feat: add Azure OpenAI provider support

.env.example

@@ -22,6 +22,14 @@ BETTER_AUTH_URL=http://localhost:3000
 OPENAI_API_KEY=sk-123456789
 ANTHROPIC_API_KEY=sk-123456789
 
+# Azure OpenAI
+# AZURE_API_KEY=your-azure-api-key
+# Either resource name or base URL — not both:
+# AZURE_RESOURCE_NAME=your-resource-name
+# AZURE_OPENAI_BASE_URL=https://your-resource.openai.azure.com
+# AZURE_API_VERSION=v1
+# AZURE_USE_DEPLOYMENT_BASED_URLS=false
+
 # AWS Bedrock
 # AWS_BEARER_TOKEN_BEDROCK=your-bearer-token
 # AWS_REGION=us-east-1

apps/backend/package.json

@@ -38,6 +38,7 @@
 	"dependencies": {
 		"@ai-sdk/amazon-bedrock": "^4.0.69",
 		"@ai-sdk/anthropic": "^3.0.15",
+		"@ai-sdk/azure": "^3.0.49",
 		"@ai-sdk/google": "^3.0.16",
 		"@ai-sdk/google-vertex": "^4.0.80",
 		"@ai-sdk/mistral": "^3.0.13",

apps/backend/src/agents/provider-meta.ts

@@ -220,12 +220,43 @@ export const PROVIDER_META: ProviderMetaMap = {
 			{ id: 'gemini-3-flash-preview', name: 'Gemini 3 Flash (Vertex)', contextWindow: 200_000 },
 		],
 	},
+	azure: {
+		auth: {
+			apiKey: 'required',
+			hint: 'Provide either a Resource Name or a Base URL — not both',
+			extraFields: [
+				{
+					name: 'resourceName',
+					label: 'Resource Name',
+					envVar: 'AZURE_RESOURCE_NAME',
+					placeholder: 'my-resource (builds https://{name}.openai.azure.com)',
+				},
+				{
+					name: 'apiVersion',
+					label: 'API Version',
+					envVar: 'AZURE_API_VERSION',
+					placeholder: 'v1',
+				},
+				{
+					name: 'useDeploymentBasedUrls',
+					label: 'Use Deployment-Based URLs',
+					envVar: 'AZURE_USE_DEPLOYMENT_BASED_URLS',
+					placeholder: 'false',
+				},
+			],
+		},
+		envVar: 'AZURE_API_KEY',
+		baseUrlEnvVar: 'AZURE_OPENAI_BASE_URL',
+		extractorModelId: '',
+		summaryModelId: '',
+		models: [],
+	},
 };
 
 export function getDefaultModelId(provider: LlmProvider): string {
 	const models = PROVIDER_META[provider].models;
 	const defaultModel = models.find((m) => m.default);
-	return defaultModel?.id ?? models[0].id;
+	return defaultModel?.id ?? models[0]?.id ?? '';
 }
 
 export function getProviderAuth(provider: LlmProvider): ProviderAuth {

apps/backend/src/agents/providers.ts

@@ -1,5 +1,6 @@
 import { createAmazonBedrock } from '@ai-sdk/amazon-bedrock';
 import { type AnthropicProviderOptions, createAnthropic } from '@ai-sdk/anthropic';
+import { createAzure } from '@ai-sdk/azure';
 import { createGoogleGenerativeAI } from '@ai-sdk/google';
 import { createVertex } from '@ai-sdk/google-vertex';
 import { createVertexAnthropic } from '@ai-sdk/google-vertex/anthropic';
@@ -115,6 +116,24 @@ export const LLM_PROVIDERS: LlmProvidersType = {
 			return createVertex(config)(modelId);
 		},
 	},
+	azure: {
+		...PROVIDER_META.azure,
+		create: (settings, modelId) => {
+			const creds = settings.credentials;
+			const resourceName = creds?.resourceName || process.env.AZURE_RESOURCE_NAME;
+			const apiVersion = creds?.apiVersion || process.env.AZURE_API_VERSION;
+			const useDeploymentBasedUrls =
+				(creds?.useDeploymentBasedUrls || process.env.AZURE_USE_DEPLOYMENT_BASED_URLS) === 'true';
+
+			return createAzure({
+				apiKey: settings.apiKey,
+				...(settings.baseURL ? { baseURL: settings.baseURL } : resourceName ? { resourceName } : {}),
+				...(apiVersion && { apiVersion }),
+				...(useDeploymentBasedUrls && { useDeploymentBasedUrls }),
+			})(modelId);
+		},
+		defaultOptions: { store: false },
+	},
 };
 
 export type ProviderModelResult = {

apps/backend/src/types/llm.ts

@@ -1,5 +1,6 @@
 import type { AmazonBedrockLanguageModelOptions } from '@ai-sdk/amazon-bedrock';
 import type { AnthropicProviderOptions } from '@ai-sdk/anthropic';
+import type { OpenAIResponsesProviderOptions as AzureOpenAIResponsesProviderOptions } from '@ai-sdk/azure';
 import type { GoogleGenerativeAIProviderOptions } from '@ai-sdk/google';
 import type { MistralLanguageModelOptions } from '@ai-sdk/mistral';
 import type { OpenAIResponsesProviderOptions } from '@ai-sdk/openai';
@@ -18,6 +19,7 @@ export const llmProviderSchema = z.enum([
 	'ollama',
 	'bedrock',
 	'vertex',
+	'azure',
 ]);
 export type LlmProvider = z.infer<typeof llmProviderSchema>;
 
@@ -47,6 +49,7 @@ export type ProviderConfigMap = {
 	ollama: Flatten<OllamaChatProviderOptions>;
 	bedrock: AmazonBedrockLanguageModelOptions;
 	vertex: GoogleGenerativeAIProviderOptions;
+	azure: AzureOpenAIResponsesProviderOptions;
 };
 
 /** Model definition with provider-specific config type */

apps/frontend/src/components/settings/usage-filters.tsx

@@ -25,6 +25,7 @@ const providerLabels: Record<LlmProvider, string> = {
 	ollama: 'Ollama',
 	bedrock: 'AWS Bedrock',
 	vertex: 'Google Vertex',
+	azure: 'Azure OpenAI',
 };
 
 export const dateFormats: Record<Granularity, string> = {

apps/frontend/src/components/ui/llm-provider-icon.tsx

@@ -1,3 +1,4 @@
+import AzureIcon from '@/components/icons/azure.svg';
 import ClaudeIcon from '@/components/icons/claude.svg';
 import GoogleIcon from '@/components/icons/google.svg';
 import MistralIcon from '@/components/icons/mistral.svg';
@@ -28,6 +29,8 @@ export function LlmProviderIcon({ provider, className: customClassName }: { prov
 			return <BedrockIcon className={className} />;
 		case 'vertex':
 			return <GoogleVertexIcon className={className} />;
+		case 'azure':
+			return <AzureIcon className={className} />;
 		default:
 			return null;
 	}