架构重构说明¶

public class GPUStackAdapter {
    public Mono<Response> chat(Request request) { ... }
    public Mono<Response> embedding(EmbeddingRequest request) { ... }
}

public interface AiServiceAdapter {
    String getServiceType();
    Mono<Response> chat(Request request);
    Mono<Response> embedding(EmbeddingRequest request);
}

public abstract class BaseAiServiceAdapter implements AiServiceAdapter {
    // 公共逻辑：重试、超时、错误处理
}

public class GPUStackAdapter extends BaseAiServiceAdapter {
    // 只关注 GPUStack 特定逻辑
}

public interface LoadBalancerStrategy {
    String select(List<String> instances, String serviceKey);
}

public class RoundRobinStrategy implements LoadBalancerStrategy { ... }
public class WeightedStrategy implements LoadBalancerStrategy { ... }
public class LeastConnectionsStrategy implements LoadBalancerStrategy { ... }

public interface RateLimiter {
    Mono<Boolean> tryAcquire(String key);
}

public class TokenBucketRateLimiter implements RateLimiter { ... }
public class SlidingWindowRateLimiter implements RateLimiter { ... }
public class LeakyBucketRateLimiter implements RateLimiter { ... }

public class RateLimiterFactory {
    public RateLimiter create(RateLimiterType type, Config config) { ... }
}