Merge 287e843737 into c3cae397a1

src/api/providers/__tests__/openai-native-usage.spec.ts

@@ -397,6 +397,60 @@ describe("OpenAiNativeHandler - normalizeUsage", () => {
 			const body = buildRequestBodyForModel(modelId)
 			expect(body.prompt_cache_retention).toBeUndefined()
 		})
+
+		it("should not set prompt_cache_retention when using Azure .azure.com endpoints", () => {
+			const azureHandler = new OpenAiNativeHandler({
+				openAiNativeApiKey: "test-key",
+				openAiNativeBaseUrl: "https://myinstance.openai.azure.com",
+				apiModelId: "gpt-5.1",
+			} as any)
+			const model = azureHandler.getModel()
+			const body = (azureHandler as any).buildRequestBody(model, [], "", model.verbosity, undefined, undefined)
+			expect(body.prompt_cache_retention).toBeUndefined()
+		})
+
+		it("should not set prompt_cache_retention when using Azure AI Foundry endpoints", () => {
+			const azureHandler = new OpenAiNativeHandler({
+				openAiNativeApiKey: "test-key",
+				openAiNativeBaseUrl: "https://myinstance.services.ai.azure.com",
+				apiModelId: "gpt-5.1",
+			} as any)
+			const model = azureHandler.getModel()
+			const body = (azureHandler as any).buildRequestBody(model, [], "", model.verbosity, undefined, undefined)
+			expect(body.prompt_cache_retention).toBeUndefined()
+		})
+
+		it("should not set prompt_cache_retention when using Azure API Management endpoints", () => {
+			const azureHandler = new OpenAiNativeHandler({
+				openAiNativeApiKey: "test-key",
+				openAiNativeBaseUrl: "https://myapi.azure-api.net/openai",
+				apiModelId: "gpt-5.1",
+			} as any)
+			const model = azureHandler.getModel()
+			const body = (azureHandler as any).buildRequestBody(model, [], "", model.verbosity, undefined, undefined)
+			expect(body.prompt_cache_retention).toBeUndefined()
+		})
+
+		it("should set prompt_cache_retention for non-Azure endpoints with eligible models", () => {
+			const normalHandler = new OpenAiNativeHandler({
+				openAiNativeApiKey: "test-key",
+				openAiNativeBaseUrl: "https://api.openai.com/v1",
+				apiModelId: "gpt-5.1",
+			} as any)
+			const model = normalHandler.getModel()
+			const body = (normalHandler as any).buildRequestBody(model, [], "", model.verbosity, undefined, undefined)
+			expect(body.prompt_cache_retention).toBe("24h")
+		})
+
+		it("should set prompt_cache_retention when no base URL is configured (default OpenAI)", () => {
+			const defaultHandler = new OpenAiNativeHandler({
+				openAiNativeApiKey: "test-key",
+				apiModelId: "gpt-5.1",
+			} as any)
+			const model = defaultHandler.getModel()
+			const body = (defaultHandler as any).buildRequestBody(model, [], "", model.verbosity, undefined, undefined)
+			expect(body.prompt_cache_retention).toBe("24h")
+		})
 	})
 
 	describe("cost calculation", () => {

src/api/providers/openai-native.ts

@@ -1399,10 +1399,16 @@ export class OpenAiNativeHandler extends BaseProvider implements SingleCompletio
 	 * The policy is driven by ModelInfo.promptCacheRetention so that model-specific details
 	 * live in the shared types layer rather than this provider. When set to "24h" and the
 	 * model supports prompt caching, extended prompt cache retention is requested.
+	 *
+	 * Note: Azure Foundry/AI does not support the prompt_cache_retention parameter,
+	 * so we skip it when using an Azure base URL.
 	 */
 	private getPromptCacheRetention(model: OpenAiNativeModel): "24h" | undefined {
 		if (!model.info.supportsPromptCache) return undefined
 
+		// Azure Foundry/AI does not support prompt_cache_retention
+		if (this.isAzureEndpoint()) return undefined
+
 		if (model.info.promptCacheRetention === "24h") {
 			return "24h"
 		}
@@ -1410,6 +1416,25 @@ export class OpenAiNativeHandler extends BaseProvider implements SingleCompletio
 		return undefined
 	}
 
+	/**
+	 * Checks if the configured base URL is an Azure endpoint.
+	 * Azure Foundry/AI endpoints use different URL patterns than OpenAI's API
+	 * and may not support all OpenAI-specific parameters like prompt_cache_retention.
+	 */
+	private isAzureEndpoint(): boolean {
+		const baseUrl = this.options.openAiNativeBaseUrl
+		if (!baseUrl) return false
+
+		try {
+			const url = new URL(baseUrl)
+			const host = url.host.toLowerCase()
+			// Match Azure OpenAI Service and Azure AI Foundry endpoints
+			return host.endsWith(".azure.com") || host.endsWith(".azure-api.net")
+		} catch {
+			return false
+		}
+	}
+
 	/**
 	 * Returns a shallow-cloned ModelInfo with pricing overridden for the given tier, if available.
 	 * If no tier or no overrides exist, the original ModelInfo is returned.