preface
In the field of AI application development, DeepSeek, as an advanced large language model, provides powerful natural language processing capabilities. This article will detail how to integrate the DeepSeek API in the HarmonyOS to implement AI conversational capabilities. Whether you are a HarmonyOS app developer or an AI technology enthusiast, you can quickly master the integration method with this tutorial.
Environment preparation
Development tools
DevEco Studio 3.1 or later (official IDE for HarmonyOS App Development)
HarmonyOS SDK API 9 or later
Account preparation
Visit the official DeepSeek platform (https://api.deepseek.com)
Sign up for a developer account and create an app
Get the API key in the console (sk-xxx format)
Description of the project structure
First, create a tool folder in the HarmonyOS project, and the example structure in this document is as follows:
plaintext
src/main/ets/
├── utils/
│ └── AIRequest_Util.ts # ai request util
├── pages/
│ └── testHttp.ets # testPage
└── entry/
└── main_pages.json # page router
Full code implementation
- AI request util（AIRequest_Util.ts）
typescript
import http from '@ohos.net.http';
import hilog from '@ohos.hilog';
/**
DeepSeek AI API请求工具类
封装HTTP请求逻辑与响应处理
*/
export class AIRequest_Util {
/**
*/
request(question: string, callback: (text: string) => void) {
hilog.info(0x0000, 'testTag', 'ALiYunHttpUtils request invoke. question: %{public}s', question);
let httpRequest = http.createHttp();
const requestData = {
model: "deepseek-chat",
messages: [
{ role: "system", content: "你是人工智能AI" },
{ role: "user", content: question }
],
"stream": false
};
httpRequest.request(
"https://api.deepseek.com/chat/completions", // DeepSeek API
{
method: http.RequestMethod.POST,
header: {
"Content-Type": "application/json",
"Authorization": "Bearer sk-3e4695d9f07c454c8cb9d59822d27987" // APIkey
},
extraData: JSON.stringify(requestData),
readTimeout: 60000,
connectTimeout: 60000,
usingProtocol: http.HttpProtocol.HTTP1_1
}, (err, data: http.HttpResponse) => {
if (err) {
hilog.error(0x0000, 'testTag', '请求失败: %{public}s', JSON.stringify(err));
httpRequest.destroy();
callback("请求失败，请检查网络");
return;
}
try {
// 5. 解析响应数据
const response = JSON.parse(data.result as string);
hilog.info(0x0000, 'testTag', '完整响应: %{public}s', JSON.stringify(response));
// 检查API错误
if (response.error) {
hilog.error(0x0000, 'testTag', 'API错误: %{public}s', response.error.message);
callback(`API错误: ${response.error.message}`);
return;
}
// 提取AI回答内容
if (response.choices && response.choices.length > 0) {
const content = response.choices[0].message.content;
if (content) {
hilog.info(0x0000, 'testTag', '提取到的文本: %{public}s', content);
callback(content);
return;
}
}
throw new Error("未找到有效响应内容");
} catch (e) {
hilog.error(0x0000, 'testTag', '解析失败: %{public}s', e.message);
callback("解析响应失败");
} finally {
// 释放资源
httpRequest.destroy();
}
}
)
}
}
// 导出单例对象
export const DeepSeek = new AIRequest_Util();
- page（testHttp.ets）
typescript
import { DeepSeek } from '../utils/AIRequest_Util'; // 导入工具类
@Entry
@Component
struct testHttp {
@State answer: string = '' // 存储AI回答
build() {
Column() {
// 交互按钮
Button('click')
.width('50%')
.onClick(() => {
// 调用DeepSeek API
DeepSeek.request("你是谁", (answer) => {
// 更新UI显示回答
this.answer = answer
})
})
// 显示回答内容
Text(this.answer)
.fontSize(30)
.margin({ top: 20 })
}
.height('100%')
.width('100%')
.justifyContent(FlexAlign.Center)
.alignItems(HorizontalAlign.Center)
}
}
- page router（main_pages.json）
json
{
"src": [
{
"pages": [
"pages/testHttp"
],
"name": "entry",
"window": {
"designWidth": 720,
"autoDesignWidth": true
}
}
]
Analysis of functional modules
- The core process of API requests
The DeepSeek API call follows these steps:
Create an instance of the HTTP request
Build an OpenAI-compliant request body (DeepSeek-compatible OpenAI interface specification)
Set the request header (including the API key and content type)
Send a POST request to the DeepSeek API endpoint
Process response data and extract AI responses
- Request parameters
Parameter Name, Type, and description
model string is the name of the model, and deepseek-chat is the dialogue model
Messages array, which contains system prompts and user questions
messages.role string is the message role, system is the system prompt, user is the user input, and assistant is the AI answer
messages.content string The content of the message
Whether stream boolean enables streaming response, false is off (used in this example)
- Error Handling Mechanism
Three layers of error handling are implemented in the code:
Network request errors (e.g., timeouts, connection failures)
API interface errors (e.g., authentication failures, parameter errors)
Response parsing errors (e.g., data format is abnormal)
Results:
Problems you may encounter:
05-31 23:37:56.504 40524-28380 A00000/testTag E API error: Insufficient Balance
If the balance of the API key is insufficient, go to the official website to charge money
Top comments (0)