ChatGPT has become popular recently, and the boss is also ready to ask us to connect ChatGPT to Feishu. After a morning of research, we finally successfully registered and connected the Feishu robot to ChatGPT.
Let me share with you the detailed steps to access Feishu.
The interaction between Feishu and chatgpt is as follows. Our custom service acts as a middleman to forward messages.
1. Enter the Feishu open platform and choose to create a self-built application for the enterprise.
#2. After creating the application, click to enter the application and add the robot.
#3. Configure message-related permissions for the robot. If you are not sure what permissions are required, you can enable them all first.
4. Configure event subscription. Event subscription requires first developing an interface for Feishu to verify. The interface needs to be accessible from the public network.
The code of this interface can be referred to as follows:
@PostMapping(value = "/message")
public FeishuEventDTO message(@RequestBody String body) {
log.info("收到消息：{}", body);
FeishuEventParams feishuEventParams = JSON.parseObject(body, FeishuEventParams.class);
FeishuEventDTO eventDTO = new FeishuEventDTO();
eventDTO.setChallenge(feishuEventParams.getChallenge());
return eventDTO;
}
@Data
public class FeishuEventParams {
private String challenge;
private String token;
private String type;
}
@Data
public class FeishuEventDTO {
private String challenge;
}One thing to note is that this verification interface and the following receive Feishu messages The interface is the same address, but the message body is different.
That is to say, the verification interface is one-time use, and this interface needs to be modified after the verification is completed.
We first publish this interface to a project that can be accessed from the public network. For example, the interface address is
https://www.php.cn/link/4aee31b0ec9f7bb7885473d95961e9a6
OK, now the configuration of Feishu is basically completed. Here is the logic we need to process.
Let me first talk about the general logic of our company’s docking for your reference.
After a user sends a message to Feishu, Feishu will forward the message to our own service.
But there will be a problem here, that is, when multiple users initiate conversations concurrently, or when many people in a large group are @our robot, we need to remember everyone's reply and query it in chatgpt After getting the result, reply to this person accurately.
Since our company is currently also using it for internal testing and does not want to implement too complicated, the idea we adopt is: when each user's session is forwarded to our service, the session content is first saved to a global ConcurrentLinkedQueue queue, and then start a thread to continuously consume this queue.
The generic type of queue is an object constructed in advance. This object stores the message ID of the current message, the sender, the question content, etc.
Every time an object is consumed, the object's question content is sent to chatgpt. After obtaining the response result, the session reply interface provided by Feishu is called to reply to the user. (If the amount of concurrency is relatively large, it can be made asynchronous here).
Okay, that’s the general idea, let’s take a look at the specific code.
1. Open our project and introduce the jar provided by chatgpt.
<dependency> <groupId>com.theokanning.openai-gpt3-java</groupId> <artifactId>service</artifactId> <version>0.10.0</version> </dependency>
2, rewrite the above verification interface and change it to receive Feishu messages. (Do not change the interface path)
@Slf4j
@RestController
@RequestMapping(value = "/query")
public class QureyController {
public static ConcurrentLinkedQueue<FeishuResponse> consumer
= new ConcurrentLinkedQueue<>();
@PostMapping(value = "/message")
public String message(@RequestBody String body) {
log.info("收到飞书消息:{}", body);
JSONObject jsonObject = JSONObject.parseObject(body);
JSONObject header = jsonObject.getJSONObject("header");
String eventType = header.getString("event_type");
if ("im.message.receive_v1".equals(eventType)) {
JSONObject event = jsonObject.getJSONObject("event");
JSONObject message = event.getJSONObject("message");
String messageType = message.getString("message_type");
if ("text".equals(messageType)) {
String messageId = message.getString("message_id");
String content = message.getString("content");
JSONObject contentJson = JSON.parseObject(content);
String text = contentJson.getString("text");
FeishuResponse feishuResponse = new FeishuResponse();
feishuResponse.setMessageId(messageId);
feishuResponse.setQuery(text);
log.info("投递用户消息,{}", JSON.toJSON(feishuResponse));
consumer.add(feishuResponse);
} else {
log.info("非文本消息");
}
}
return "suc";
}
}The structure of FeishuResponse is as follows.
@Data
public class FeishuResponse {
private String messageId;
private String query;
}3, write a task thread.
@Slf4j
public class AutoSendTask implements Runnable {
//你的chatgpt的key
public static final String token = "";
public static OpenAiService openAiService = null;
static {
openAiService = new OpenAiService(token, Duration.ofSeconds(60));
}
@Override
public void run() {
while (true) {
try {
FeishuResponse poll = consumer.poll();
if (poll == null) {
log.info("no query,sleep 2s");
TimeUnit.SECONDS.sleep(2);
} else {
String query = this.query(poll.getQuery());
this.reply(poll, query);
}
} catch (InterruptedException e) {
log.error("Thread exception...", e);
}
}
}
private String query(String q) {
log.info("开始提问:{}", q);
CompletionRequest completionRequest = CompletionRequest.builder()
.prompt(q)
.model("text-davinci-003")
.maxTokens(2048)
.echo(false)
.build();
StringBuilder sb = new StringBuilder();
CompletionResult completion = openAiService.createCompletion(completionRequest);
log.info("q:{},获取响应:{}", q, JSON.toJSONString(completion));
completion.getChoices().forEach(v -> {
sb.append(v.getText());
});
String rs = sb.toString();
if (rs.startsWith("？")) {
rs = rs.replaceFirst("？", "");
}
if (rs.startsWith("nn")) {
rs = rs.replaceFirst("nn", "");
}
log.info("格式化后的rs:{}", rs);
return rs;
}
private String reply(FeishuResponse poll, String rs) {
JSONObject params = new JSONObject();
params.put("uuid", RandomUtil.randomNumbers(10));
params.put("msg_type", "text");
JSONObject content = new JSONObject();
content.put("text", rs);
params.put("content", content.toJSONString());
String url = String.format("https://open.feishu.cn/open-apis/im/v1/messages/%s/reply",
poll.getMessageId());
String tenantAccessToken = FeishuUtils.getTenantAccessToken();
String body = null;
try (HttpResponse authorization = HttpUtil.createPost(url)
.header("Authorization", "Bearer " + tenantAccessToken)
.body(params.toJSONString())
.execute()) {
body = authorization.body();
}
return body;
}
}The tool class for obtaining Feishu token is as follows:
@Slf4j
public class FeishuUtils {
public static final String tokenUrl
= "https://open.feishu.cn/open-apis/auth/v3/app_access_token/internal/";
//构建一个cache 缓存飞书的token
static Cache<String, String> tokenCache =
CacheBuilder.newBuilder().expireAfterWrite(Duration.ofSeconds(3500)).build();
//这个是飞书应用的appid和key，可以在创建的飞书应用中找到
public static final String appId = "";
public static final String appKey = "";
public static String getTenantAccessToken() {
String token = null;
try {
token = tokenCache.get("token", () -> {
JSONObject params = new JSONObject();
params.put("app_id", appId);
params.put("app_secret", appKey);
String body;
try (HttpResponse execute = HttpUtil.createPost(tokenUrl)
.body(params.toJSONString()).execute()) {
body = execute.body();
}
log.info("获取飞书token:{}", body);
if (StrUtil.isNotBlank(body)) {
String tenantAccessToken = JSON.parseObject(body).getString("tenant_access_token");
tokenCache.put("token", tenantAccessToken);
return tenantAccessToken;
}
return null;
});
} catch (ExecutionException e) {
throw new RuntimeException(e);
}
return token;
}
}4. Just start the thread class.
Finally, for the sake of privacy, the effect of chatgpt group conversation will not be displayed. Let’s show the effect of direct conversation with the robot.
Since we introduced chatgpt just with a trying attitude, the code is relatively rough. If there is anything wrong with the writing , I hope everyone will understand.
The code in this article also introduces additional jars: guava, hutool-all, and fastjson.
The above is the detailed content of Code attached, detailed steps for connecting ChatGPT to Feishu. For more information, please follow other related articles on the PHP Chinese website!
![[Web front-end] Node.js quick start](https://img.php.cn/upload/course/000/000/067/662b5d34ba7c0227.png)
