fix: Fix the issue of Moonshot's maximum return token count being limited to 1024 (#7673)

Co-authored-by: ellipsis-dev[bot] <65095814+ellipsis-dev[bot]@users.noreply.github.com>
Co-authored-by: roomote[bot] <219738659+roomote[bot]@users.noreply.github.com>

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

@@ -294,4 +294,66 @@ describe("MoonshotHandler", () => {
 			expect(result.cacheReadTokens).toBeUndefined()
 		})
 	})
+
+	describe("addMaxTokensIfNeeded", () => {
+		it("should always add max_tokens regardless of includeMaxTokens option", () => {
+			// Create a test subclass to access the protected method
+			class TestMoonshotHandler extends MoonshotHandler {
+				public testAddMaxTokensIfNeeded(requestOptions: any, modelInfo: any) {
+					this.addMaxTokensIfNeeded(requestOptions, modelInfo)
+				}
+			}
+
+			const testHandler = new TestMoonshotHandler(mockOptions)
+			const requestOptions: any = {}
+			const modelInfo = {
+				maxTokens: 32_000,
+			}
+
+			// Test with includeMaxTokens set to false - should still add max tokens
+			testHandler.testAddMaxTokensIfNeeded(requestOptions, modelInfo)
+
+			expect(requestOptions.max_tokens).toBe(32_000)
+		})
+
+		it("should use modelMaxTokens when provided", () => {
+			class TestMoonshotHandler extends MoonshotHandler {
+				public testAddMaxTokensIfNeeded(requestOptions: any, modelInfo: any) {
+					this.addMaxTokensIfNeeded(requestOptions, modelInfo)
+				}
+			}
+
+			const customMaxTokens = 5000
+			const testHandler = new TestMoonshotHandler({
+				...mockOptions,
+				modelMaxTokens: customMaxTokens,
+			})
+			const requestOptions: any = {}
+			const modelInfo = {
+				maxTokens: 32_000,
+			}
+
+			testHandler.testAddMaxTokensIfNeeded(requestOptions, modelInfo)
+
+			expect(requestOptions.max_tokens).toBe(customMaxTokens)
+		})
+
+		it("should fall back to modelInfo.maxTokens when modelMaxTokens is not provided", () => {
+			class TestMoonshotHandler extends MoonshotHandler {
+				public testAddMaxTokensIfNeeded(requestOptions: any, modelInfo: any) {
+					this.addMaxTokensIfNeeded(requestOptions, modelInfo)
+				}
+			}
+
+			const testHandler = new TestMoonshotHandler(mockOptions)
+			const requestOptions: any = {}
+			const modelInfo = {
+				maxTokens: 16_000,
+			}
+
+			testHandler.testAddMaxTokensIfNeeded(requestOptions, modelInfo)
+
+			expect(requestOptions.max_tokens).toBe(16_000)
+		})
+	})
 })

src/api/providers/moonshot.ts

@@ -1,4 +1,5 @@
-import { moonshotModels, moonshotDefaultModelId } from "@roo-code/types"
+import OpenAI from "openai"
+import { moonshotModels, moonshotDefaultModelId, type ModelInfo } from "@roo-code/types"
 
 import type { ApiHandlerOptions } from "../../shared/api"
 
@@ -36,4 +37,15 @@ export class MoonshotHandler extends OpenAiHandler {
 			cacheReadTokens: usage?.cached_tokens,
 		}
 	}
+
+	// Override to always include max_tokens for Moonshot (not max_completion_tokens)
+	protected override addMaxTokensIfNeeded(
+		requestOptions:
+			| OpenAI.Chat.Completions.ChatCompletionCreateParamsStreaming
+			| OpenAI.Chat.Completions.ChatCompletionCreateParamsNonStreaming,
+		modelInfo: ModelInfo,
+	): void {
+		// Moonshot uses max_tokens instead of max_completion_tokens
+		requestOptions.max_tokens = this.options.modelMaxTokens || modelInfo.maxTokens
+	}
 }

src/api/providers/openai.ts

@@ -435,7 +435,7 @@ export class OpenAiHandler extends BaseProvider implements SingleCompletionHandl
 	 * Note: max_tokens is deprecated in favor of max_completion_tokens as per OpenAI documentation
 	 * O3 family models handle max_tokens separately in handleO3FamilyMessage
 	 */
-	private addMaxTokensIfNeeded(
+	protected addMaxTokensIfNeeded(
 		requestOptions:
 			| OpenAI.Chat.Completions.ChatCompletionCreateParamsStreaming
 			| OpenAI.Chat.Completions.ChatCompletionCreateParamsNonStreaming,