fix: update Azure AI deployment handling to support dynamic model IDs and custom deployment names

src/api/providers/__tests__/azure-ai.test.ts

@@ -5,167 +5,180 @@ import ModelClient from "@azure-rest/ai-inference"
 
 // Mock the Azure AI client
 jest.mock("@azure-rest/ai-inference", () => {
-  return {
+	const mockClient = jest.fn().mockImplementation(() => ({
-    __esModule: true,
+		path: jest.fn().mockReturnValue({
-    default: jest.fn().mockImplementation(() => ({
+			post: jest.fn(),
-      path: jest.fn().mockReturnValue({
+		}),
-        post: jest.fn()
+	}))
-      })
+
-    })),
+	return {
-    isUnexpected: jest.fn()
+		__esModule: true,
-  }
+		default: mockClient,
+		isUnexpected: jest.fn(),
+	}
 })
 
 describe("AzureAiHandler", () => {
-  const mockOptions: ApiHandlerOptions = {
+	const mockOptions: ApiHandlerOptions = {
-    apiProvider: "azure-ai",
+		apiModelId: "azure-gpt-35",
-    apiModelId: "azure-gpt-35",
+		azureAiEndpoint: "https://test-resource.inference.azure.com",
-    azureAiEndpoint: "https://test-resource.inference.azure.com",
+		azureAiKey: "test-key",
-    azureAiKey: "test-key",
+	}
-    azureAiDeployments: {
-      "azure-gpt-35": {
-        name: "custom-gpt35",
-        apiVersion: "2024-02-15-preview",
-        modelMeshName: "test-mesh-model"
-      }
-    }
-  }
 
-  beforeEach(() => {
+	beforeEach(() => {
-    jest.clearAllMocks()
+		jest.clearAllMocks()
-  })
+	})
 
-  test("constructs with required options", () => {
+	test("constructs with required options", () => {
-    const handler = new AzureAiHandler(mockOptions)
+		const handler = new AzureAiHandler(mockOptions)
-    expect(handler).toBeInstanceOf(AzureAiHandler)
+		expect(handler).toBeInstanceOf(AzureAiHandler)
-  })
+	})
 
-  test("throws error without endpoint", () => {
+	test("throws error without endpoint", () => {
-    const invalidOptions = { ...mockOptions }
+		const invalidOptions = { ...mockOptions }
-    delete invalidOptions.azureAiEndpoint
+		delete invalidOptions.azureAiEndpoint
-    expect(() => new AzureAiHandler(invalidOptions)).toThrow("Azure AI endpoint is required")
+		expect(() => new AzureAiHandler(invalidOptions)).toThrow("Azure AI endpoint is required")
-  })
+	})
 
-  test("throws error without API key", () => {
+	test("throws error without API key", () => {
-    const invalidOptions = { ...mockOptions }
+		const invalidOptions = { ...mockOptions }
-    delete invalidOptions.azureAiKey
+		delete invalidOptions.azureAiKey
-    expect(() => new AzureAiHandler(invalidOptions)).toThrow("Azure AI key is required")
+		expect(() => new AzureAiHandler(invalidOptions)).toThrow("Azure AI key is required")
-  })
+	})
 
-  test("creates chat completion correctly", async () => {
+	test("creates chat completion correctly", async () => {
-    const handler = new AzureAiHandler(mockOptions)
+		const handler = new AzureAiHandler(mockOptions)
-    const mockResponse = {
+		const mockResponse = {
-      body: {
+			body: {
-        choices: [
+				choices: [
-          {
+					{
-            message: {
+						message: {
-              content: "test response"
+							content: "test response",
-            }
+						},
-          }
+					},
-        ]
+				],
-      }
+			},
-    }
+		}
-    
-    const mockClient = ModelClient as jest.MockedClass<typeof ModelClient>
-    mockClient.prototype.path.mockReturnValue({
-      post: jest.fn().mockResolvedValue(mockResponse)
-    })
 
-    const result = await handler.completePrompt("test prompt")
+		const mockClient = ModelClient as jest.MockedFunction<typeof ModelClient>
-    expect(result).toBe("test response")
+		mockClient.mockReturnValue({
+			path: jest.fn().mockReturnValue({
+				post: jest.fn().mockResolvedValue(mockResponse),
+			}),
+		} as any)
 
-    expect(mockClient.prototype.path).toHaveBeenCalledWith("/chat/completions")
+		const result = await handler.completePrompt("test prompt")
-    expect(mockClient.prototype.path().post).toHaveBeenCalledWith({
+		expect(result).toBe("test response")
-      body: {
+	})
-        messages: [{ role: "user", content: "test prompt" }],
-        temperature: 0
-      }
-    })
-  })
 
-  test("handles streaming responses correctly", async () => {
+	test("handles streaming responses correctly", async () => {
-    const handler = new AzureAiHandler(mockOptions)
+		const handler = new AzureAiHandler(mockOptions)
-    const mockStream = Readable.from([
+		const mockStream = new Readable({
-      'data: {"choices":[{"delta":{"content":"Hello"},"finish_reason":null}]}\n\n',
+			read() {
-      'data: {"choices":[{"delta":{"content":" world"},"finish_reason":"stop"}],"usage":{"prompt_tokens":10,"completion_tokens":2}}\n\n',
+				this.push('data: {"choices":[{"delta":{"content":"Hello"},"finish_reason":null}]}\n\n')
-      'data: [DONE]\n\n'
+				this.push(
-    ])
+					'data: {"choices":[{"delta":{"content":" world"},"finish_reason":"stop"}],"usage":{"prompt_tokens":10,"completion_tokens":2}}\n\n',
+				)
+				this.push("data: [DONE]\n\n")
+				this.push(null)
+			},
+		})
 
-    const mockClient = ModelClient as jest.MockedClass<typeof ModelClient>
+		const mockResponse = {
-    mockClient.prototype.path.mockReturnValue({
+			status: 200,
-      post: jest.fn().mockResolvedValue({
+			body: mockStream,
-        status: 200,
+		}
-        body: mockStream,
-      })
-    })
 
-    const messages = []
+		const mockClient = ModelClient as jest.MockedFunction<typeof ModelClient>
-    for await (const message of handler.createMessage("system prompt", [])) {
+		mockClient.mockReturnValue({
-      messages.push(message)
+			path: jest.fn().mockReturnValue({
-    }
+				post: jest.fn().mockReturnValue({
+					asNodeStream: () => Promise.resolve(mockResponse),
+				}),
+			}),
+		} as any)
 
-    expect(messages).toEqual([
+		const messages = []
-      { type: "text", text: "Hello" },
+		for await (const message of handler.createMessage("system prompt", [])) {
-      { type: "text", text: " world" },
+			messages.push(message)
-      { type: "usage", inputTokens: 10, outputTokens: 2 }
+		}
-    ])
 
-    expect(mockClient.prototype.path().post).toHaveBeenCalledWith({
+		expect(messages).toEqual([
-      body: {
+			{ type: "text", text: "Hello" },
-        messages: [{ role: "system", content: "system prompt" }],
+			{ type: "text", text: " world" },
-        temperature: 0,
+			{ type: "usage", inputTokens: 10, outputTokens: 2 },
-        stream: true,
+		])
-        max_tokens: expect.any(Number)
+	})
-      }
-    })
-  })
 
-  test("handles rate limit errors", async () => {
+	test("handles rate limit errors", async () => {
-    const handler = new AzureAiHandler(mockOptions)
+		const handler = new AzureAiHandler(mockOptions)
-    const mockError = new Error("Rate limit exceeded")
+		const mockError = new Error("Rate limit exceeded")
-    Object.assign(mockError, { status: 429 })
+		Object.assign(mockError, { status: 429 })
 
-    const mockClient = ModelClient as jest.MockedClass<typeof ModelClient>
+		const mockClient = ModelClient as jest.MockedFunction<typeof ModelClient>
-    mockClient.prototype.path.mockReturnValue({
+		mockClient.mockReturnValue({
-      post: jest.fn().mockRejectedValue(mockError)
+			path: jest.fn().mockReturnValue({
-    })
+				post: jest.fn().mockRejectedValue(mockError),
+			}),
+		} as any)
 
-    await expect(handler.completePrompt("test")).rejects.toThrow(
+		await expect(handler.completePrompt("test")).rejects.toThrow(
-      "Azure AI rate limit exceeded. Please try again later."
+			"Azure AI rate limit exceeded. Please try again later.",
-    )
+		)
-  })
+	})
 
-  test("handles content safety errors", async () => {
+	test("handles content safety errors", async () => {
-    const handler = new AzureAiHandler(mockOptions)
+		const handler = new AzureAiHandler(mockOptions)
-    const mockError = {
+		const mockError = {
-      status: 400,
+			status: 400,
-      body: {
+			body: {
-        error: {
+				error: {
-          code: "ContentFilterError",
+					code: "ContentFilterError",
-          message: "Content was flagged by content safety filters"
+					message: "Content was flagged by content safety filters",
-        }
+				},
-      }
+			},
-    }
+		}
 
-    const mockClient = ModelClient as jest.MockedClass<typeof ModelClient>
+		const mockClient = ModelClient as jest.MockedFunction<typeof ModelClient>
-    mockClient.prototype.path.mockReturnValue({
+		mockClient.mockReturnValue({
-      post: jest.fn().mockRejectedValue(mockError)
+			path: jest.fn().mockReturnValue({
-    })
+				post: jest.fn().mockRejectedValue(mockError),
+			}),
+		} as any)
 
-    await expect(handler.completePrompt("test")).rejects.toThrow(
+		await expect(handler.completePrompt("test")).rejects.toThrow(
-      "Azure AI completion error: Content was flagged by content safety filters"
+			"Content was flagged by Azure AI content safety filters",
-    )
+		)
-  })
+	})
 
-  test("falls back to default model configuration", async () => {
+	test("falls back to default model configuration", () => {
-    const options = { ...mockOptions }
+		const handler = new AzureAiHandler({
-    delete options.azureAiDeployments
+			azureAiEndpoint: "https://test.azure.com",
+			azureAiKey: "test-key",
+		})
+		const model = handler.getModel()
 
-    const handler = new AzureAiHandler(options)
+		expect(model.id).toBe("azure-gpt-35")
-    const model = handler.getModel()
+		expect(model.info).toBeDefined()
+	})
 
-    expect(model.id).toBe("azure-gpt-35")
+	test("supports custom deployment names", async () => {
-    expect(model.info).toBeDefined()
+		const customOptions = {
-    expect(model.info.defaultDeployment.name).toBe("azure-gpt-35")
+			...mockOptions,
-  })
+			apiModelId: "custom-model",
-})
+			azureAiDeployments: {
+				"custom-model": {
+					name: "my-custom-deployment",
+					apiVersion: "2024-02-15-preview",
+					modelMeshName: "my-custom-model",
+				},
+			},
+		}
+
+		const handler = new AzureAiHandler(customOptions)
+		const model = handler.getModel()
+
+		expect(model.id).toBe("custom-model")
+		expect(model.info).toBeDefined()
+	})
+})

src/api/providers/azure-ai.ts

@@ -2,22 +2,17 @@ import { Anthropic } from "@anthropic-ai/sdk"
 import ModelClient from "@azure-rest/ai-inference"
 import { isUnexpected } from "@azure-rest/ai-inference"
 import { AzureKeyCredential } from "@azure/core-auth"
-import {
+import { ApiHandlerOptions, ModelInfo, AzureDeploymentConfig } from "../../shared/api"
-	ApiHandlerOptions,
-	ModelInfo,
-	azureAiDefaultModelId,
-	AzureAiModelId,
-	azureAiModels,
-	AzureDeploymentConfig,
-} from "../../shared/api"
 import { ApiHandler, SingleCompletionHandler } from "../index"
 import { convertToOpenAiMessages } from "../transform/openai-format"
 import { ApiStream } from "../transform/stream"
-import { createSseStream } from "@azure/core-rest-pipeline"
+
+const DEFAULT_API_VERSION = "2024-02-15-preview"
+const DEFAULT_MAX_TOKENS = 4096
 
 export class AzureAiHandler implements ApiHandler, SingleCompletionHandler {
 	private options: ApiHandlerOptions
-	private client: ModelClient
+	private client: ReturnType<typeof ModelClient>
 
 	constructor(options: ApiHandlerOptions) {
 		this.options = options
@@ -30,22 +25,36 @@ export class AzureAiHandler implements ApiHandler, SingleCompletionHandler {
 			throw new Error("Azure AI key is required")
 		}
 
-		this.client = new ModelClient(options.azureAiEndpoint, new AzureKeyCredential(options.azureAiKey))
+		this.client = ModelClient(options.azureAiEndpoint, new AzureKeyCredential(options.azureAiKey))
 	}
 
 	private getDeploymentConfig(): AzureDeploymentConfig {
-		const model = this.getModel()
+		const modelId = this.options.apiModelId
-		const defaultConfig = azureAiModels[model.id].defaultDeployment
+		if (!modelId) {
+			return {
+				name: "gpt-35-turbo", // Default deployment name if none specified
+				apiVersion: DEFAULT_API_VERSION,
+			}
+		}
 
+		const customConfig = this.options.azureAiDeployments?.[modelId]
+		if (customConfig) {
+			return {
+				name: customConfig.name,
+				apiVersion: customConfig.apiVersion || DEFAULT_API_VERSION,
+				modelMeshName: customConfig.modelMeshName,
+			}
+		}
+
+		// If no custom config, use model ID as deployment name
 		return {
-			name: this.options.azureAiDeployments?.[model.id]?.name || defaultConfig.name,
+			name: modelId,
-			apiVersion: this.options.azureAiDeployments?.[model.id]?.apiVersion || defaultConfig.apiVersion,
+			apiVersion: DEFAULT_API_VERSION,
-			modelMeshName: this.options.azureAiDeployments?.[model.id]?.modelMeshName,
 		}
 	}
 
 	async *createMessage(systemPrompt: string, messages: Anthropic.Messages.MessageParam[]): ApiStream {
-		const modelInfo = this.getModel().info
+		const deployment = this.getDeploymentConfig()
 		const chatMessages = [{ role: "system", content: systemPrompt }, ...convertToOpenAiMessages(messages)]
 
 		try {
@@ -56,12 +65,12 @@ export class AzureAiHandler implements ApiHandler, SingleCompletionHandler {
 						messages: chatMessages,
 						temperature: 0,
 						stream: true,
-						max_tokens: modelInfo.maxTokens,
+						max_tokens: DEFAULT_MAX_TOKENS,
-						response_format: { type: "text" }, // Ensure text format for chat
+						response_format: { type: "text" },
 					},
-					headers: this.getDeploymentConfig().modelMeshName
+					headers: deployment.modelMeshName
 						? {
-								"x-ms-model-mesh-model-name": this.getDeploymentConfig().modelMeshName,
+								"x-ms-model-mesh-model-name": deployment.modelMeshName,
 							}
 						: undefined,
 				})
@@ -69,22 +78,22 @@ export class AzureAiHandler implements ApiHandler, SingleCompletionHandler {
 
 			const stream = response.body
 			if (!stream) {
-				throw new Error(`Failed to get chat completions with status: ${response.status}`)
+				throw new Error("Failed to get chat completions stream")
 			}
 
-			if (response.status !== 200) {
+			const statusCode = Number(response.status)
-				throw new Error(`Failed to get chat completions: ${response.body.error}`)
+			if (statusCode !== 200) {
+				throw new Error(`Failed to get chat completions: HTTP ${statusCode}`)
 			}
 
-			const sseStream = createSseStream(stream)
+			for await (const chunk of stream) {
-
+				const chunkStr = chunk.toString()
-			for await (const event of sseStream) {
+				if (chunkStr === "data: [DONE]\n\n") {
-				if (event.data === "[DONE]") {
 					return
 				}
 
 				try {
-					const data = JSON.parse(event.data)
+					const data = JSON.parse(chunkStr.replace("data: ", ""))
 					const delta = data.choices[0]?.delta
 
 					if (delta?.content) {
@@ -124,26 +133,29 @@ export class AzureAiHandler implements ApiHandler, SingleCompletionHandler {
 		}
 	}
 
-	getModel(): { id: AzureAiModelId; info: ModelInfo } {
+	getModel(): { id: string; info: ModelInfo } {
-		const modelId = this.options.apiModelId
+		return {
-		if (modelId && modelId in azureAiModels) {
+			id: this.options.apiModelId || "gpt-35-turbo",
-			const id = modelId as AzureAiModelId
+			info: {
-			return { id, info: azureAiModels[id] }
+				maxTokens: DEFAULT_MAX_TOKENS,
+				contextWindow: 16385, // Conservative default
+				supportsPromptCache: true,
+			},
 		}
-		return { id: azureAiDefaultModelId, info: azureAiModels[azureAiDefaultModelId] }
 	}
 
 	async completePrompt(prompt: string): Promise<string> {
 		try {
+			const deployment = this.getDeploymentConfig()
 			const response = await this.client.path("/chat/completions").post({
 				body: {
 					messages: [{ role: "user", content: prompt }],
 					temperature: 0,
 					response_format: { type: "text" },
 				},
-				headers: this.getDeploymentConfig().modelMeshName
+				headers: deployment.modelMeshName
 					? {
-							"x-ms-model-mesh-model-name": this.getDeploymentConfig().modelMeshName,
+							"x-ms-model-mesh-model-name": deployment.modelMeshName,
 						}
 					: undefined,
 			})

src/core/webview/ClineProvider.ts

@@ -86,7 +86,7 @@ type GlobalStateKey =
 	| "lmStudioBaseUrl"
 	| "anthropicBaseUrl"
 	| "azureApiVersion"
-  | "azureAiDeployments"
+	| "azureAiDeployments"
 	| "openAiStreamingEnabled"
 	| "openRouterModelId"
 	| "openRouterModelInfo"
@@ -1075,16 +1075,25 @@ export class ClineProvider implements vscode.WebviewViewProvider {
 						await this.updateGlobalState("autoApprovalEnabled", message.bool ?? false)
 						await this.postStateToWebview()
 						break
-      case "updateAzureAiDeployment":
+					case "updateAzureAiDeployment":
-        if (message.azureAiDeployment) {
+						if (message.azureAiDeployment) {
-          const deployments = await this.getGlobalState("azureAiDeployments") || {}
+							const deployments = ((await this.getGlobalState("azureAiDeployments")) || {}) as Record<
-          deployments[message.azureAiDeployment.modelId] = {
+								string,
-            ...message.azureAiDeployment,
+								{
-          }
+									name: string
-          await this.updateGlobalState("azureAiDeployments", deployments)
+									apiVersion: string
-          await this.postStateToWebview()
+									modelMeshName?: string
-        }
+								}
-        break
+							>
+							deployments[message.azureAiDeployment.modelId] = {
+								name: message.azureAiDeployment.name,
+								apiVersion: message.azureAiDeployment.apiVersion,
+								modelMeshName: message.azureAiDeployment.modelMeshName,
+							}
+							await this.updateGlobalState("azureAiDeployments", deployments)
+							await this.postStateToWebview()
+						}
+						break
 					case "enhancePrompt":
 						if (message.text) {
 							try {
@@ -1517,7 +1526,7 @@ export class ClineProvider implements vscode.WebviewViewProvider {
 		await this.storeSecret("openAiNativeApiKey", openAiNativeApiKey)
 		await this.storeSecret("deepSeekApiKey", deepSeekApiKey)
 		await this.updateGlobalState("azureApiVersion", azureApiVersion)
-    await this.updateGlobalState("azureAiDeployments", apiConfiguration.azureAiDeployments)
+		await this.updateGlobalState("azureAiDeployments", apiConfiguration.azureAiDeployments)
 		await this.updateGlobalState("openAiStreamingEnabled", openAiStreamingEnabled)
 		await this.updateGlobalState("openRouterModelId", openRouterModelId)
 		await this.updateGlobalState("openRouterModelInfo", openRouterModelInfo)
@@ -2159,7 +2168,7 @@ export class ClineProvider implements vscode.WebviewViewProvider {
 			openAiNativeApiKey,
 			deepSeekApiKey,
 			mistralApiKey,
-    azureAiDeployments,
+			azureAiDeployments,
 			azureApiVersion,
 			openAiStreamingEnabled,
 			openRouterModelId,
@@ -2234,7 +2243,7 @@ export class ClineProvider implements vscode.WebviewViewProvider {
 			this.getSecret("openAiNativeApiKey") as Promise<string | undefined>,
 			this.getSecret("deepSeekApiKey") as Promise<string | undefined>,
 			this.getSecret("mistralApiKey") as Promise<string | undefined>,
-    this.getGlobalState("azureAiDeployments") as Promise<Record<string, any> | undefined>,
+			this.getGlobalState("azureAiDeployments") as Promise<Record<string, any> | undefined>,
 			this.getGlobalState("azureApiVersion") as Promise<string | undefined>,
 			this.getGlobalState("openAiStreamingEnabled") as Promise<boolean | undefined>,
 			this.getGlobalState("openRouterModelId") as Promise<string | undefined>,
@@ -2327,7 +2336,7 @@ export class ClineProvider implements vscode.WebviewViewProvider {
 				deepSeekApiKey,
 				mistralApiKey,
 				azureApiVersion,
-    azureAiDeployments,
+				azureAiDeployments,
 				openAiStreamingEnabled,
 				openRouterModelId,
 				openRouterModelInfo,

src/shared/api.ts

@@ -15,7 +15,7 @@ export type ApiProvider =
 	| "vscode-lm"
 	| "mistral"
 	| "unbound"
-| "azure-ai"
+	| "azure-ai"
 
 export interface ApiHandlerOptions {
 	apiModelId?: string
@@ -61,15 +61,17 @@ export interface ApiHandlerOptions {
 	includeMaxTokens?: boolean
 	unboundApiKey?: string
 	unboundModelId?: string
-  azureAiEndpoint?: string
+	azureAiEndpoint?: string
-  azureAiKey?: string
+	azureAiKey?: string
-  azureAiDeployments?: {
+	azureAiDeployments?:
-    [key in AzureAiModelId]?: {
+		| {
-      name: string
+				[key: string]: {
-      apiVersion: string
+					name: string
-      modelMeshName?: string
+					apiVersion: string
-    }
+					modelMeshName?: string
-  }
+				}
+		  }
+		| undefined
 }
 
 export type ApiConfiguration = ApiHandlerOptions & {
@@ -650,45 +652,45 @@ export const unboundModels = {
 export type AzureAiModelId = "azure-gpt-35" | "azure-gpt-4" | "azure-gpt-4-turbo"
 
 export interface AzureDeploymentConfig {
-  name: string
+	name: string
-  apiVersion: string
+	apiVersion: string
-  modelMeshName?: string // For Model-Mesh deployments
+	modelMeshName?: string // For Model-Mesh deployments
 }
 
 export const azureAiModels: Record<AzureAiModelId, ModelInfo & { defaultDeployment: AzureDeploymentConfig }> = {
-  "azure-gpt-35": {
+	"azure-gpt-35": {
-    maxTokens: 4096,
+		maxTokens: 4096,
-    contextWindow: 16385,
+		contextWindow: 16385,
-    supportsPromptCache: true,
+		supportsPromptCache: true,
-    inputPrice: 0.0015,
+		inputPrice: 0.0015,
-    outputPrice: 0.002,
+		outputPrice: 0.002,
-    defaultDeployment: {
+		defaultDeployment: {
-      name: "azure-gpt-35",
+			name: "azure-gpt-35",
-      apiVersion: "2024-02-15-preview"
+			apiVersion: "2024-02-15-preview",
-    }
+		},
-  },
+	},
-  "azure-gpt-4": {
+	"azure-gpt-4": {
-    maxTokens: 8192,
+		maxTokens: 8192,
-    contextWindow: 8192,
+		contextWindow: 8192,
-    supportsPromptCache: true,
+		supportsPromptCache: true,
-    inputPrice: 0.03,
+		inputPrice: 0.03,
-    outputPrice: 0.06,
+		outputPrice: 0.06,
-    defaultDeployment: {
+		defaultDeployment: {
-      name: "azure-gpt-4",
+			name: "azure-gpt-4",
-      apiVersion: "2024-02-15-preview"
+			apiVersion: "2024-02-15-preview",
-    }
+		},
-  },
+	},
-  "azure-gpt-4-turbo": {
+	"azure-gpt-4-turbo": {
-    maxTokens: 4096,
+		maxTokens: 4096,
-    contextWindow: 128000,
+		contextWindow: 128000,
-    supportsPromptCache: true,
+		supportsPromptCache: true,
-    inputPrice: 0.01,
+		inputPrice: 0.01,
-    outputPrice: 0.03,
+		outputPrice: 0.03,
-    defaultDeployment: {
+		defaultDeployment: {
-      name: "azure-gpt-4-turbo",
+			name: "azure-gpt-4-turbo",
-      apiVersion: "2024-02-15-preview"
+			apiVersion: "2024-02-15-preview",
-    }
+		},
-  }
+	},
 } as const satisfies Record<AzureAiModelId, ModelInfo & { defaultDeployment: AzureDeploymentConfig }>
 
 export const azureAiDefaultModelId: AzureAiModelId = "azure-gpt-35"