Langchain4J调用自定义Base URL的Anthropic模型时授权头缺失报错

环境信息 #

• Langchain4J版本：0.36.2
• Java版本：21
• 部署场景：企业自定义Base URL的OpenAI与Anthropic模型，其中OpenAI模型可正常调用，Anthropic模型抛出授权头缺失错误

正常运行的OpenAI模型代码 #

// 此代码可正常运行无报错
import dev.langchain4j.model.chat.ChatLanguageModel;
import dev.langchain4j.model.openai.OpenAiChatModel;
import dev.langchain4j.model.output.Response;
import dev.langchain4j.model.chat.ChatMessage;
import dev.langchain4j.model.input.Prompt;

public class OpenAIChatExample {

    public static void main(String[] args) {
        
        String clientId = "myclientid";
        String clientSecret = "myclientsecret";
        
        TokenAuthenticator authenticator = new TokenAuthenticator(clientId, clientSecret);

        
        ChatLanguageModel model = OpenAiChatModel.builder()
                .baseUrl("https://mycustombaseurl")
                .apiKey(authenticator.getValidToken())
                .modelName("gpt-4o") 
                .build();

        try {
            
            String response = model.generate("What is LLM?");
            System.out.println("Simple Response: " + response);


        } catch (Exception e) {
            System.err.println("Error occurred: " + e.getMessage());
            e.printStackTrace();
        }
    }    
}

运行失败的Anthropic模型代码 #

// 此代码抛出错误："Required Authorization header is missing"
import dev.langchain4j.model.chat.ChatLanguageModel;
import dev.langchain4j.model.anthropic.AnthropicChatModel;
import dev.langchain4j.model.output.Response;
import dev.langchain4j.model.chat.ChatMessage;
import dev.langchain4j.model.input.Prompt;

public class AnthropicChatExample {

    public static void main(String[] args) {
        
        String clientId = "myclientid";
        String clientSecret = "myclientsecret";
        
        TokenAuthenticator authenticator = new TokenAuthenticator(clientId, clientSecret);

        
        AnthropicChatModel model = AnthropicChatModel.builder()
                .baseUrl("https://mycustombaseurl")
                .apiKey(authenticator.getValidToken())
                .modelName("claude-3-5-sonnet") 
                .build();

        try {
            
            String response = model.generate("What is LLM?");
            System.out.println("Simple Response: " + response);


        } catch (Exception e) {
            System.err.println("Error occurred: " + e.getMessage());
            e.printStackTrace();
        }
    }    
}

报错信息 #

Caused by: dev.langchain4j.model.anthropic.internal.client.AnthropicHttpException: {"Error": "Required Authorization header is missing"}

build.gradle配置 #

plugins {
    id 'java'
}

group = 'org.example'
version = '1.0-SNAPSHOT'

repositories {
    mavenCentral()
}

dependencies {
    testImplementation platform('org.junit:junit-bom:5.10.0')
    testImplementation 'org.junit.jupiter:junit-jupiter'
    implementation 'dev.langchain4j:langchain4j:0.36.2'
    implementation 'dev.langchain4j:langchain4j-core:0.36.2'
    implementation 'dev.langchain4j:langchain4j-open-ai:0.36.2'
    implementation group: 'dev.langchain4j', name: 'langchain4j-anthropic', version: '0.36.2'
    implementation 'dev.langchain4j:langchain4j-embeddings-all-minilm-l6-v2:0.36.2'
    implementation 'org.apache.httpcomponents.client5:httpclient5:5.2.1'
    implementation 'com.fasterxml.jackson.core:jackson-databind:2.15.2'
}

test {
    useJUnitPlatform()
}

排查分析与解决方案 #

可能原因 #

1. Langchain4J Anthropic客户端版本bug：0.36.2版本的langchain4j-anthropic模块在自定义Base URL场景下，未正确将配置的apiKey注入到请求的Authorization头中，与OpenAI客户端的请求头构造逻辑存在差异。
2. 企业网关授权格式要求差异：部分企业网关对Anthropic接口的授权头格式要求与原生Anthropic不同（原生为Bearer <token>），但OpenAI网关兼容当前token格式，导致Anthropic请求被拦截。

解决方案 #

方案1：手动配置自定义HttpClient，强制添加授权头

通过自定义Apache HttpClient拦截器，手动注入Authorization头，绕过Langchain4J Anthropic客户端的默认头构造逻辑：

import org.apache.hc.client5.http.classic.HttpClient;
import org.apache.hc.client5.http.impl.classic.HttpClients;
import org.apache.hc.core5.http.message.BasicHeader;

public class AnthropicChatExample {

    public static void main(String[] args) {
        
        String clientId = "myclientid";
        String clientSecret = "myclientsecret";
        TokenAuthenticator authenticator = new TokenAuthenticator(clientId, clientSecret);
        String validToken = authenticator.getValidToken();

        // 构建自定义HttpClient，添加Authorization头拦截器
        HttpClient httpClient = HttpClients.custom()
                .addRequestInterceptorFirst((request, context) -> {
                    request.addHeader(new BasicHeader("Authorization", "Bearer " + validToken));
                    return null;
                })
                .build();

        AnthropicChatModel model = AnthropicChatModel.builder()
                .baseUrl("https://mycustombaseurl")
                .httpClient(httpClient)
                .modelName("claude-3-5-sonnet")
                .build();

        try {
            String response = model.generate("What is LLM?");
            System.out.println("Simple Response: " + response);
        } catch (Exception e) {
            System.err.println("Error occurred: " + e.getMessage());
            e.printStackTrace();
        }
    }    
}

方案2：升级Langchain4J版本

检查Langchain4J更新日志，确认后续版本是否修复了自定义Base URL下Anthropic授权头的问题，尝试升级到最新稳定版本（如0.37+）后重新测试。

方案3：验证企业网关授权要求

联系企业网关管理员，确认Anthropic接口所需的授权头格式：

• 是否需要使用X-API-Key而非Authorization？
• token是否需要添加特定前缀（如ApiKey 而非Bearer ）？
  若格式不同，可在自定义HttpClient中调整头的名称和值。

内容的提问来源于stack exchange，提问作者Sabari