Add native tool calling support to OpenAI-compatible

packages/types/src/providers/openai.ts

@@ -436,6 +436,7 @@ export const openAiModelInfoSaneDefaults: ModelInfo = {
 	supportsPromptCache: false,
 	inputPrice: 0,
 	outputPrice: 0,
+	supportsNativeTools: true,
 }
 
 // https://learn.microsoft.com/en-us/azure/ai-services/openai/api-version-deprecation

src/api/providers/__tests__/openai.spec.ts

@@ -157,6 +157,55 @@ describe("OpenAiHandler", () => {
 			expect(usageChunk?.outputTokens).toBe(5)
 		})
 
+		it("should handle tool calls in non-streaming mode", async () => {
+			mockCreate.mockResolvedValueOnce({
+				choices: [
+					{
+						message: {
+							role: "assistant",
+							content: null,
+							tool_calls: [
+								{
+									id: "call_1",
+									type: "function",
+									function: {
+										name: "test_tool",
+										arguments: '{"arg":"value"}',
+									},
+								},
+							],
+						},
+						finish_reason: "tool_calls",
+					},
+				],
+				usage: {
+					prompt_tokens: 10,
+					completion_tokens: 5,
+					total_tokens: 15,
+				},
+			})
+
+			const handler = new OpenAiHandler({
+				...mockOptions,
+				openAiStreamingEnabled: false,
+			})
+
+			const stream = handler.createMessage(systemPrompt, messages)
+			const chunks: any[] = []
+			for await (const chunk of stream) {
+				chunks.push(chunk)
+			}
+
+			const toolCallChunks = chunks.filter((chunk) => chunk.type === "tool_call")
+			expect(toolCallChunks).toHaveLength(1)
+			expect(toolCallChunks[0]).toEqual({
+				type: "tool_call",
+				id: "call_1",
+				name: "test_tool",
+				arguments: '{"arg":"value"}',
+			})
+		})
+
 		it("should handle streaming responses", async () => {
 			const stream = handler.createMessage(systemPrompt, messages)
 			const chunks: any[] = []
@@ -170,6 +219,66 @@ describe("OpenAiHandler", () => {
 			expect(textChunks[0].text).toBe("Test response")
 		})
 
+		it("should handle tool calls in streaming responses", async () => {
+			mockCreate.mockImplementation(async (options) => {
+				return {
+					[Symbol.asyncIterator]: async function* () {
+						yield {
+							choices: [
+								{
+									delta: {
+										tool_calls: [
+											{
+												index: 0,
+												id: "call_1",
+												function: { name: "test_tool", arguments: "" },
+											},
+										],
+									},
+									finish_reason: null,
+								},
+							],
+						}
+						yield {
+							choices: [
+								{
+									delta: {
+										tool_calls: [{ index: 0, function: { arguments: '{"arg":' } }],
+									},
+									finish_reason: null,
+								},
+							],
+						}
+						yield {
+							choices: [
+								{
+									delta: {
+										tool_calls: [{ index: 0, function: { arguments: '"value"}' } }],
+									},
+									finish_reason: "tool_calls",
+								},
+							],
+						}
+					},
+				}
+			})
+
+			const stream = handler.createMessage(systemPrompt, messages)
+			const chunks: any[] = []
+			for await (const chunk of stream) {
+				chunks.push(chunk)
+			}
+
+			const toolCallChunks = chunks.filter((chunk) => chunk.type === "tool_call")
+			expect(toolCallChunks).toHaveLength(1)
+			expect(toolCallChunks[0]).toEqual({
+				type: "tool_call",
+				id: "call_1",
+				name: "test_tool",
+				arguments: '{"arg":"value"}',
+			})
+		})
+
 		it("should include reasoning_effort when reasoning effort is enabled", async () => {
 			const reasoningOptions: ApiHandlerOptions = {
 				...mockOptions,
@@ -618,6 +727,58 @@ describe("OpenAiHandler", () => {
 			)
 		})
 
+		it("should handle tool calls with O3 model in streaming mode", async () => {
+			const o3Handler = new OpenAiHandler(o3Options)
+
+			mockCreate.mockImplementation(async (options) => {
+				return {
+					[Symbol.asyncIterator]: async function* () {
+						yield {
+							choices: [
+								{
+									delta: {
+										tool_calls: [
+											{
+												index: 0,
+												id: "call_1",
+												function: { name: "test_tool", arguments: "" },
+											},
+										],
+									},
+									finish_reason: null,
+								},
+							],
+						}
+						yield {
+							choices: [
+								{
+									delta: {
+										tool_calls: [{ index: 0, function: { arguments: "{}" } }],
+									},
+									finish_reason: "tool_calls",
+								},
+							],
+						}
+					},
+				}
+			})
+
+			const stream = o3Handler.createMessage("system", [])
+			const chunks: any[] = []
+			for await (const chunk of stream) {
+				chunks.push(chunk)
+			}
+
+			const toolCallChunks = chunks.filter((chunk) => chunk.type === "tool_call")
+			expect(toolCallChunks).toHaveLength(1)
+			expect(toolCallChunks[0]).toEqual({
+				type: "tool_call",
+				id: "call_1",
+				name: "test_tool",
+				arguments: "{}",
+			})
+		})
+
 		it("should handle O3 model with streaming and exclude max_tokens when includeMaxTokens is false", async () => {
 			const o3Handler = new OpenAiHandler({
 				...o3Options,
@@ -705,6 +866,55 @@ describe("OpenAiHandler", () => {
 			expect(callArgs).not.toHaveProperty("stream")
 		})
 
+		it("should handle tool calls with O3 model in non-streaming mode", async () => {
+			const o3Handler = new OpenAiHandler({
+				...o3Options,
+				openAiStreamingEnabled: false,
+			})
+
+			mockCreate.mockResolvedValueOnce({
+				choices: [
+					{
+						message: {
+							role: "assistant",
+							content: null,
+							tool_calls: [
+								{
+									id: "call_1",
+									type: "function",
+									function: {
+										name: "test_tool",
+										arguments: "{}",
+									},
+								},
+							],
+						},
+						finish_reason: "tool_calls",
+					},
+				],
+				usage: {
+					prompt_tokens: 10,
+					completion_tokens: 5,
+					total_tokens: 15,
+				},
+			})
+
+			const stream = o3Handler.createMessage("system", [])
+			const chunks: any[] = []
+			for await (const chunk of stream) {
+				chunks.push(chunk)
+			}
+
+			const toolCallChunks = chunks.filter((chunk) => chunk.type === "tool_call")
+			expect(toolCallChunks).toHaveLength(1)
+			expect(toolCallChunks[0]).toEqual({
+				type: "tool_call",
+				id: "call_1",
+				name: "test_tool",
+				arguments: "{}",
+			})
+		})
+
 		it("should use default temperature of 0 when not specified for O3 models", async () => {
 			const o3Handler = new OpenAiHandler({
 				...o3Options,

src/api/providers/base-openai-compatible-provider.ts

@@ -90,6 +90,8 @@ export abstract class BaseOpenAiCompatibleProvider<ModelName extends string>
 			messages: [{ role: "system", content: systemPrompt }, ...convertToOpenAiMessages(messages)],
 			stream: true,
 			stream_options: { include_usage: true },
+			...(metadata?.tools && { tools: this.convertToolsForOpenAI(metadata.tools) }),
+			...(metadata?.tool_choice && { tool_choice: metadata.tool_choice }),
 		}
 
 		try {
@@ -115,6 +117,8 @@ export abstract class BaseOpenAiCompatibleProvider<ModelName extends string>
 				}) as const,
 		)
 
+		const toolCallAccumulator = new Map<number, { id: string; name: string; arguments: string }>()
+
 		for await (const chunk of stream) {
 			// Check for provider-specific error responses (e.g., MiniMax base_resp)
 			const chunkAny = chunk as any
@@ -125,6 +129,7 @@ export abstract class BaseOpenAiCompatibleProvider<ModelName extends string>
 			}
 
 			const delta = chunk.choices?.[0]?.delta
+			const finishReason = chunk.choices?.[0]?.finish_reason
 
 			if (delta?.content) {
 				for (const processedChunk of matcher.update(delta.content)) {
@@ -139,6 +144,37 @@ export abstract class BaseOpenAiCompatibleProvider<ModelName extends string>
 				}
 			}
 
+			if (delta?.tool_calls) {
+				for (const toolCall of delta.tool_calls) {
+					const index = toolCall.index
+					const existing = toolCallAccumulator.get(index)
+
+					if (existing) {
+						if (toolCall.function?.arguments) {
+							existing.arguments += toolCall.function.arguments
+						}
+					} else {
+						toolCallAccumulator.set(index, {
+							id: toolCall.id || "",
+							name: toolCall.function?.name || "",
+							arguments: toolCall.function?.arguments || "",
+						})
+					}
+				}
+			}
+
+			if (finishReason === "tool_calls") {
+				for (const toolCall of toolCallAccumulator.values()) {
+					yield {
+						type: "tool_call",
+						id: toolCall.id,
+						name: toolCall.name,
+						arguments: toolCall.arguments,
+					}
+				}
+				toolCallAccumulator.clear()
+			}
+
 			if (chunk.usage) {
 				yield {
 					type: "usage",

src/api/providers/base-provider.ts

@@ -18,6 +18,75 @@ export abstract class BaseProvider implements ApiHandler {
 
 	abstract getModel(): { id: string; info: ModelInfo }
 
+	/**
+	 * Converts an array of tools to be compatible with OpenAI's strict mode.
+	 * Filters for function tools and applies schema conversion to their parameters.
+	 */
+	protected convertToolsForOpenAI(tools: any[] | undefined): any[] | undefined {
+		if (!tools) {
+			return undefined
+		}
+
+		return tools.map((tool) =>
+			tool.type === "function"
+				? {
+						...tool,
+						function: {
+							...tool.function,
+							parameters: this.convertToolSchemaForOpenAI(tool.function.parameters),
+						},
+					}
+				: tool,
+		)
+	}
+
+	/**
+	 * Converts tool schemas to be compatible with OpenAI's strict mode by:
+	 * - Ensuring all properties are in the required array (strict mode requirement)
+	 * - Converting nullable types (["type", "null"]) to non-nullable ("type")
+	 * - Recursively processing nested objects and arrays
+	 *
+	 * This matches the behavior of ensureAllRequired in openai-native.ts
+	 */
+	protected convertToolSchemaForOpenAI(schema: any): any {
+		if (!schema || typeof schema !== "object" || schema.type !== "object") {
+			return schema
+		}
+
+		const result = { ...schema }
+
+		if (result.properties) {
+			const allKeys = Object.keys(result.properties)
+			// OpenAI strict mode requires ALL properties to be in required array
+			result.required = allKeys
+
+			// Recursively process nested objects and convert nullable types
+			const newProps = { ...result.properties }
+			for (const key of allKeys) {
+				const prop = newProps[key]
+
+				// Handle nullable types by removing null
+				if (prop && Array.isArray(prop.type) && prop.type.includes("null")) {
+					const nonNullTypes = prop.type.filter((t: string) => t !== "null")
+					prop.type = nonNullTypes.length === 1 ? nonNullTypes[0] : nonNullTypes
+				}
+
+				// Recursively process nested objects
+				if (prop && prop.type === "object") {
+					newProps[key] = this.convertToolSchemaForOpenAI(prop)
+				} else if (prop && prop.type === "array" && prop.items?.type === "object") {
+					newProps[key] = {
+						...prop,
+						items: this.convertToolSchemaForOpenAI(prop.items),
+					}
+				}
+			}
+			result.properties = newProps
+		}
+
+		return result
+	}
+
 	/**
 	 * Default token counting implementation using tiktoken.
 	 * Providers can override this to use their native token counting endpoints.