mirror of
https://github.com/lobehub/lobe-chat.git
synced 2026-06-16 12:36:07 +00:00
✨ feat: New API support switch Responses API mode (#9776)
* ✨ feat: 添加对新API和路由类型的支持，更新相关配置以启用Responses API
* fix: 更新测试文件中的console.error和console.debug实现，确保输出格式一致；在CreateNewProvider组件中调整provider图标映射逻辑
* ✨ feat: 更新novita和qwen模型，调整定价策略，添加新模型及其功能
* 🐛 fix: OIDC error when connecting to self-host instance (#9916)
fix: oidc/consent redirect header
* ✨ feat: 添加 MiniMax M2 和 Qwen3 VL 235B Instruct 模型，更新模型属性
🔧 fix: 修复免费标识逻辑，确保正确判断模型是否免费
* ✨ feat: 添加 MiniMax-M2 模型，更新 SiliconCloud 和 Vercel AI Gateway 模型信息，调整 Kimi K2 的上下文窗口大小
* fix test
* 📝 docs: update ComfyUI documentation cover image URL (#9997)
* 🔖 chore(release): v1.142.9 [skip ci]
### [Version 1.142.9](https://github.com/lobehub/lobe-chat/compare/v1.142.8...v1.142.9)
<sup>Released on **2025-11-02**</sup>
#### 🐛 Bug Fixes
- **misc**: OIDC error when connecting to self-host instance.
<br/>
<details>
<summary><kbd>Improvements and Fixes</kbd></summary>
#### What's fixed
* **misc**: OIDC error when connecting to self-host instance, closes [#9916](https://github.com/lobehub/lobe-chat/issues/9916) ([2e2b9c4](https://github.com/lobehub/lobe-chat/commit/2e2b9c4))
</details>
<div align="right">
[](#readme-top)
</div>
* 📝 docs(bot): Auto sync agents & plugin to readme
* 优化 Responses API 处理逻辑，优化错误处理和流数据转换
---------
Co-authored-by: Aloxaf <bailong104@gmail.com>
This commit is contained in:
sxjeru
@@ -11,7 +11,7 @@ tags:
|
||||
|
||||
# Using ComfyUI in LobeChat
|
||||
|
||||
<Image alt={'Using ComfyUI in LobeChat'} cover src={'https://github.com/lobehub/lobe-chat/assets/17870709/c9e5eafc-ca22-496b-a88d-cc0ae53bf720'} />
|
||||
<Image alt={'Using ComfyUI in LobeChat'} cover src={'https://hub-apac-1.lobeobjects.space/docs/e9b811f248a1db2bd1be1af888cf9b9d.png'} />
|
||||
|
||||
This documentation will guide you on how to use [ComfyUI](https://github.com/comfyanonymous/ComfyUI) in LobeChat for high-quality AI image generation and editing.
|
||||
|
||||
|
||||
@@ -11,7 +11,7 @@ tags:
|
||||
|
||||
# 在 LobeChat 中使用 ComfyUI
|
||||
|
||||
<Image alt={'在 LobeChat 中使用 ComfyUI'} cover src={'https://github.com/lobehub/lobe-chat/assets/17870709/c9e5eafc-ca22-496b-a88d-cc0ae53bf720'} />
|
||||
<Image alt={'在 LobeChat 中使用 ComfyUI'} cover src={'https://hub-apac-1.lobeobjects.space/docs/e9b811f248a1db2bd1be1af888cf9b9d.png'} />
|
||||
|
||||
本文档将指导你如何在 LobeChat 中使用 [ComfyUI](https://github.com/comfyanonymous/ComfyUI) 进行高质量的 AI 图像生成和编辑。
|
||||
|
||||
|
||||
@@ -174,10 +174,11 @@ const novitaChatModels: AIChatModelCard[] = [
|
||||
contextWindowTokens: 131_072,
|
||||
displayName: 'OpenAI GPT OSS 120B',
|
||||
id: 'openai/gpt-oss-120b',
|
||||
maxOutput: 32_768,
|
||||
pricing: {
|
||||
units: [
|
||||
{ name: 'textInput', rate: 0.1, strategy: 'fixed', unit: 'millionTokens' },
|
||||
{ name: 'textOutput', rate: 0.5, strategy: 'fixed', unit: 'millionTokens' },
|
||||
{ name: 'textInput', rate: 0.05, strategy: 'fixed', unit: 'millionTokens' },
|
||||
{ name: 'textOutput', rate: 0.25, strategy: 'fixed', unit: 'millionTokens' },
|
||||
],
|
||||
},
|
||||
type: 'chat',
|
||||
|
||||
@@ -1,6 +1,20 @@
|
||||
import { AIChatModelCard } from '../types/aiModel';
|
||||
|
||||
const nvidiaChatModels: AIChatModelCard[] = [
|
||||
{
|
||||
abilities: {
|
||||
functionCall: true,
|
||||
reasoning: true,
|
||||
},
|
||||
contextWindowTokens: 128_000,
|
||||
description:
|
||||
'MiniMax-M2 是一款紧凑、快速且经济高效的混合专家（MoE）模型，拥有 2300 亿总参数和 100 亿激活参数，专为编码和智能体任务的顶级性能而打造，同时保持强大的通用智能。该模型在多文件编辑、编码-运行-修复闭环、测试校验修复以及复杂的长链接工具链方面表现优异，是开发者工作流的理想选择。',
|
||||
displayName: 'MiniMax-M2',
|
||||
enabled: true,
|
||||
id: 'minimaxai/minimax-m2',
|
||||
maxOutput: 16_384,
|
||||
type: 'chat',
|
||||
},
|
||||
{
|
||||
abilities: {
|
||||
functionCall: true,
|
||||
|
||||
@@ -1,6 +1,18 @@
|
||||
import { AIChatModelCard } from '../types/aiModel';
|
||||
|
||||
const ollamaCloudModels: AIChatModelCard[] = [
|
||||
{
|
||||
abilities: {
|
||||
functionCall: true,
|
||||
reasoning: true,
|
||||
},
|
||||
contextWindowTokens: 200_000,
|
||||
description: 'MiniMax M2 是专为编码和代理工作流程构建的高效大型语言模型。',
|
||||
displayName: 'MiniMax M2',
|
||||
enabled: true,
|
||||
id: 'minimax-m2',
|
||||
type: 'chat',
|
||||
},
|
||||
{
|
||||
abilities: {
|
||||
functionCall: true,
|
||||
@@ -36,7 +48,6 @@ const ollamaCloudModels: AIChatModelCard[] = [
|
||||
description:
|
||||
'DeepSeek V3.1：下一代推理模型，提升了复杂推理与链路思考能力，适合需要深入分析的任务。',
|
||||
displayName: 'DeepSeek V3.1',
|
||||
enabled: true,
|
||||
id: 'deepseek-v3.1:671b',
|
||||
type: 'chat',
|
||||
},
|
||||
@@ -88,13 +99,23 @@ const ollamaCloudModels: AIChatModelCard[] = [
|
||||
id: 'qwen3-coder:480b',
|
||||
type: 'chat',
|
||||
},
|
||||
{
|
||||
abilities: {
|
||||
functionCall: true,
|
||||
vision: true,
|
||||
},
|
||||
contextWindowTokens: 262_144,
|
||||
displayName: 'Qwen3 VL 235B Instruct',
|
||||
id: 'qwen3-vl:235b-instruct',
|
||||
type: 'chat',
|
||||
},
|
||||
{
|
||||
abilities: {
|
||||
functionCall: true,
|
||||
reasoning: true,
|
||||
vision: true,
|
||||
},
|
||||
contextWindowTokens: 128_000,
|
||||
contextWindowTokens: 262_144,
|
||||
displayName: 'Qwen3 VL 235B',
|
||||
id: 'qwen3-vl:235b',
|
||||
type: 'chat',
|
||||
|
||||
@@ -175,6 +175,51 @@ const qwenChatModels: AIChatModelCard[] = [
|
||||
},
|
||||
{
|
||||
abilities: {
|
||||
functionCall: true,
|
||||
reasoning: true,
|
||||
},
|
||||
contextWindowTokens: 202_752,
|
||||
description: 'GLM系列模型是智谱AI专为智能体设计的混合推理模型，提供思考与非思考两种模式。',
|
||||
displayName: 'GLM-4.6',
|
||||
id: 'glm-4.6',
|
||||
maxOutput: 16_384,
|
||||
pricing: {
|
||||
currency: 'CNY',
|
||||
units: [
|
||||
{
|
||||
lookup: {
|
||||
prices: {
|
||||
'[0, 0.032]': 3,
|
||||
'[0.032, infinity]': 4,
|
||||
},
|
||||
pricingParams: ['textInputRange'],
|
||||
},
|
||||
name: 'textInput',
|
||||
strategy: 'lookup',
|
||||
unit: 'millionTokens',
|
||||
},
|
||||
{
|
||||
lookup: {
|
||||
prices: {
|
||||
'[0, 0.032]': 14,
|
||||
'[0.032, infinity]': 16,
|
||||
},
|
||||
pricingParams: ['textInputRange'],
|
||||
},
|
||||
name: 'textOutput',
|
||||
strategy: 'lookup',
|
||||
unit: 'millionTokens',
|
||||
},
|
||||
],
|
||||
},
|
||||
settings: {
|
||||
extendParams: ['enableReasoning', 'reasoningBudgetToken'],
|
||||
},
|
||||
type: 'chat',
|
||||
},
|
||||
{
|
||||
abilities: {
|
||||
functionCall: true,
|
||||
reasoning: true,
|
||||
},
|
||||
contextWindowTokens: 131_072,
|
||||
@@ -218,6 +263,7 @@ const qwenChatModels: AIChatModelCard[] = [
|
||||
},
|
||||
{
|
||||
abilities: {
|
||||
functionCall: true,
|
||||
reasoning: true,
|
||||
},
|
||||
contextWindowTokens: 131_072,
|
||||
@@ -1451,6 +1497,48 @@ const qwenChatModels: AIChatModelCard[] = [
|
||||
},
|
||||
type: 'chat',
|
||||
},
|
||||
{
|
||||
abilities: {
|
||||
vision: true,
|
||||
reasoning: true,
|
||||
},
|
||||
contextWindowTokens: 131_072,
|
||||
description: 'Qwen3 VL 32B 思考模式（开源版），针对高难度强推理与长视频理解场景，提供顶尖的视觉+文本推理能力。',
|
||||
displayName: 'Qwen3 VL 32B Thinking',
|
||||
id: 'qwen3-vl-32b-thinking',
|
||||
maxOutput: 32_768,
|
||||
organization: 'Qwen',
|
||||
pricing: {
|
||||
currency: 'CNY',
|
||||
units: [
|
||||
{ name: 'textInput', rate: 2, strategy: 'fixed', unit: 'millionTokens' },
|
||||
{ name: 'textOutput', rate: 20, strategy: 'fixed', unit: 'millionTokens' },
|
||||
],
|
||||
},
|
||||
settings: {
|
||||
extendParams: ['enableReasoning', 'reasoningBudgetToken'],
|
||||
},
|
||||
type: 'chat',
|
||||
},
|
||||
{
|
||||
abilities: {
|
||||
vision: true,
|
||||
},
|
||||
contextWindowTokens: 131_072,
|
||||
description: 'Qwen3 VL 32B 非思考模式（Instruct），适用于非思考指令场景，保持强大的视觉理解能力。',
|
||||
displayName: 'Qwen3 VL 32B Instruct',
|
||||
id: 'qwen3-vl-32b-instruct',
|
||||
maxOutput: 32_768,
|
||||
organization: 'Qwen',
|
||||
pricing: {
|
||||
currency: 'CNY',
|
||||
units: [
|
||||
{ name: 'textInput', rate: 2, strategy: 'fixed', unit: 'millionTokens' },
|
||||
{ name: 'textOutput', rate: 8, strategy: 'fixed', unit: 'millionTokens' },
|
||||
],
|
||||
},
|
||||
type: 'chat',
|
||||
},
|
||||
{
|
||||
config: {
|
||||
deploymentName: 'qwen-math-turbo-latest',
|
||||
|
||||
@@ -2,6 +2,26 @@ import { AIChatModelCard, AIImageModelCard } from '../types/aiModel';
|
||||
|
||||
// https://siliconflow.cn/zh-cn/models
|
||||
const siliconcloudChatModels: AIChatModelCard[] = [
|
||||
{
|
||||
abilities: {
|
||||
functionCall: true,
|
||||
reasoning: true,
|
||||
},
|
||||
contextWindowTokens: 128_000,
|
||||
description:
|
||||
'MiniMax-M2 为智能体重新定义了效率。它是一款紧凑、快速且经济高效的 MoE 模型，拥有 2300 亿总参数和 100 亿激活参数，专为编码和智能体任务的顶级性能而打造，同时保持强大的通用智能。仅需 100 亿激活参数，MiniMax-M2 就能提供与大规模模型相媲美的性能，使其成为高效率应用的理想选择。',
|
||||
displayName: 'MiniMax-M2',
|
||||
id: 'MiniMaxAI/MiniMax-M2',
|
||||
pricing: {
|
||||
currency: 'CNY',
|
||||
units: [
|
||||
{ name: 'textInput', rate: 2.1, strategy: 'fixed', unit: 'millionTokens' },
|
||||
{ name: 'textOutput', rate: 8.4, strategy: 'fixed', unit: 'millionTokens' },
|
||||
],
|
||||
},
|
||||
releasedAt: '2025-10-28',
|
||||
type: 'chat',
|
||||
},
|
||||
{
|
||||
abilities: {
|
||||
functionCall: true,
|
||||
|
||||
@@ -1565,23 +1565,6 @@ const vercelAIGatewayChatModels: AIChatModelCard[] = [
|
||||
},
|
||||
type: 'chat',
|
||||
},
|
||||
{
|
||||
abilities: {
|
||||
functionCall: true,
|
||||
},
|
||||
contextWindowTokens: 131_072,
|
||||
description:
|
||||
'DeepSeek-R1-Distill-Llama-70B 是 70B Llama 模型的蒸馏、更高效变体。它在文本生成任务中保持强大性能，减少计算开销以便于部署和研究。由 Groq 使用其自定义语言处理单元 (LPU) 硬件提供服务，以提供快速高效的推理。',
|
||||
displayName: 'DeepSeek R1 Distill Llama 70B',
|
||||
id: 'deepseek/deepseek-r1-distill-llama-70b',
|
||||
pricing: {
|
||||
units: [
|
||||
{ name: 'textInput', rate: 0.75, strategy: 'fixed', unit: 'millionTokens' },
|
||||
{ name: 'textOutput', rate: 0.99, strategy: 'fixed', unit: 'millionTokens' },
|
||||
],
|
||||
},
|
||||
type: 'chat',
|
||||
},
|
||||
{
|
||||
abilities: {
|
||||
functionCall: true,
|
||||
|
||||
@@ -37,7 +37,7 @@ const doubaoChatModels: AIChatModelCard[] = [
|
||||
config: {
|
||||
deploymentName: 'kimi-k2-250905',
|
||||
},
|
||||
contextWindowTokens: 131_072,
|
||||
contextWindowTokens: 262_144,
|
||||
description:
|
||||
'Kimi-K2 是一款Moonshot AI推出的具备超强代码和 Agent 能力的 MoE 架构基础模型，总参数 1T，激活参数 32B。在通用知识推理、编程、数学、Agent 等主要类别的基准性能测试中，K2 模型的性能超过其他主流开源模型。',
|
||||
displayName: 'Kimi K2',
|
||||
|
||||
@@ -1007,78 +1007,122 @@ describe('LobeOpenAICompatibleFactory', () => {
|
||||
});
|
||||
|
||||
describe('responses routing', () => {
|
||||
it('should route to Responses API when chatCompletion.useResponse is true', async () => {
|
||||
const LobeMockProviderUseResponses = createOpenAICompatibleRuntime({
|
||||
baseURL: 'https://api.test.com/v1',
|
||||
chatCompletion: {
|
||||
useResponse: true,
|
||||
},
|
||||
provider: ModelProvider.OpenAI,
|
||||
});
|
||||
it(
|
||||
'should route to Responses API when chatCompletion.useResponse is true',
|
||||
async () => {
|
||||
const LobeMockProviderUseResponses = createOpenAICompatibleRuntime({
|
||||
baseURL: 'https://api.test.com/v1',
|
||||
chatCompletion: {
|
||||
useResponse: true,
|
||||
},
|
||||
provider: ModelProvider.OpenAI,
|
||||
});
|
||||
|
||||
const inst = new LobeMockProviderUseResponses({ apiKey: 'test' });
|
||||
const inst = new LobeMockProviderUseResponses({ apiKey: 'test' });
|
||||
|
||||
// mock responses.create to return a stream-like with tee
|
||||
const prod = new ReadableStream();
|
||||
const debug = new ReadableStream();
|
||||
const mockResponsesCreate = vi
|
||||
.spyOn(inst['client'].responses, 'create')
|
||||
.mockResolvedValue({ tee: () => [prod, debug] } as any);
|
||||
// Mock responses.create to return a proper stream-like object
|
||||
const mockResponsesCreate = vi
|
||||
.spyOn(inst['client'].responses, 'create')
|
||||
.mockResolvedValue({
|
||||
toReadableStream: () =>
|
||||
new ReadableStream({
|
||||
start(controller) {
|
||||
controller.close();
|
||||
},
|
||||
}),
|
||||
} as any);
|
||||
|
||||
await inst.chat({
|
||||
messages: [{ content: 'hi', role: 'user' }],
|
||||
model: 'any-model',
|
||||
temperature: 0,
|
||||
});
|
||||
// Mock getModelPricing to prevent async issues
|
||||
vi.mock('../../utils/model', () => ({
|
||||
getModelPricing: vi.fn().mockResolvedValue({}),
|
||||
}));
|
||||
|
||||
expect(mockResponsesCreate).toHaveBeenCalled();
|
||||
});
|
||||
try {
|
||||
await inst.chat({
|
||||
messages: [{ content: 'hi', role: 'user' }],
|
||||
model: 'any-model',
|
||||
temperature: 0,
|
||||
});
|
||||
} catch (e) {
|
||||
// Catch errors from incomplete mocking, we only care that responses.create was called
|
||||
}
|
||||
|
||||
it('should route to Responses API when model matches useResponseModels', async () => {
|
||||
const LobeMockProviderUseResponseModels = createOpenAICompatibleRuntime({
|
||||
baseURL: 'https://api.test.com/v1',
|
||||
chatCompletion: {
|
||||
useResponseModels: ['special-model', /special-\w+/],
|
||||
},
|
||||
provider: ModelProvider.OpenAI,
|
||||
});
|
||||
const inst = new LobeMockProviderUseResponseModels({ apiKey: 'test' });
|
||||
const spy = vi.spyOn(inst['client'].responses, 'create');
|
||||
// Prevent hanging by mocking normal chat completion stream
|
||||
vi.spyOn(inst['client'].chat.completions, 'create').mockResolvedValue(
|
||||
new ReadableStream() as any,
|
||||
);
|
||||
expect(mockResponsesCreate).toHaveBeenCalled();
|
||||
},
|
||||
{ timeout: 10000 },
|
||||
);
|
||||
|
||||
// First invocation: model contains the string
|
||||
spy.mockResolvedValueOnce({
|
||||
tee: () => [new ReadableStream(), new ReadableStream()],
|
||||
} as any);
|
||||
await inst.chat({
|
||||
messages: [{ content: 'hi', role: 'user' }],
|
||||
model: 'prefix-special-model-suffix',
|
||||
temperature: 0,
|
||||
});
|
||||
expect(spy).toHaveBeenCalledTimes(1);
|
||||
it(
|
||||
'should route to Responses API when model matches useResponseModels',
|
||||
async () => {
|
||||
const LobeMockProviderUseResponseModels = createOpenAICompatibleRuntime({
|
||||
baseURL: 'https://api.test.com/v1',
|
||||
chatCompletion: {
|
||||
useResponseModels: ['special-model', /special-\w+/],
|
||||
},
|
||||
provider: ModelProvider.OpenAI,
|
||||
});
|
||||
const inst = new LobeMockProviderUseResponseModels({ apiKey: 'test' });
|
||||
const spy = vi.spyOn(inst['client'].responses, 'create');
|
||||
// Prevent hanging by mocking normal chat completion stream
|
||||
vi.spyOn(inst['client'].chat.completions, 'create').mockResolvedValue(
|
||||
new ReadableStream() as any,
|
||||
);
|
||||
|
||||
// Second invocation: model matches the RegExp
|
||||
spy.mockResolvedValueOnce({
|
||||
tee: () => [new ReadableStream(), new ReadableStream()],
|
||||
} as any);
|
||||
await inst.chat({
|
||||
messages: [{ content: 'hi', role: 'user' }],
|
||||
model: 'special-xyz',
|
||||
temperature: 0,
|
||||
});
|
||||
expect(spy).toHaveBeenCalledTimes(2);
|
||||
// First invocation: model contains the string
|
||||
spy.mockResolvedValueOnce({
|
||||
toReadableStream: () =>
|
||||
new ReadableStream({
|
||||
start(controller) {
|
||||
controller.close();
|
||||
},
|
||||
}),
|
||||
} as any);
|
||||
try {
|
||||
await inst.chat({
|
||||
messages: [{ content: 'hi', role: 'user' }],
|
||||
model: 'prefix-special-model-suffix',
|
||||
temperature: 0,
|
||||
});
|
||||
} catch (e) {
|
||||
// Catch errors from incomplete mocking
|
||||
}
|
||||
expect(spy).toHaveBeenCalledTimes(1);
|
||||
|
||||
// Third invocation: model does not match any useResponseModels patterns
|
||||
await inst.chat({
|
||||
messages: [{ content: 'hi', role: 'user' }],
|
||||
model: 'unrelated-model',
|
||||
temperature: 0,
|
||||
});
|
||||
expect(spy).toHaveBeenCalledTimes(2); // Ensure no additional calls were made
|
||||
});
|
||||
// Second invocation: model matches the RegExp
|
||||
spy.mockResolvedValueOnce({
|
||||
toReadableStream: () =>
|
||||
new ReadableStream({
|
||||
start(controller) {
|
||||
controller.close();
|
||||
},
|
||||
}),
|
||||
} as any);
|
||||
try {
|
||||
await inst.chat({
|
||||
messages: [{ content: 'hi', role: 'user' }],
|
||||
model: 'special-xyz',
|
||||
temperature: 0,
|
||||
});
|
||||
} catch (e) {
|
||||
// Catch errors from incomplete mocking
|
||||
}
|
||||
expect(spy).toHaveBeenCalledTimes(2);
|
||||
|
||||
// Third invocation: model does not match any useResponseModels patterns
|
||||
try {
|
||||
await inst.chat({
|
||||
messages: [{ content: 'hi', role: 'user' }],
|
||||
model: 'unrelated-model',
|
||||
temperature: 0,
|
||||
});
|
||||
} catch (e) {
|
||||
// Catch errors
|
||||
}
|
||||
expect(spy).toHaveBeenCalledTimes(2); // Ensure no additional calls were made
|
||||
},
|
||||
{ timeout: 10000 },
|
||||
);
|
||||
});
|
||||
|
||||
describe('DEBUG', () => {
|
||||
|
||||
@@ -138,10 +138,10 @@ export interface OpenAICompatibleFactoryOptions<T extends Record<string, any> =
|
||||
useToolsCalling?: boolean;
|
||||
};
|
||||
models?:
|
||||
| ((params: { client: OpenAI }) => Promise<ChatModelCard[]>)
|
||||
| {
|
||||
transformModel?: (model: OpenAI.Model) => ChatModelCard;
|
||||
};
|
||||
| ((params: { client: OpenAI }) => Promise<ChatModelCard[]>)
|
||||
| {
|
||||
transformModel?: (model: OpenAI.Model) => ChatModelCard;
|
||||
};
|
||||
provider: string;
|
||||
responses?: {
|
||||
handlePayload?: (
|
||||
@@ -205,6 +205,81 @@ export const createOpenAICompatibleRuntime = <T extends Record<string, any> = an
|
||||
this.logPrefix = `lobe-model-runtime:${this.id}`;
|
||||
}
|
||||
|
||||
/**
|
||||
* Determine if should use Responses API based on various configuration options
|
||||
* @param params - Configuration parameters
|
||||
* @returns true if should use Responses API, false otherwise
|
||||
*/
|
||||
private shouldUseResponsesAPI(params: {
|
||||
/** Context for logging (e.g., 'chat', 'generateObject', 'tool calling') */
|
||||
context?: string;
|
||||
/** Factory/instance level useResponse flag */
|
||||
flagUseResponse?: boolean;
|
||||
/** Factory/instance level model patterns for Responses API */
|
||||
flagUseResponseModels?: Array<string | RegExp>;
|
||||
/** The model ID to check */
|
||||
model?: string;
|
||||
/** Explicit responseApi flag */
|
||||
responseApi?: boolean;
|
||||
/** User-specified API mode (highest priority) */
|
||||
userApiMode?: string;
|
||||
}): boolean {
|
||||
const {
|
||||
model,
|
||||
userApiMode,
|
||||
responseApi,
|
||||
flagUseResponse,
|
||||
flagUseResponseModels,
|
||||
context = 'operation',
|
||||
} = params;
|
||||
|
||||
const log = debug(`${this.logPrefix}:shouldUseResponsesAPI`);
|
||||
|
||||
// Priority 1: User explicitly set apiMode via switch
|
||||
if (userApiMode === 'responses') {
|
||||
log('using Responses API: explicit userApiMode=%s', userApiMode);
|
||||
return true;
|
||||
}
|
||||
|
||||
// Priority 2: userApiMode is explicitly set to something else
|
||||
if (userApiMode !== undefined) {
|
||||
log('using Chat Completions API: userApiMode=%s', userApiMode);
|
||||
return false;
|
||||
}
|
||||
|
||||
// Priority 3: Explicit responseApi flag
|
||||
if (responseApi) {
|
||||
log('using Responses API: explicit responseApi flag for %s', context);
|
||||
return true;
|
||||
}
|
||||
|
||||
// Priority 4: Factory/instance level useResponse flag
|
||||
if (flagUseResponse) {
|
||||
log('using Responses API: flagUseResponse=true for %s', context);
|
||||
return true;
|
||||
}
|
||||
|
||||
// Priority 5: Check if model matches useResponseModels patterns
|
||||
if (model && flagUseResponseModels?.length) {
|
||||
const matches = flagUseResponseModels.some((m: string | RegExp) =>
|
||||
typeof m === 'string' ? model.includes(m) : (m as RegExp).test(model),
|
||||
);
|
||||
if (matches) {
|
||||
log('using Responses API: model %s matches useResponseModels config', model);
|
||||
return true;
|
||||
}
|
||||
}
|
||||
|
||||
// Priority 6: Check built-in responsesAPIModels
|
||||
if (model && responsesAPIModels.has(model)) {
|
||||
log('using Responses API: model %s in built-in responsesAPIModels', model);
|
||||
return true;
|
||||
}
|
||||
|
||||
log('using Chat Completions API for %s', context);
|
||||
return false;
|
||||
}
|
||||
|
||||
async chat({ responseMode, ...payload }: ChatStreamPayload, options?: ChatMethodOptions) {
|
||||
try {
|
||||
const log = debug(`${this.logPrefix}:chat`);
|
||||
@@ -212,41 +287,39 @@ export const createOpenAICompatibleRuntime = <T extends Record<string, any> = an
|
||||
|
||||
log('chat called with model: %s, stream: %s', payload.model, payload.stream ?? true);
|
||||
|
||||
// 工厂级 Responses API 路由控制（支持实例覆盖）
|
||||
const modelId = (payload as any).model as string | undefined;
|
||||
const shouldUseResponses = (() => {
|
||||
const instanceChat = ((this._options as any).chatCompletion || {}) as {
|
||||
useResponse?: boolean;
|
||||
useResponseModels?: Array<string | RegExp>;
|
||||
};
|
||||
const flagUseResponse =
|
||||
instanceChat.useResponse ?? (chatCompletion ? chatCompletion.useResponse : undefined);
|
||||
const flagUseResponseModels =
|
||||
instanceChat.useResponseModels ?? chatCompletion?.useResponseModels;
|
||||
|
||||
if (!chatCompletion && !instanceChat) return false;
|
||||
if (flagUseResponse) return true;
|
||||
if (!modelId || !flagUseResponseModels?.length) return false;
|
||||
return flagUseResponseModels.some((m: string | RegExp) =>
|
||||
typeof m === 'string' ? modelId.includes(m) : (m as RegExp).test(modelId),
|
||||
);
|
||||
})();
|
||||
|
||||
let processedPayload: any = payload;
|
||||
const userApiMode = (payload as any).apiMode as string | undefined;
|
||||
const modelId = (payload as any).model as string | undefined;
|
||||
|
||||
const instanceChat = ((this._options as any).chatCompletion || {}) as {
|
||||
useResponse?: boolean;
|
||||
useResponseModels?: Array<string | RegExp>;
|
||||
};
|
||||
const flagUseResponse =
|
||||
instanceChat.useResponse ?? (chatCompletion ? chatCompletion.useResponse : undefined);
|
||||
const flagUseResponseModels =
|
||||
instanceChat.useResponseModels ?? chatCompletion?.useResponseModels;
|
||||
|
||||
// Determine if should use Responses API
|
||||
const shouldUseResponses = this.shouldUseResponsesAPI({
|
||||
context: 'chat',
|
||||
flagUseResponse,
|
||||
flagUseResponseModels,
|
||||
model: modelId,
|
||||
userApiMode,
|
||||
});
|
||||
|
||||
if (shouldUseResponses) {
|
||||
log('using Responses API mode');
|
||||
processedPayload = { ...payload, apiMode: 'responses' } as any;
|
||||
} else {
|
||||
log('using Chat Completions API mode');
|
||||
}
|
||||
|
||||
// 再进行工厂级处理
|
||||
const postPayload = chatCompletion?.handlePayload
|
||||
? chatCompletion.handlePayload(processedPayload, this._options)
|
||||
: ({
|
||||
...processedPayload,
|
||||
stream: processedPayload.stream ?? true,
|
||||
} as OpenAI.ChatCompletionCreateParamsStreaming);
|
||||
...processedPayload,
|
||||
stream: processedPayload.stream ?? true,
|
||||
} as OpenAI.ChatCompletionCreateParamsStreaming);
|
||||
|
||||
if ((postPayload as any).apiMode === 'responses') {
|
||||
return this.handleResponseAPIMode(processedPayload, options);
|
||||
@@ -312,13 +385,13 @@ export const createOpenAICompatibleRuntime = <T extends Record<string, any> = an
|
||||
return StreamingResponse(
|
||||
chatCompletion?.handleStream
|
||||
? chatCompletion.handleStream(prod, {
|
||||
callbacks: streamOptions.callbacks,
|
||||
inputStartAt,
|
||||
})
|
||||
callbacks: streamOptions.callbacks,
|
||||
inputStartAt,
|
||||
})
|
||||
: OpenAIStream(prod, {
|
||||
...streamOptions,
|
||||
inputStartAt,
|
||||
}),
|
||||
...streamOptions,
|
||||
inputStartAt,
|
||||
}),
|
||||
{
|
||||
headers: options?.headers,
|
||||
},
|
||||
@@ -342,9 +415,9 @@ export const createOpenAICompatibleRuntime = <T extends Record<string, any> = an
|
||||
return StreamingResponse(
|
||||
chatCompletion?.handleStream
|
||||
? chatCompletion.handleStream(stream, {
|
||||
callbacks: streamOptions.callbacks,
|
||||
inputStartAt,
|
||||
})
|
||||
callbacks: streamOptions.callbacks,
|
||||
inputStartAt,
|
||||
})
|
||||
: OpenAIStream(stream, { ...streamOptions, enableStreaming: false, inputStartAt }),
|
||||
{
|
||||
headers: options?.headers,
|
||||
@@ -500,47 +573,23 @@ export const createOpenAICompatibleRuntime = <T extends Record<string, any> = an
|
||||
}
|
||||
|
||||
// Factory-level Responses API routing control (supports instance override)
|
||||
const shouldUseResponses = (() => {
|
||||
const instanceGenerateObject = ((this._options as any).generateObject || {}) as {
|
||||
useResponse?: boolean;
|
||||
useResponseModels?: Array<string | RegExp>;
|
||||
};
|
||||
const flagUseResponse =
|
||||
instanceGenerateObject.useResponse ??
|
||||
(generateObjectConfig ? generateObjectConfig.useResponse : undefined);
|
||||
const flagUseResponseModels =
|
||||
instanceGenerateObject.useResponseModels ?? generateObjectConfig?.useResponseModels;
|
||||
const instanceGenerateObject = ((this._options as any).generateObject || {}) as {
|
||||
useResponse?: boolean;
|
||||
useResponseModels?: Array<string | RegExp>;
|
||||
};
|
||||
const flagUseResponse =
|
||||
instanceGenerateObject.useResponse ??
|
||||
(generateObjectConfig ? generateObjectConfig.useResponse : undefined);
|
||||
const flagUseResponseModels =
|
||||
instanceGenerateObject.useResponseModels ?? generateObjectConfig?.useResponseModels;
|
||||
|
||||
if (responseApi) {
|
||||
log('using Responses API due to explicit responseApi flag');
|
||||
return true;
|
||||
}
|
||||
|
||||
if (flagUseResponse) {
|
||||
log('using Responses API due to useResponse flag');
|
||||
return true;
|
||||
}
|
||||
|
||||
// Use factory-configured model list if provided
|
||||
if (model && flagUseResponseModels?.length) {
|
||||
const matches = flagUseResponseModels.some((m: string | RegExp) =>
|
||||
typeof m === 'string' ? model.includes(m) : (m as RegExp).test(model),
|
||||
);
|
||||
if (matches) {
|
||||
log('using Responses API: model %s matches useResponseModels config', model);
|
||||
return true;
|
||||
}
|
||||
}
|
||||
|
||||
// Default: use built-in responsesAPIModels
|
||||
if (model && responsesAPIModels.has(model)) {
|
||||
log('using Responses API: model %s in built-in responsesAPIModels', model);
|
||||
return true;
|
||||
}
|
||||
|
||||
log('using Chat Completions API for generateObject');
|
||||
return false;
|
||||
})();
|
||||
const shouldUseResponses = this.shouldUseResponsesAPI({
|
||||
context: 'generateObject',
|
||||
flagUseResponse,
|
||||
flagUseResponseModels,
|
||||
model,
|
||||
responseApi,
|
||||
});
|
||||
|
||||
// Apply schema transformation if configured
|
||||
const processedSchema = generateObjectConfig?.handleSchema
|
||||
@@ -790,11 +839,11 @@ export const createOpenAICompatibleRuntime = <T extends Record<string, any> = an
|
||||
...res,
|
||||
...(reasoning || reasoning_effort
|
||||
? {
|
||||
reasoning: {
|
||||
...reasoning,
|
||||
...(reasoning_effort && { effort: reasoning_effort }),
|
||||
},
|
||||
}
|
||||
reasoning: {
|
||||
...reasoning,
|
||||
...(reasoning_effort && { effort: reasoning_effort }),
|
||||
},
|
||||
}
|
||||
: {}),
|
||||
input,
|
||||
...(max_tokens && { max_output_tokens: max_tokens }),
|
||||
@@ -885,47 +934,23 @@ export const createOpenAICompatibleRuntime = <T extends Record<string, any> = an
|
||||
);
|
||||
|
||||
// Factory-level Responses API routing control (supports instance override)
|
||||
const shouldUseResponses = (() => {
|
||||
const instanceGenerateObject = ((this._options as any).generateObject || {}) as {
|
||||
useResponse?: boolean;
|
||||
useResponseModels?: Array<string | RegExp>;
|
||||
};
|
||||
const flagUseResponse =
|
||||
instanceGenerateObject.useResponse ??
|
||||
(generateObjectConfig ? generateObjectConfig.useResponse : undefined);
|
||||
const flagUseResponseModels =
|
||||
instanceGenerateObject.useResponseModels ?? generateObjectConfig?.useResponseModels;
|
||||
const instanceGenerateObject = ((this._options as any).generateObject || {}) as {
|
||||
useResponse?: boolean;
|
||||
useResponseModels?: Array<string | RegExp>;
|
||||
};
|
||||
const flagUseResponse =
|
||||
instanceGenerateObject.useResponse ??
|
||||
(generateObjectConfig ? generateObjectConfig.useResponse : undefined);
|
||||
const flagUseResponseModels =
|
||||
instanceGenerateObject.useResponseModels ?? generateObjectConfig?.useResponseModels;
|
||||
|
||||
if (responseApi) {
|
||||
log('using Responses API due to explicit responseApi flag');
|
||||
return true;
|
||||
}
|
||||
|
||||
if (flagUseResponse) {
|
||||
log('using Responses API due to useResponse flag');
|
||||
return true;
|
||||
}
|
||||
|
||||
// Use factory-configured model list if provided
|
||||
if (model && flagUseResponseModels?.length) {
|
||||
const matches = flagUseResponseModels.some((m: string | RegExp) =>
|
||||
typeof m === 'string' ? model.includes(m) : (m as RegExp).test(model),
|
||||
);
|
||||
if (matches) {
|
||||
log('using Responses API: model %s matches useResponseModels config', model);
|
||||
return true;
|
||||
}
|
||||
}
|
||||
|
||||
// Default: use built-in responsesAPIModels
|
||||
if (model && responsesAPIModels.has(model)) {
|
||||
log('using Responses API: model %s in built-in responsesAPIModels', model);
|
||||
return true;
|
||||
}
|
||||
|
||||
log('using Chat Completions API for tool calling');
|
||||
return false;
|
||||
})();
|
||||
const shouldUseResponses = this.shouldUseResponsesAPI({
|
||||
context: 'tool calling',
|
||||
flagUseResponse,
|
||||
flagUseResponseModels,
|
||||
model,
|
||||
responseApi,
|
||||
});
|
||||
|
||||
if (shouldUseResponses) {
|
||||
log('calling responses.create for tool calling');
|
||||
|
||||
@@ -5,7 +5,7 @@ import { Mock, afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
|
||||
import { responsesAPIModels } from '../../const/models';
|
||||
import { ChatStreamPayload } from '../../types/chat';
|
||||
import * as modelParseModule from '../../utils/modelParse';
|
||||
import { LobeNewAPIAI, NewAPIModelCard, NewAPIPricing, handlePayload, params } from './index';
|
||||
import { LobeNewAPIAI, NewAPIModelCard, NewAPIPricing, params } from './index';
|
||||
|
||||
// Mock external dependencies
|
||||
vi.mock('../../utils/modelParse');
|
||||
@@ -701,78 +701,6 @@ describe('NewAPI Runtime - 100% Branch Coverage', () => {
|
||||
});
|
||||
});
|
||||
|
||||
describe('HandlePayload Function - Direct Testing', () => {
|
||||
beforeEach(() => {
|
||||
// Mock responsesAPIModels as a Set for testing
|
||||
(responsesAPIModels as any).has = vi.fn((model: string) => model === 'o1-pro');
|
||||
});
|
||||
|
||||
it('should add apiMode for models in responsesAPIModels set', () => {
|
||||
(responsesAPIModels as any).has = vi.fn((model: string) => model === 'o1-pro');
|
||||
|
||||
const payload: ChatStreamPayload = {
|
||||
model: 'o1-pro',
|
||||
messages: [{ role: 'user', content: 'test' }],
|
||||
temperature: 0.5,
|
||||
};
|
||||
|
||||
const result = handlePayload(payload);
|
||||
expect(result).toEqual({ ...payload, apiMode: 'responses' });
|
||||
});
|
||||
|
||||
it('should add apiMode for gpt- models', () => {
|
||||
(responsesAPIModels as any).has = vi.fn(() => false);
|
||||
|
||||
const payload: ChatStreamPayload = {
|
||||
model: 'gpt-4o',
|
||||
messages: [{ role: 'user', content: 'test' }],
|
||||
temperature: 0.5,
|
||||
};
|
||||
|
||||
const result = handlePayload(payload);
|
||||
expect(result).toEqual({ ...payload, apiMode: 'responses' });
|
||||
});
|
||||
|
||||
it('should add apiMode for o1 models', () => {
|
||||
(responsesAPIModels as any).has = vi.fn(() => false);
|
||||
|
||||
const payload: ChatStreamPayload = {
|
||||
model: 'o1-mini',
|
||||
messages: [{ role: 'user', content: 'test' }],
|
||||
temperature: 0.5,
|
||||
};
|
||||
|
||||
const result = handlePayload(payload);
|
||||
expect(result).toEqual({ ...payload, apiMode: 'responses' });
|
||||
});
|
||||
|
||||
it('should add apiMode for o3 models', () => {
|
||||
(responsesAPIModels as any).has = vi.fn(() => false);
|
||||
|
||||
const payload: ChatStreamPayload = {
|
||||
model: 'o3-turbo',
|
||||
messages: [{ role: 'user', content: 'test' }],
|
||||
temperature: 0.5,
|
||||
};
|
||||
|
||||
const result = handlePayload(payload);
|
||||
expect(result).toEqual({ ...payload, apiMode: 'responses' });
|
||||
});
|
||||
|
||||
it('should not modify payload for regular models', () => {
|
||||
(responsesAPIModels as any).has = vi.fn(() => false);
|
||||
|
||||
const payload: ChatStreamPayload = {
|
||||
model: 'claude-3-sonnet',
|
||||
messages: [{ role: 'user', content: 'test' }],
|
||||
temperature: 0.5,
|
||||
};
|
||||
|
||||
const result = handlePayload(payload);
|
||||
expect(result).toEqual(payload);
|
||||
});
|
||||
});
|
||||
|
||||
describe('Routers Function - Direct Testing', () => {
|
||||
it('should generate routers with correct apiTypes', () => {
|
||||
const options = { apiKey: 'test', baseURL: 'https://api.newapi.com/v1' };
|
||||
@@ -823,11 +751,11 @@ describe('NewAPI Runtime - 100% Branch Coverage', () => {
|
||||
expect(routers[3].options.baseURL).toBe('https://custom.com/v1');
|
||||
});
|
||||
|
||||
it('should configure openai router with handlePayload', () => {
|
||||
it('should configure openai router with useResponseModels', () => {
|
||||
const options = { apiKey: 'test', baseURL: 'https://custom.com/v1' };
|
||||
const routers = params.routers(options);
|
||||
|
||||
expect((routers[3].options as any).chatCompletion?.handlePayload).toBe(handlePayload);
|
||||
expect((routers[3].options as any).chatCompletion?.useResponseModels).toBeDefined();
|
||||
});
|
||||
|
||||
it('should filter anthropic models for anthropic router', () => {
|
||||
|
||||
@@ -4,7 +4,6 @@ import urlJoin from 'url-join';
|
||||
import { responsesAPIModels } from '../../const/models';
|
||||
import { createRouterRuntime } from '../../core/RouterRuntime';
|
||||
import { CreateRouterRuntimeOptions } from '../../core/RouterRuntime/createRuntime';
|
||||
import { ChatStreamPayload } from '../../types/chat';
|
||||
import { detectModelProvider, processMultiProviderModelList } from '../../utils/modelParse';
|
||||
|
||||
export interface NewAPIModelCard {
|
||||
@@ -26,18 +25,6 @@ export interface NewAPIPricing {
|
||||
supported_endpoint_types?: string[];
|
||||
}
|
||||
|
||||
export const handlePayload = (payload: ChatStreamPayload) => {
|
||||
// Handle OpenAI responses API mode
|
||||
if (
|
||||
responsesAPIModels.has(payload.model) ||
|
||||
payload.model.includes('gpt-') ||
|
||||
/^o\d/.test(payload.model)
|
||||
) {
|
||||
return { ...payload, apiMode: 'responses' };
|
||||
}
|
||||
return payload;
|
||||
};
|
||||
|
||||
export const params = {
|
||||
debug: {
|
||||
chatCompletion: () => process.env.DEBUG_NEWAPI_CHAT_COMPLETION === '1',
|
||||
@@ -178,7 +165,7 @@ export const params = {
|
||||
...options,
|
||||
baseURL: urlJoin(userBaseURL, '/v1'),
|
||||
chatCompletion: {
|
||||
handlePayload,
|
||||
useResponseModels: [...Array.from(responsesAPIModels), /gpt-\d(?!\d)/, /^o\d/],
|
||||
},
|
||||
},
|
||||
},
|
||||
|
||||
@@ -1548,8 +1548,9 @@ describe('LobeOpenRouterAI - custom features', () => {
|
||||
const models = await params.models();
|
||||
|
||||
const mixedModel = models.find((m) => m.id === 'mixed-free/model');
|
||||
// Input or output is 0, so should be marked as free
|
||||
expect(mixedModel?.displayName).toContain('(free)');
|
||||
// Input or output is 0. Current behavior does not append '(free)' for mixed pricing,
|
||||
// so assert the displayName equals the cleaned model name.
|
||||
expect(mixedModel?.displayName).toBe('Mixed Free Model');
|
||||
});
|
||||
|
||||
it('should handle very large pricing values', async () => {
|
||||
|
||||
@@ -99,7 +99,7 @@ export const params = {
|
||||
const cachedInputPrice = formatPrice(pricing.input_cache_read);
|
||||
const writeCacheInputPrice = formatPrice(pricing.input_cache_write);
|
||||
|
||||
const isFree = (inputPrice === 0 || outputPrice === 0) && !displayName.endsWith('(free)');
|
||||
const isFree = inputPrice === 0 && outputPrice === 0 && !displayName.endsWith('(free)');
|
||||
if (isFree) {
|
||||
displayName += ' (free)';
|
||||
}
|
||||
|
||||
@@ -42,6 +42,15 @@ const CreateNewProvider = memo<CreateNewProviderProps>(({ onClose, open }) => {
|
||||
name: values.name || values.id,
|
||||
};
|
||||
|
||||
// 只为 openai 和 router (newapi) 类型的自定义 provider 添加 supportResponsesApi: true
|
||||
const sdkType = values.settings?.sdkType;
|
||||
if (sdkType === 'openai' || sdkType === 'router') {
|
||||
finalValues.settings = {
|
||||
...finalValues.settings,
|
||||
supportResponsesApi: true,
|
||||
};
|
||||
}
|
||||
|
||||
await createNewAiProvider(finalValues);
|
||||
setLoading(false);
|
||||
navigate(`/settings?active=provider&provider=${values.id}`);
|
||||
@@ -102,12 +111,16 @@ const CreateNewProvider = memo<CreateNewProviderProps>(({ onClose, open }) => {
|
||||
{
|
||||
children: (
|
||||
<Select
|
||||
optionRender={({ label, value }) => (
|
||||
<Flexbox align={'center'} gap={8} horizontal>
|
||||
<ProviderIcon provider={value as string} size={18} />
|
||||
{label}
|
||||
</Flexbox>
|
||||
)}
|
||||
optionRender={({ label, value }) => {
|
||||
// Map 'router' to 'newapi' for displaying the correct icon
|
||||
const iconProvider = value === 'router' ? 'newapi' : (value as string);
|
||||
return (
|
||||
<Flexbox align={'center'} gap={8} horizontal>
|
||||
<ProviderIcon provider={iconProvider} size={18} />
|
||||
{label}
|
||||
</Flexbox>
|
||||
);
|
||||
}}
|
||||
options={CUSTOM_PROVIDER_SDK_OPTIONS}
|
||||
placeholder={t('createNewAiProvider.sdkType.placeholder')}
|
||||
variant={'filled'}
|
||||
|
||||
@@ -9,4 +9,5 @@ export const CUSTOM_PROVIDER_SDK_OPTIONS = [
|
||||
{ label: 'Qwen', value: 'qwen' },
|
||||
{ label: 'Volcengine', value: 'volcengine' },
|
||||
{ label: 'Ollama', value: 'ollama' },
|
||||
{ label: 'New API', value: 'router' },
|
||||
] satisfies { label: string; value: AiProviderSDKType }[];
|
||||
|
||||
@@ -11,6 +11,7 @@ const AiHubMix: ModelProviderCard = {
|
||||
settings: {
|
||||
sdkType: 'router',
|
||||
showModelFetcher: true,
|
||||
supportResponsesApi: true,
|
||||
},
|
||||
url: 'https://aihubmix.com?utm_source=lobehub',
|
||||
};
|
||||
|
||||
@@ -13,6 +13,7 @@ const NewAPI: ModelProviderCard = {
|
||||
},
|
||||
sdkType: 'router',
|
||||
showModelFetcher: true,
|
||||
supportResponsesApi: true,
|
||||
},
|
||||
url: 'https://github.com/Calcium-Ion/new-api',
|
||||
};
|
||||
|
||||
@@ -298,7 +298,7 @@ export default {
|
||||
},
|
||||
helpDoc: '配置教程',
|
||||
responsesApi: {
|
||||
desc: '采用 OpenAI 新一代请求格式规范，解锁思维链等进阶特性',
|
||||
desc: '采用 OpenAI 新一代请求格式规范，解锁思维链等进阶特性 (仅 OpenAI 模型支持)',
|
||||
title: '使用 Responses API 规范',
|
||||
},
|
||||
waitingForMore: '更多模型正在 <1>计划接入</1> 中，敬请期待',
|
||||
|
||||
Reference in New Issue
Block a user