaseProvider.stream() pre-merges base tools + external tools into options.tools
+        const allTools = options.tools ||
+            (await this.getAllTools());
         // Convert tools to Ollama format
         const ollamaTools = this.convertToolsToOllamaFormat(allTools);
         // Validate that PDFs are not provided
@@ -698,7 +699,6 @@ export class OllamaProvider extends BaseProvider {
                             };
                             conversationHistory.push(toolMessage);
                             iteration++;
-                            continue; // Next iteration
                         }
                         else if (finishReason === "length") {
                             // Max tokens reached, continue conversation
@@ -708,7 +708,6 @@ export class OllamaProvider extends BaseProvider {
                                 content: "Please continue.",
                             });
                             iteration++;
-                            continue;
                         }
                         else {
                             // Unknown finish reason, end conversation

package/dist/lib/providers/openAI.d.ts

@@ -1,9 +1,9 @@
 import { type LanguageModelV1 } from "ai";
-import type { ValidationSchema } from "../types/typeAliases.js";
 import { AIProviderName } from "../constants/enums.js";
-import type { StreamOptions, StreamResult } from "../types/streamTypes.js";
 import { BaseProvider } from "../core/baseProvider.js";
 import type { NeuroLink } from "../neurolink.js";
+import type { StreamOptions, StreamResult } from "../types/streamTypes.js";
+import type { ValidationSchema } from "../types/typeAliases.js";
 /**
  * OpenAI Provider v2 - BaseProvider Implementation
  * Migrated to use factory pattern with exact Google AI provider pattern
@@ -17,6 +17,11 @@ export declare class OpenAIProvider extends BaseProvider {
     supportsTools(): boolean;
     getProviderName(): AIProviderName;
     getDefaultModel(): string;
+    /**
+     * Get the default embedding model for OpenAI
+     * @returns The default OpenAI embedding model name
+     */
+    protected getDefaultEmbeddingModel(): string;
     /**
      * Returns the Vercel AI SDK model instance for OpenAI
      */
@@ -47,5 +52,12 @@ export declare class OpenAIProvider extends BaseProvider {
      * and the migration guide in the project repository.
      */
     protected executeStream(options: StreamOptions, _analysisSchema?: ValidationSchema): Promise<StreamResult>;
+    /**
+     * Generate embeddings for text using OpenAI text-embedding models
+     * @param text - The text to embed
+     * @param modelName - The embedding model to use (default: text-embedding-3-small)
+     * @returns Promise resolving to the embedding vector
+     */
+    embed(text: string, modelName?: string): Promise<number[]>;
 }
 export default OpenAIProvider;

package/dist/lib/providers/openAI.js

@@ -2,14 +2,14 @@ import { createOpenAI } from "@ai-sdk/openai";
 import { streamText } from "ai";
 import { AIProviderName } from "../constants/enums.js";
 import { BaseProvider } from "../core/baseProvider.js";
-import { logger } from "../utils/logger.js";
-import { createTimeoutController, TimeoutError } from "../utils/timeout.js";
-import { AuthenticationError, InvalidModelError, NetworkError, ProviderError, RateLimitError, } from "../types/errors.js";
 import { DEFAULT_MAX_STEPS } from "../core/constants.js";
-import { validateApiKey, createOpenAIConfig, getProviderModel, } from "../utils/providerConfig.js";
 import { streamAnalyticsCollector } from "../core/streamAnalytics.js";
 import { createProxyFetch } from "../proxy/proxyFetch.js";
+import { AuthenticationError, InvalidModelError, NetworkError, ProviderError, RateLimitError, } from "../types/errors.js";
+import { logger } from "../utils/logger.js";
+import { createOpenAIConfig, getProviderModel, validateApiKey, } from "../utils/providerConfig.js";
 import { isZodSchema } from "../utils/schemaConversion.js";
+import { createTimeoutController, TimeoutError } from "../utils/timeout.js";
 // Configuration helpers - now using consolidated utility
 const getOpenAIApiKey = () => {
     return validateApiKey(createOpenAIConfig());
@@ -54,6 +54,13 @@ export class OpenAIProvider extends BaseProvider {
     getDefaultModel() {
         return getOpenAIModel();
     }
+    /**
+     * Get the default embedding model for OpenAI
+     * @returns The default OpenAI embedding model name
+     */
+    getDefaultEmbeddingModel() {
+        return process.env.OPENAI_EMBEDDING_MODEL || "text-embedding-3-small";
+    }
     /**
      * Returns the Vercel AI SDK model instance for OpenAI
      */
@@ -218,9 +225,12 @@ export class OpenAIProvider extends BaseProvider {
         const timeout = this.getTimeout(options);
         const timeoutController = createTimeoutController(timeout, this.providerName, "stream");
         try {
-            // Get tools consistently with generate method
+            // Get tools - options.tools is pre-merged by BaseProvider.stream() with
+            // base tools (MCP/built-in) + user-provided tools (RAG, etc.)
             const shouldUseTools = !options.disableTools && this.supportsTools();
-            const allTools = shouldUseTools ? await this.getAllTools() : {};
+            const allTools = shouldUseTools
+                ? options.tools || (await this.getAllTools())
+                : {};
             // OpenAI-specific fix: Validate tools format and filter out problematic ones
             let tools = this.validateAndFilterToolsForOpenAI(allTools);
             // OpenAI max tools limit - configurable via environment variable
@@ -446,6 +456,50 @@ export class OpenAIProvider extends BaseProvider {
             throw this.handleProviderError(error);
         }
     }
+    /**
+     * Generate embeddings for text using OpenAI text-embedding models
+     * @param text - The text to embed
+     * @param modelName - The embedding model to use (default: text-embedding-3-small)
+     * @returns Promise resolving to the embedding vector
+     */
+    async embed(text, modelName) {
+        const embeddingModelName = modelName || "text-embedding-3-small";
+        logger.debug("Generating embedding", {
+            provider: this.providerName,
+            model: embeddingModelName,
+            textLength: text.length,
+        });
+        try {
+            // Create embedding model using the AI SDK
+            const { embed } = await import("ai");
+            // Create the OpenAI provider
+            const openai = createOpenAI({
+                apiKey: getOpenAIApiKey(),
+                fetch: createProxyFetch(),
+            });
+            // Get the text embedding model
+            const embeddingModel = openai.textEmbeddingModel(embeddingModelName);
+            // Generate the embedding
+            const result = await embed({
+                model: embeddingModel,
+                value: text,
+            });
+            logger.debug("Embedding generated successfully", {
+                provider: this.providerName,
+                model: embeddingModelName,
+                embeddingDimension: result.embedding.length,
+            });
+            return result.embedding;
+        }
+        catch (error) {
+            logger.error("Embedding generation failed", {
+                error: error instanceof Error ? error.message : String(error),
+                model: embeddingModelName,
+                textLength: text.length,
+            });
+            throw this.handleProviderError(error);
+        }
+    }
 }
 // Export for factory registration
 export default OpenAIProvider;

package/dist/lib/providers/openRouter.d.ts

@@ -1,8 +1,8 @@
+import { type LanguageModelV1, type Schema } from "ai";
 import type { ZodType, ZodTypeDef } from "zod";
-import { type Schema, type LanguageModelV1 } from "ai";
 import { AIProviderName } from "../constants/enums.js";
-import type { StreamOptions, StreamResult } from "../types/streamTypes.js";
 import { BaseProvider } from "../core/baseProvider.js";
+import type { StreamOptions, StreamResult } from "../types/streamTypes.js";
 /**
  * OpenRouter Provider - BaseProvider Implementation
  * Provides access to 300+ models from 60+ providers via OpenRouter unified gateway

package/dist/lib/providers/openRouter.js

@@ -1,13 +1,13 @@
 import { createOpenRouter } from "@openrouter/ai-sdk-provider";
-import { streamText, Output } from "ai";
+import { Output, streamText } from "ai";
 import { AIProviderName } from "../constants/enums.js";
 import { BaseProvider } from "../core/baseProvider.js";
-import { logger } from "../utils/logger.js";
-import { createTimeoutController, TimeoutError } from "../utils/timeout.js";
-import { getProviderModel } from "../utils/providerConfig.js";
-import { streamAnalyticsCollector } from "../core/streamAnalytics.js";
 import { DEFAULT_MAX_STEPS } from "../core/constants.js";
+import { streamAnalyticsCollector } from "../core/streamAnalytics.js";
 import { createProxyFetch } from "../proxy/proxyFetch.js";
+import { logger } from "../utils/logger.js";
+import { getProviderModel } from "../utils/providerConfig.js";
+import { createTimeoutController, TimeoutError } from "../utils/timeout.js";
 // Constants
 const MODELS_DISCOVERY_TIMEOUT_MS = 5000; // 5 seconds for model discovery
 // Configuration helpers
@@ -203,8 +203,12 @@ export class OpenRouterProvider extends BaseProvider {
             const messages = await this.buildMessagesForStream(options);
             const model = await this.getAISDKModelWithMiddleware(options);
             // Get all available tools (direct + MCP + external) for streaming
+            // BaseProvider.stream() pre-merges base tools + external tools into options.tools
             const shouldUseTools = !options.disableTools && this.supportsTools();
-            const tools = shouldUseTools ? await this.getAllTools() : {};
+            const tools = shouldUseTools
+                ? options.tools ||
+                    (await this.getAllTools())
+                : {};
             logger.debug(`OpenRouter: Tools for streaming`, {
                 shouldUseTools,
                 toolCount: Object.keys(tools).length,