diff --git a/packages/ai-providers/server-ai-langchain/__tests__/LangChainProvider.test.ts b/packages/ai-providers/server-ai-langchain/__tests__/LangChainProvider.test.ts
index 7568eff456..b6a6a5f8e7 100644
--- a/packages/ai-providers/server-ai-langchain/__tests__/LangChainProvider.test.ts
+++ b/packages/ai-providers/server-ai-langchain/__tests__/LangChainProvider.test.ts
@@ -155,6 +155,66 @@ describe('LangChainProvider', () => {
       expect(result.message.content).toBe('');
       expect(mockLogger.warn).toHaveBeenCalledTimes(1);
     });
+
+    it('returns success=false when model invocation throws an error', async () => {
+      const error = new Error('Model invocation failed');
+      mockLLM.invoke.mockRejectedValue(error);
+
+      const messages = [{ role: 'user' as const, content: 'Hello' }];
+      const result = await provider.invokeModel(messages);
+
+      expect(result.metrics.success).toBe(false);
+      expect(result.message.content).toBe('');
+      expect(result.message.role).toBe('assistant');
+      expect(mockLogger.error).toHaveBeenCalledWith('LangChain model invocation failed:', error);
+    });
+  });
+
+  describe('invokeStructuredModel', () => {
+    let mockLLM: any;
+    let provider: LangChainProvider;
+
+    beforeEach(() => {
+      mockLLM = {
+        withStructuredOutput: jest.fn(),
+      };
+      provider = new LangChainProvider(mockLLM, mockLogger);
+      jest.clearAllMocks();
+    });
+
+    it('returns success=true for successful invocation', async () => {
+      const mockResponse = { result: 'structured data' };
+      const mockInvoke = jest.fn().mockResolvedValue(mockResponse);
+      mockLLM.withStructuredOutput.mockReturnValue({ invoke: mockInvoke });
+
+      const messages = [{ role: 'user' as const, content: 'Hello' }];
+      const responseStructure = { type: 'object', properties: {} };
+      const result = await provider.invokeStructuredModel(messages, responseStructure);
+
+      expect(result.metrics.success).toBe(true);
+      expect(result.data).toEqual(mockResponse);
+      expect(result.rawResponse).toBe(JSON.stringify(mockResponse));
+      expect(mockLogger.error).not.toHaveBeenCalled();
+    });
+
+    it('returns success=false when structured model invocation throws an error', async () => {
+      const error = new Error('Structured invocation failed');
+      const mockInvoke = jest.fn().mockRejectedValue(error);
+      mockLLM.withStructuredOutput.mockReturnValue({ invoke: mockInvoke });
+
+      const messages = [{ role: 'user' as const, content: 'Hello' }];
+      const responseStructure = { type: 'object', properties: {} };
+      const result = await provider.invokeStructuredModel(messages, responseStructure);
+
+      expect(result.metrics.success).toBe(false);
+      expect(result.data).toEqual({});
+      expect(result.rawResponse).toBe('');
+      expect(result.metrics.usage).toEqual({ total: 0, input: 0, output: 0 });
+      expect(mockLogger.error).toHaveBeenCalledWith(
+        'LangChain structured model invocation failed:',
+        error,
+      );
+    });
   });
 
   describe('mapProvider', () => {
diff --git a/packages/ai-providers/server-ai-langchain/package.json b/packages/ai-providers/server-ai-langchain/package.json
index 9c7c975e47..5cb5425c8e 100644
--- a/packages/ai-providers/server-ai-langchain/package.json
+++ b/packages/ai-providers/server-ai-langchain/package.json
@@ -28,7 +28,7 @@
   "license": "Apache-2.0",
   "devDependencies": {
     "@langchain/core": "^0.3.0",
-    "@launchdarkly/server-sdk-ai": "^0.12.3",
+    "@launchdarkly/server-sdk-ai": "^0.12.0",
     "@trivago/prettier-plugin-sort-imports": "^4.1.1",
     "@types/jest": "^29.5.3",
     "@typescript-eslint/eslint-plugin": "^6.20.0",
@@ -48,7 +48,7 @@
   },
   "peerDependencies": {
     "@langchain/core": "^0.2.0 || ^0.3.0",
-    "@launchdarkly/server-sdk-ai": "^0.12.2",
+    "@launchdarkly/server-sdk-ai": "^0.12.0",
     "langchain": "^0.2.0 || ^0.3.0"
   }
 }
diff --git a/packages/ai-providers/server-ai-langchain/src/LangChainProvider.ts b/packages/ai-providers/server-ai-langchain/src/LangChainProvider.ts
index 01df295384..20f7a929e1 100644
--- a/packages/ai-providers/server-ai-langchain/src/LangChainProvider.ts
+++ b/packages/ai-providers/server-ai-langchain/src/LangChainProvider.ts
@@ -10,6 +10,7 @@ import type {
   LDLogger,
   LDMessage,
   LDTokenUsage,
+  StructuredResponse,
 } from '@launchdarkly/server-sdk-ai';
 
 /**
@@ -44,39 +45,102 @@ export class LangChainProvider extends AIProvider {
    * Invoke the LangChain model with an array of messages.
    */
   async invokeModel(messages: LDMessage[]): Promise<ChatResponse> {
-    // Convert LDMessage[] to LangChain messages
-    const langchainMessages = LangChainProvider.convertMessagesToLangChain(messages);
-
-    // Get the LangChain response
-    const response: AIMessage = await this._llm.invoke(langchainMessages);
-
-    // Generate metrics early (assumes success by default)
-    const metrics = LangChainProvider.createAIMetrics(response);
-
-    // Extract text content from the response
-    let content: string = '';
-    if (typeof response.content === 'string') {
-      content = response.content;
-    } else {
-      // Log warning for non-string content (likely multimodal)
-      this.logger?.warn(
-        `Multimodal response not supported, expecting a string. Content type: ${typeof response.content}, Content:`,
-        JSON.stringify(response.content, null, 2),
-      );
-      // Update metrics to reflect content loss
-      metrics.success = false;
+    try {
+      // Convert LDMessage[] to LangChain messages
+      const langchainMessages = LangChainProvider.convertMessagesToLangChain(messages);
+
+      // Get the LangChain response
+      const response: AIMessage = await this._llm.invoke(langchainMessages);
+
+      // Generate metrics early (assumes success by default)
+      const metrics = LangChainProvider.createAIMetrics(response);
+
+      // Extract text content from the response
+      let content: string = '';
+      if (typeof response.content === 'string') {
+        content = response.content;
+      } else {
+        // Log warning for non-string content (likely multimodal)
+        this.logger?.warn(
+          `Multimodal response not supported, expecting a string. Content type: ${typeof response.content}, Content:`,
+          JSON.stringify(response.content, null, 2),
+        );
+        // Update metrics to reflect content loss
+        metrics.success = false;
+      }
+
+      // Create the assistant message
+      const assistantMessage: LDMessage = {
+        role: 'assistant',
+        content,
+      };
+
+      return {
+        message: assistantMessage,
+        metrics,
+      };
+    } catch (error) {
+      this.logger?.warn('LangChain model invocation failed:', error);
+
+      return {
+        message: {
+          role: 'assistant',
+          content: '',
+        },
+        metrics: {
+          success: false,
+        },
+      };
     }
+  }
 
-    // Create the assistant message
-    const assistantMessage: LDMessage = {
-      role: 'assistant',
-      content,
-    };
+  /**
+   * Invoke the LangChain model with structured output support.
+   */
+  async invokeStructuredModel(
+    messages: LDMessage[],
+    responseStructure: Record<string, unknown>,
+  ): Promise<StructuredResponse> {
+    try {
+      // Convert LDMessage[] to LangChain messages
+      const langchainMessages = LangChainProvider.convertMessagesToLangChain(messages);
+
+      // Get the LangChain response
+      const response = await this._llm
+        .withStructuredOutput(responseStructure)
+        .invoke(langchainMessages);
+
+      // Using structured output doesn't support metrics
+      const metrics = {
+        success: true,
+        usage: {
+          total: 0,
+          input: 0,
+          output: 0,
+        },
+      };
 
-    return {
-      message: assistantMessage,
-      metrics,
-    };
+      return {
+        data: response,
+        rawResponse: JSON.stringify(response),
+        metrics,
+      };
+    } catch (error) {
+      this.logger?.warn('LangChain structured model invocation failed:', error);
+
+      return {
+        data: {},
+        rawResponse: '',
+        metrics: {
+          success: false,
+          usage: {
+            total: 0,
+            input: 0,
+            output: 0,
+          },
+        },
+      };
+    }
   }
 
   /**