Skip to content

Commit 669463d

Browse files
committed
✨ feat: add vertex ai route
1 parent 66c820b commit 669463d

File tree

11 files changed

+316
-14
lines changed

11 files changed

+316
-14
lines changed

package.json

+1
Original file line numberDiff line numberDiff line change
@@ -117,6 +117,7 @@
117117
"@clerk/themes": "^2.1.37",
118118
"@codesandbox/sandpack-react": "^2.19.9",
119119
"@cyntler/react-doc-viewer": "^1.17.0",
120+
"@google-cloud/vertexai": "^1.9.0",
120121
"@google/generative-ai": "^0.21.0",
121122
"@huggingface/inference": "^2.8.1",
122123
"@icons-pack/react-simple-icons": "9.6.0",
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,38 @@
1+
import { readFileSync } from 'node:fs';
2+
import { resolve } from 'node:path';
3+
4+
import { getLLMConfig } from '@/config/llm';
5+
import { AgentRuntime, ModelProvider } from '@/libs/agent-runtime';
6+
import { LobeVertexAI } from '@/libs/agent-runtime/vertexai';
7+
import { safeParseJSON } from '@/utils/safeParseJSON';
8+
9+
import { POST as UniverseRoute } from '../[provider]/route';
10+
11+
export const POST = async (req: Request) =>
12+
UniverseRoute(req, {
13+
createRuntime: () => {
14+
const {
15+
VERTEXAI_PROJECT,
16+
VERTEXAI_LOCATION,
17+
VERTEXAI_CREDENTIALS,
18+
VERTEXAI_CREDENTIALS_PATH,
19+
} = getLLMConfig();
20+
21+
const credentialsContent =
22+
VERTEXAI_CREDENTIALS ??
23+
(VERTEXAI_CREDENTIALS_PATH
24+
? readFileSync(resolve(process.cwd(), VERTEXAI_CREDENTIALS_PATH), 'utf8')
25+
: undefined);
26+
27+
const googleAuthOptions = credentialsContent ? safeParseJSON(credentialsContent) : undefined;
28+
29+
const instance = LobeVertexAI.initFromVertexAI({
30+
googleAuthOptions: googleAuthOptions,
31+
location: VERTEXAI_LOCATION,
32+
project: VERTEXAI_PROJECT,
33+
});
34+
35+
return new AgentRuntime(instance);
36+
},
37+
params: { provider: ModelProvider.VertexAI },
38+
});

src/app/(main)/settings/llm/ProviderList/providers.tsx

+3-1
Original file line numberDiff line numberDiff line change
@@ -26,6 +26,7 @@ import {
2626
TogetherAIProviderCard,
2727
UpstageProviderCard,
2828
XAIProviderCard,
29+
VertexAIProviderCard,
2930
ZeroOneProviderCard,
3031
ZhiPuProviderCard,
3132
} from '@/config/modelProviders';
@@ -38,8 +39,8 @@ import { useGithubProvider } from './Github';
3839
import { useHuggingFaceProvider } from './HuggingFace';
3940
import { useOllamaProvider } from './Ollama';
4041
import { useOpenAIProvider } from './OpenAI';
41-
import { useWenxinProvider } from './Wenxin';
4242
import { useSenseNovaProvider } from './SenseNova';
43+
import { useWenxinProvider } from './Wenxin';
4344

4445
export const useProviderList = (): ProviderItem[] => {
4546
const AzureProvider = useAzureProvider();
@@ -60,6 +61,7 @@ export const useProviderList = (): ProviderItem[] => {
6061
AnthropicProviderCard,
6162
BedrockProvider,
6263
GoogleProviderCard,
64+
VertexAIProviderCard,
6365
DeepSeekProviderCard,
6466
HuggingFaceProvider,
6567
OpenRouterProviderCard,

src/config/llm.ts

+17-1
Original file line numberDiff line numberDiff line change
@@ -24,6 +24,13 @@ export const getLLMConfig = () => {
2424
ENABLED_GOOGLE: z.boolean(),
2525
GOOGLE_API_KEY: z.string().optional(),
2626

27+
ENABLED_VERTEXAI: z.boolean(),
28+
VERTEXAI_CREDENTIALS: z.string().optional(),
29+
VERTEXAI_CREDENTIALS_PATH: z.string().optional(),
30+
VERTEXAI_PROJECT: z.string().optional(),
31+
VERTEXAI_LOCATION: z.string().optional(),
32+
VERTEXAI_MODEL_LIST: z.string().optional(),
33+
2734
ENABLED_MOONSHOT: z.boolean(),
2835
MOONSHOT_API_KEY: z.string().optional(),
2936

@@ -142,6 +149,14 @@ export const getLLMConfig = () => {
142149
ENABLED_GOOGLE: !!process.env.GOOGLE_API_KEY,
143150
GOOGLE_API_KEY: process.env.GOOGLE_API_KEY,
144151

152+
ENABLED_VERTEXAI:
153+
!!process.env.VERTEXAI_CREDENTIALS_PATH || !!process.env.VERTEXAI_CREDENTIALS,
154+
VERTEXAI_CREDENTIALS_PATH: process.env.VERTEXAI_CREDENTIALS_PATH,
155+
VERTEXAI_CREDENTIALS: process.env.VERTEXAI_CREDENTIALS,
156+
VERTEXAI_LOCATION: process.env.VERTEXAI_LOCATION,
157+
VERTEXAI_PROJECT: process.env.VERTEXAI_PROJECT,
158+
VERTEXAI_MODEL_LIST: process.env.VERTEXAI_MODEL_LIST,
159+
145160
ENABLED_PERPLEXITY: !!process.env.PERPLEXITY_API_KEY,
146161
PERPLEXITY_API_KEY: process.env.PERPLEXITY_API_KEY,
147162

@@ -231,7 +246,8 @@ export const getLLMConfig = () => {
231246
ENABLED_HUGGINGFACE: !!process.env.HUGGINGFACE_API_KEY,
232247
HUGGINGFACE_API_KEY: process.env.HUGGINGFACE_API_KEY,
233248

234-
ENABLED_SENSENOVA: !!process.env.SENSENOVA_ACCESS_KEY_ID && !!process.env.SENSENOVA_ACCESS_KEY_SECRET,
249+
ENABLED_SENSENOVA:
250+
!!process.env.SENSENOVA_ACCESS_KEY_ID && !!process.env.SENSENOVA_ACCESS_KEY_SECRET,
235251
SENSENOVA_ACCESS_KEY_ID: process.env.SENSENOVA_ACCESS_KEY_ID,
236252
SENSENOVA_ACCESS_KEY_SECRET: process.env.SENSENOVA_ACCESS_KEY_SECRET,
237253

src/config/modelProviders/index.ts

+4
Original file line numberDiff line numberDiff line change
@@ -32,6 +32,7 @@ import StepfunProvider from './stepfun';
3232
import TaichuProvider from './taichu';
3333
import TogetherAIProvider from './togetherai';
3434
import UpstageProvider from './upstage';
35+
import VertexAIProvider from './vertexai';
3536
import WenxinProvider from './wenxin';
3637
import XAIProvider from './xai';
3738
import ZeroOneProvider from './zeroone';
@@ -68,6 +69,7 @@ export const LOBE_DEFAULT_MODEL_LIST: ChatModelCard[] = [
6869
GiteeAIProvider.chatModels,
6970
UpstageProvider.chatModels,
7071
SparkProvider.chatModels,
72+
VertexAIProvider.chatModels,
7173
Ai21Provider.chatModels,
7274
HunyuanProvider.chatModels,
7375
WenxinProvider.chatModels,
@@ -112,6 +114,7 @@ export const DEFAULT_MODEL_PROVIDER_LIST = [
112114
InternLMProvider,
113115
SiliconCloudProvider,
114116
GiteeAIProvider,
117+
VertexAIProvider,
115118
];
116119

117120
export const filterEnabledModels = (provider: ModelProviderCard) => {
@@ -155,6 +158,7 @@ export { default as StepfunProviderCard } from './stepfun';
155158
export { default as TaichuProviderCard } from './taichu';
156159
export { default as TogetherAIProviderCard } from './togetherai';
157160
export { default as UpstageProviderCard } from './upstage';
161+
export { default as VertexAIProviderCard } from './vertexai';
158162
export { default as WenxinProviderCard } from './wenxin';
159163
export { default as XAIProviderCard } from './xai';
160164
export { default as ZeroOneProviderCard } from './zeroone';

src/config/modelProviders/vertexai.ts

+217
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,217 @@
1+
import { ModelProviderCard } from '@/types/llm';
2+
3+
// ref: https://ai.google.dev/gemini-api/docs/models/gemini
4+
const VertexAI: ModelProviderCard = {
5+
chatModels: [
6+
{
7+
description:
8+
'Gemini 1.5 Flash 是Google最新的多模态AI模型,具备快速处理能力,支持文本、图像和视频输入,适用于多种任务的高效扩展。',
9+
displayName: 'Gemini 1.5 Flash',
10+
enabled: true,
11+
functionCall: true,
12+
id: 'gemini-1.5-flash-latest',
13+
maxOutput: 8192,
14+
pricing: {
15+
cachedInput: 0.018_75,
16+
input: 0.075,
17+
output: 0.3,
18+
},
19+
tokens: 1_000_000 + 8192,
20+
vision: true,
21+
},
22+
{
23+
description: 'Gemini 1.5 Flash 002 是一款高效的多模态模型,支持广泛应用的扩展。',
24+
displayName: 'Gemini 1.5 Flash 002',
25+
enabled: true,
26+
functionCall: true,
27+
id: 'gemini-1.5-flash-002',
28+
maxOutput: 8192,
29+
pricing: {
30+
cachedInput: 0.018_75,
31+
input: 0.075,
32+
output: 0.3,
33+
},
34+
releasedAt: '2024-09-25',
35+
tokens: 1_000_000 + 8192,
36+
vision: true,
37+
},
38+
{
39+
description: 'Gemini 1.5 Flash 001 是一款高效的多模态模型,支持广泛应用的扩展。',
40+
displayName: 'Gemini 1.5 Flash 001',
41+
functionCall: true,
42+
id: 'gemini-1.5-flash-001',
43+
maxOutput: 8192,
44+
pricing: {
45+
cachedInput: 0.018_75,
46+
input: 0.075,
47+
output: 0.3,
48+
},
49+
tokens: 1_000_000 + 8192,
50+
vision: true,
51+
},
52+
{
53+
description: 'Gemini 1.5 Flash 0827 提供了优化后的多模态处理能力,适用多种复杂任务场景。',
54+
displayName: 'Gemini 1.5 Flash 0827',
55+
functionCall: true,
56+
id: 'gemini-1.5-flash-exp-0827',
57+
maxOutput: 8192,
58+
pricing: {
59+
cachedInput: 0.018_75,
60+
input: 0.075,
61+
output: 0.3,
62+
},
63+
releasedAt: '2024-08-27',
64+
tokens: 1_000_000 + 8192,
65+
vision: true,
66+
},
67+
68+
{
69+
description:
70+
'Gemini 1.5 Flash 8B 0924 是最新的实验性模型,在文本和多模态用例中都有显著的性能提升。',
71+
displayName: 'Gemini 1.5 Flash 8B 0924',
72+
functionCall: true,
73+
id: 'gemini-1.5-flash-8b-exp-0924',
74+
maxOutput: 8192,
75+
pricing: {
76+
cachedInput: 0.018_75,
77+
input: 0.075,
78+
output: 0.3,
79+
},
80+
releasedAt: '2024-09-24',
81+
tokens: 1_000_000 + 8192,
82+
vision: true,
83+
},
84+
{
85+
description:
86+
'Gemini 1.5 Pro 支持高达200万个tokens,是中型多模态模型的理想选择,适用于复杂任务的多方面支持。',
87+
displayName: 'Gemini 1.5 Pro',
88+
enabled: true,
89+
functionCall: true,
90+
id: 'gemini-1.5-pro-latest',
91+
maxOutput: 8192,
92+
pricing: {
93+
cachedInput: 0.875,
94+
input: 3.5,
95+
output: 10.5,
96+
},
97+
releasedAt: '2024-02-15',
98+
tokens: 2_000_000 + 8192,
99+
vision: true,
100+
},
101+
{
102+
description:
103+
'Gemini 1.5 Pro 002 是最新的生产就绪模型,提供更高质量的输出,特别在数学、长上下文和视觉任务方面有显著提升。',
104+
displayName: 'Gemini 1.5 Pro 002',
105+
enabled: true,
106+
functionCall: true,
107+
id: 'gemini-1.5-pro-002',
108+
maxOutput: 8192,
109+
pricing: {
110+
cachedInput: 0.315,
111+
input: 1.25,
112+
output: 2.5,
113+
},
114+
releasedAt: '2024-09-24',
115+
tokens: 2_000_000 + 8192,
116+
vision: true,
117+
},
118+
{
119+
description: 'Gemini 1.5 Pro 001 是可扩展的多模态AI解决方案,支持广泛的复杂任务。',
120+
displayName: 'Gemini 1.5 Pro 001',
121+
functionCall: true,
122+
id: 'gemini-1.5-pro-001',
123+
maxOutput: 8192,
124+
pricing: {
125+
cachedInput: 0.875,
126+
input: 3.5,
127+
output: 10.5,
128+
},
129+
releasedAt: '2024-02-15',
130+
tokens: 2_000_000 + 8192,
131+
vision: true,
132+
},
133+
{
134+
description: 'Gemini 1.5 Pro 0827 结合最新优化技术,带来更高效的多模态数据处理能力。',
135+
displayName: 'Gemini 1.5 Pro 0827',
136+
functionCall: true,
137+
id: 'gemini-1.5-pro-exp-0827',
138+
maxOutput: 8192,
139+
pricing: {
140+
cachedInput: 0.875,
141+
input: 3.5,
142+
output: 10.5,
143+
},
144+
releasedAt: '2024-08-27',
145+
tokens: 2_000_000 + 8192,
146+
vision: true,
147+
},
148+
{
149+
description: 'Gemini 1.5 Pro 0801 提供出色的多模态处理能力,为应用开发带来更大灵活性。',
150+
displayName: 'Gemini 1.5 Pro 0801',
151+
functionCall: true,
152+
id: 'gemini-1.5-pro-exp-0801',
153+
maxOutput: 8192,
154+
pricing: {
155+
cachedInput: 0.875,
156+
input: 3.5,
157+
output: 10.5,
158+
},
159+
releasedAt: '2024-08-01',
160+
tokens: 2_000_000 + 8192,
161+
vision: true,
162+
},
163+
{
164+
description: 'Gemini 1.0 Pro 是Google的高性能AI模型,专为广泛任务扩展而设计。',
165+
displayName: 'Gemini 1.0 Pro',
166+
id: 'gemini-1.0-pro-latest',
167+
maxOutput: 2048,
168+
pricing: {
169+
input: 0.5,
170+
output: 1.5,
171+
},
172+
releasedAt: '2023-12-06',
173+
tokens: 30_720 + 2048,
174+
},
175+
{
176+
description:
177+
'Gemini 1.0 Pro 001 (Tuning) 提供稳定并可调优的性能,是复杂任务解决方案的理想选择。',
178+
displayName: 'Gemini 1.0 Pro 001 (Tuning)',
179+
functionCall: true,
180+
id: 'gemini-1.0-pro-001',
181+
maxOutput: 2048,
182+
pricing: {
183+
input: 0.5,
184+
output: 1.5,
185+
},
186+
releasedAt: '2023-12-06',
187+
tokens: 30_720 + 2048,
188+
},
189+
{
190+
description: 'Gemini 1.0 Pro 002 (Tuning) 提供出色的多模态支持,专注于复杂任务的有效解决。',
191+
displayName: 'Gemini 1.0 Pro 002 (Tuning)',
192+
id: 'gemini-1.0-pro-002',
193+
maxOutput: 2048,
194+
pricing: {
195+
input: 0.5,
196+
output: 1.5,
197+
},
198+
releasedAt: '2023-12-06',
199+
tokens: 30_720 + 2048,
200+
},
201+
],
202+
checkModel: 'gemini-1.5-flash-latest',
203+
description:
204+
'Google 的 Gemini 系列是其最先进、通用的 AI模型,由 Google DeepMind 打造,专为多模态设计,支持文本、代码、图像、音频和视频的无缝理解与处理。适用于从数据中心到移动设备的多种环境,极大提升了AI模型的效率与应用广泛性。',
205+
id: 'vertexai',
206+
modelsUrl: 'https://cloud.google.com/vertex-ai/generative-ai/docs/learn/model-versioning',
207+
name: 'VertexAI',
208+
showApiKey: false,
209+
showChecker: false,
210+
smoothing: {
211+
speed: 2,
212+
text: true,
213+
},
214+
url: 'https://cloud.google.com/vertex-ai',
215+
};
216+
217+
export default VertexAI;

src/libs/agent-runtime/error.ts

+1
Original file line numberDiff line numberDiff line change
@@ -13,6 +13,7 @@ export const AgentRuntimeErrorType = {
1313
OllamaBizError: 'OllamaBizError',
1414

1515
InvalidBedrockCredentials: 'InvalidBedrockCredentials',
16+
InvalidVertexCredentials: 'InvalidVertexCredentials',
1617
StreamChunkError: 'StreamChunkError',
1718

1819
InvalidGithubToken: 'InvalidGithubToken',

0 commit comments

Comments
 (0)