Skip to content

Enhanced ChatGPT Clone: Features Agents, MCP, Skills, DeepSeek, Anthropic, AWS, OpenAI, Responses API, Azure, Groq, o1, GPT-5, Mistral, OpenRouter, Vertex AI, Gemini, Artifacts, AI model switching, message search, Code Interpreter, langchain, DALL-E-3, OpenAPI Actions, Functions, Secure Multi-User Auth, Presets, open-source for self-hosting. Active

🪙 fix: Max Output Tokens Refactor for Responses API (#8972)

Dustin Healy21e00168b107
diff --git a/api/app/clients/OpenAIClient.js b/api/app/clients/OpenAIClient.js
index 2eda322640..700fca6a7b 100644
--- a/api/app/clients/OpenAIClient.js
+++ b/api/app/clients/OpenAIClient.js
@@ -1222,7 +1222,9 @@ ${convo}
}
if (this.isOmni === true && modelOptions.max_tokens != null) {
- modelOptions.max_completion_tokens = modelOptions.max_tokens;
+ const paramName =
+ modelOptions.useResponsesApi === true ? 'max_output_tokens' : 'max_completion_tokens';
+ modelOptions[paramName] = modelOptions.max_tokens;
delete modelOptions.max_tokens;
}
if (this.isOmni === true && modelOptions.temperature != null) {
diff --git a/api/server/controllers/agents/client.js b/api/server/controllers/agents/client.js
index 25d98aa3fc..803475e2e3 100644
--- a/api/server/controllers/agents/client.js
+++ b/api/server/controllers/agents/client.js
@@ -1154,6 +1154,8 @@ class AgentClient extends BaseClient {
}
if (shouldRemoveMaxTokens && clientOptions?.modelKwargs?.max_completion_tokens != null) {
delete clientOptions.modelKwargs.max_completion_tokens;
+ } else if (shouldRemoveMaxTokens && clientOptions?.modelKwargs?.max_output_tokens != null) {
+ delete clientOptions.modelKwargs.max_output_tokens;
}
clientOptions = Object.assign(
diff --git a/api/server/controllers/agents/client.test.js b/api/server/controllers/agents/client.test.js
index 5079bcebd8..27a54c9874 100644
--- a/api/server/controllers/agents/client.test.js
+++ b/api/server/controllers/agents/client.test.js
@@ -786,6 +786,28 @@ describe('AgentClient - titleConvo', () => {
expect(clientOptions.temperature).toBe(0.7); // Other options should remain
});
+ it('should move maxTokens to modelKwargs.max_output_tokens for GPT-5 models with useResponsesApi', () => {
+ const clientOptions = {
+ model: 'gpt-5',
+ maxTokens: 2048,
+ temperature: 0.7,
+ useResponsesApi: true,
+ };
+
+ if (/\bgpt-[5-9]\b/i.test(clientOptions.model) && clientOptions.maxTokens != null) {
+ clientOptions.modelKwargs = clientOptions.modelKwargs ?? {};
+ const paramName =
+ clientOptions.useResponsesApi === true ? 'max_output_tokens' : 'max_completion_tokens';
+ clientOptions.modelKwargs[paramName] = clientOptions.maxTokens;
+ delete clientOptions.maxTokens;
+ }
+
+ expect(clientOptions.maxTokens).toBeUndefined();
+ expect(clientOptions.modelKwargs).toBeDefined();
+ expect(clientOptions.modelKwargs.max_output_tokens).toBe(2048);
+ expect(clientOptions.temperature).toBe(0.7); // Other options should remain
+ });
+
it('should handle GPT-5+ models with existing modelKwargs', () => {
const clientOptions = {
model: 'gpt-6',
@@ -866,6 +888,45 @@ describe('AgentClient - titleConvo', () => {
});
});
+ it('should not swap max token param for older models when using useResponsesApi', () => {
+ const testCases = [
+ { model: 'gpt-5', shouldTransform: true },
+ { model: 'gpt-5-turbo', shouldTransform: true },
+ { model: 'gpt-6', shouldTransform: true },
+ { model: 'gpt-7-preview', shouldTransform: true },
+ { model: 'gpt-8', shouldTransform: true },
+ { model: 'gpt-9-mini', shouldTransform: true },
+ { model: 'gpt-4', shouldTransform: false },
+ { model: 'gpt-4o', shouldTransform: false },
+ { model: 'gpt-3.5-turbo', shouldTransform: false },
+ { model: 'claude-3', shouldTransform: false },
+ ];
+
+ testCases.forEach(({ model, shouldTransform }) => {
+ const clientOptions = {
+ model,
+ maxTokens: 1000,
+ useResponsesApi: true,
+ };
+
+ if (/\bgpt-[5-9]\b/i.test(clientOptions.model) && clientOptions.maxTokens != null) {
+ clientOptions.modelKwargs = clientOptions.modelKwargs ?? {};
+ const paramName =
+ clientOptions.useResponsesApi === true ? 'max_output_tokens' : 'max_completion_tokens';
+ clientOptions.modelKwargs[paramName] = clientOptions.maxTokens;
+ delete clientOptions.maxTokens;
+ }
+
+ if (shouldTransform) {
+ expect(clientOptions.maxTokens).toBeUndefined();
+ expect(clientOptions.modelKwargs?.max_output_tokens).toBe(1000);
+ } else {
+ expect(clientOptions.maxTokens).toBe(1000);
+ expect(clientOptions.modelKwargs).toBeUndefined();
+ }
+ });
+ });
+
it('should not transform if maxTokens is null or undefined', () => {
const testCases = [
{ model: 'gpt-5', maxTokens: null },
diff --git a/packages/api/src/agents/__tests__/memory.test.ts b/packages/api/src/agents/__tests__/memory.test.ts
index 7b69992792..74cd0f4354 100644
--- a/packages/api/src/agents/__tests__/memory.test.ts
+++ b/packages/api/src/agents/__tests__/memory.test.ts
@@ -398,4 +398,72 @@ describe('processMemory - GPT-5+ handling', () => {
}),
);
});
+
+ it('should use max_output_tokens when useResponsesApi is true', async () => {
+ await processMemory({
+ res: mockRes as Response,
+ userId: 'test-user',
+ setMemory: mockSetMemory,
+ deleteMemory: mockDeleteMemory,
+ messages: [],
+ memory: 'Test memory',
+ messageId: 'msg-123',
+ conversationId: 'conv-123',
+ instructions: 'Test instructions',
+ llmConfig: {
+ provider: Providers.OPENAI,
+ model: 'gpt-5',
+ maxTokens: 1000,
+ useResponsesApi: true,
+ },
+ });
+
+ const { Run } = jest.requireMock('@librechat/agents');
+ expect(Run.create).toHaveBeenCalledWith(
+ expect.objectContaining({
+ graphConfig: expect.objectContaining({
+ llmConfig: expect.objectContaining({
+ model: 'gpt-5',
+ modelKwargs: {
+ max_output_tokens: 1000,
+ },
+ }),
+ }),
+ }),
+ );
+ });
+
+ it('should use max_completion_tokens when useResponsesApi is false or undefined', async () => {
+ await processMemory({
+ res: mockRes as Response,
+ userId: 'test-user',
+ setMemory: mockSetMemory,
+ deleteMemory: mockDeleteMemory,
+ messages: [],
+ memory: 'Test memory',
+ messageId: 'msg-123',
+ conversationId: 'conv-123',
+ instructions: 'Test instructions',
+ llmConfig: {
+ provider: Providers.OPENAI,
+ model: 'gpt-5',
+ maxTokens: 1000,
+ useResponsesApi: false,
+ },
+ });
+
+ const { Run } = jest.requireMock('@librechat/agents');
+ expect(Run.create).toHaveBeenCalledWith(
+ expect.objectContaining({
+ graphConfig: expect.objectContaining({
+ llmConfig: expect.objectContaining({
+ model: 'gpt-5',
+ modelKwargs: {
+ max_completion_tokens: 1000,
+ },
+ }),
+ }),
+ }),
+ );
+ });
});
diff --git a/packages/api/src/agents/memory.ts b/packages/api/src/agents/memory.ts
index 298ebc04f6..e5fb55ef43 100644
--- a/packages/api/src/agents/memory.ts
+++ b/packages/api/src/agents/memory.ts
@@ -352,7 +352,11 @@ ${memory ?? 'No existing memories'}`;
// Move maxTokens to modelKwargs for GPT-5+ models
if ('maxTokens' in finalLLMConfig && finalLLMConfig.maxTokens != null) {
const modelKwargs = (finalLLMConfig as OpenAIClientOptions).modelKwargs ?? {};
- modelKwargs.max_completion_tokens = finalLLMConfig.maxTokens;
+ const paramName =
+ (finalLLMConfig as OpenAIClientOptions).useResponsesApi === true
+ ? 'max_output_tokens'
+ : 'max_completion_tokens';
+ modelKwargs[paramName] = finalLLMConfig.maxTokens;
delete finalLLMConfig.maxTokens;
(finalLLMConfig as OpenAIClientOptions).modelKwargs = modelKwargs;
}
diff --git a/packages/api/src/endpoints/openai/llm.spec.ts b/packages/api/src/endpoints/openai/llm.spec.ts
index 79d9e05627..90a96c118c 100644
--- a/packages/api/src/endpoints/openai/llm.spec.ts
+++ b/packages/api/src/endpoints/openai/llm.spec.ts
@@ -373,7 +373,7 @@ describe('getOpenAIConfig', () => {
text: {
verbosity: Verbosity.medium,
},
- max_completion_tokens: 1500,
+ max_output_tokens: 1500,
});
});
diff --git a/packages/api/src/endpoints/openai/llm.ts b/packages/api/src/endpoints/openai/llm.ts
index e5c1b49c53..dca5ff7f11 100644
--- a/packages/api/src/endpoints/openai/llm.ts
+++ b/packages/api/src/endpoints/openai/llm.ts
@@ -300,7 +300,9 @@ export function getOpenAIConfig(
}
if (llmConfig.model && /\bgpt-[5-9]\b/i.test(llmConfig.model) && llmConfig.maxTokens != null) {
- modelKwargs.max_completion_tokens = llmConfig.maxTokens;
+ const paramName =
+ llmConfig.useResponsesApi === true ? 'max_output_tokens' : 'max_completion_tokens';
+ modelKwargs[paramName] = llmConfig.maxTokens;
delete llmConfig.maxTokens;
hasModelKwargs = true;
}