feat: add max_completion_tokens support for O3 family models

- O3 models now include max_completion_tokens when includeMaxTokens is true
- Updated tests to reflect that O3 models support max_completion_tokens
- This addresses PR feedback that O3 models should use addMaxTokensIfNeeded()

Author: AlexandruSmirnov

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

@@ -566,7 +566,7 @@ describe("OpenAiHandler", () => {
 			},
 		}
 
-		it("should handle O3 model with streaming and NOT include max_tokens even when includeMaxTokens is true", async () => {
+		it("should handle O3 model with streaming and include max_completion_tokens when includeMaxTokens is true", async () => {
 			const o3Handler = new OpenAiHandler({
 				...o3Options,
 				includeMaxTokens: true,
@@ -601,7 +601,8 @@ describe("OpenAiHandler", () => {
 					stream_options: { include_usage: true },
 					reasoning_effort: "medium",
 					temperature: 0.5,
-					// O3 models do not support max_tokens
+					// O3 models do not support deprecated max_tokens but do support max_completion_tokens
+					max_completion_tokens: 32000,
 				}),
 				{},
 			)
@@ -650,7 +651,7 @@ describe("OpenAiHandler", () => {
 			expect(callArgs).not.toHaveProperty("max_completion_tokens")
 		})
 
-		it("should handle O3 model non-streaming with reasoning_effort but NO max_tokens", async () => {
+		it("should handle O3 model non-streaming with reasoning_effort and max_completion_tokens when includeMaxTokens is true", async () => {
 			const o3Handler = new OpenAiHandler({
 				...o3Options,
 				openAiStreamingEnabled: false,
@@ -683,7 +684,8 @@ describe("OpenAiHandler", () => {
 					],
 					reasoning_effort: "medium",
 					temperature: 0.3,
-					// O3 models do not support max_tokens
+					// O3 models do not support deprecated max_tokens but do support max_completion_tokens
+					max_completion_tokens: 65536, // Using default maxTokens from o3Options
 				}),
 				{},
 			)

src/api/providers/openai.ts

@@ -310,6 +310,8 @@ export class OpenAiHandler extends BaseProvider implements SingleCompletionHandl
 			}
 
 			// O3 family models do not support max_tokens parameter
+			// but they do support max_completion_tokens
+			this.addMaxTokensIfNeeded(requestOptions, modelInfo)
 
 			const stream = await this.client.chat.completions.create(
 				requestOptions,
@@ -332,6 +334,8 @@ export class OpenAiHandler extends BaseProvider implements SingleCompletionHandl
 			}
 
 			// O3 family models do not support max_tokens parameter
+			// but they do support max_completion_tokens
+			this.addMaxTokensIfNeeded(requestOptions, modelInfo)
 
 			const response = await this.client.chat.completions.create(
 				requestOptions,