Component Details

Platform Adapter

• Handles incoming messages from Telegram (Phase 1) or X (Phase 2)

• Manages platform-specific formatting

• Implements rate limiting and queueing

• Handles media processing (images, files, etc.)

Agent Runtime

• Core processing engine that manages conversation flow

• Implements context window management

• Handles message formatting and tokenization

• Manages agent state and session tracking

LLM Provider Integration

• Abstracts away provider-specific API differences

• Implements fallback mechanisms for API failures

• Optimizes prompt construction for different models

• Manages token usage and quota monitoring

State Storage

• Maintains conversation history and agent state

• Implements configurable retention policies

• Provides persistence across container restarts

• Options for local, IPFS, or Arweave storage

// Agent Runtime Interface
interface AgentRuntime {
  initialize(): Promise<void>;
  
  handleMessage(message: IncomingMessage): Promise<OutgoingMessage[]>;
  
  getState(): Promise<AgentState>;
  
  setState(partial: Partial<AgentState>): Promise<void>;
  
  shutdown(): Promise<void>;
}

// LLM Provider Interface
interface LLMProvider {
  generateResponse(
    prompt: string,
    parameters: {
      temperature: number;
      maxTokens: number;
      topP?: number;
      frequencyPenalty?: number;
      presencePenalty?: number;
    }
  ): Promise<{
    text: string;
    usage: {
      promptTokens: number;
      completionTokens: number;
      totalTokens: number;
    }
  }>;
  
  countTokens(text: string): number;
}