Skip to main content

Overview

When you need to implement private Agent streaming API protocol integration, you can create a custom Agent service class by inheriting from BaseAgentService.

What is AgentService​

AgentService is the core component for handling communication with AI conversation services, responsible for:

  • Sending user messages to AI service endpoints
  • Receiving and processing streaming response data
  • Managing conversation session states
  • Handling connection exceptions and retry logic

Core Features​

1. Streaming Conversation Processing​

Implement real-time conversation responses through SSE (Server-Sent Events) or WebSocket.

2. Session Management​

Automatically maintain conversation session IDs to ensure continuity across multiple conversation rounds.

3. Error Handling​

Provides complete error handling mechanisms, including network exceptions, server errors, etc.

4. Extensibility​

Supports custom request parameters, response processing logic, and business-specific features.

Use Cases​

  • Integrate with privately deployed AI conversation services
  • Scenarios requiring custom request/response formats
  • Implement complex business logic processing
  • Add custom monitoring and tracking

Configuration Notes​

⚠️ Important: When using custom agentService, standard agentConfig configuration will not be available. All related configurations need to be implemented in the custom service class.

import {createAaaSPilotKit, BaseAgentService} from '@bdky/aaas-pilot-kit';

// Create custom AgentService
class CustomAgentService extends BaseAgentService {
    // Implement specific logic
}

// Bind in Kit initialization, <CustomAgentService> TS generic parameter recommended for type constraint
const kit = createAaaSPilotKit<CustomAgentService>({
    // Other configurations...
    agentService: CustomAgentService
});

Next Steps​

Choose the HTTP client implementation suitable for your project: