Factory

Overview

The LLM Factory implements a flexible provider pattern for managing multiple language model integrations. It provides a unified interface for generating responses while supporting provider-specific features like function calling and streaming.

Architecture

Core Interfaces

interface LLMResponse {
  content: string;
  totalTokens?: number;
  toolCalls?: Array<{
    id: string;
    type: 'function';
    function: {
      name: string;
      arguments: string;
    }
  }>;
}

interface LLMProvider {
  generateResponse(
    messages: Array<{ role: string; content: string }>, 
    cleanContent: string
  ): Promise<LLMResponse>;
}

export type ModelType = 'OPENAI' | 'GEMINI' | 'GROK' | 'DEEPSEEK';

Factory Implementation

export class LLMFactory {
  private static instance: LLMFactory;
  private providers: Map<ModelType, LLMProvider>;
  private currentProvider: LLMProvider | null;

  private constructor() {
    this.providers = new Map();
    this.providers.set('OPENAI', new OpenAIProvider());
    this.providers.set('GEMINI', new GeminiProvider());
    this.providers.set('GROK', new GrokProvider());
    this.providers.set('DEEPSEEK', new DeepSeekProvider());
    this.currentProvider = this.providers.get('OPENAI') || null;
  }

  public static getInstance(): LLMFactory;
  public getAvailableModels(): ModelType[];
  public setModel(model: ModelType): boolean;
  public getCurrentModel(): ModelType | null;
  public getCurrentProvider(): LLMProvider | null;
  public generateResponse(
    messages: Array<{ role: string; content: string }>,
    cleanContent: string
  ): Promise<LLMResponse>;
}

Providers

OpenAI Provider

class OpenAIProvider implements LLMProvider {
  private client: OpenAI;

  constructor() {
    this.client = new OpenAI({
      apiKey: process.env.OPENAI_API_KEY,
    });
  }

  async generateResponse(
    messages: Array<{ role: string; content: string }>
  ): Promise<LLMResponse>;
}

• Model: gpt-4o

• Full function calling support

• Token usage tracking

• Error handling and retries

• Standard OpenAI SDK integration

Gemini Provider

class GeminiProvider implements LLMProvider {
  private model: GenerativeModel;

  constructor() {
    const genAI = new GoogleGenerativeAI(process.env.GEMINI_API_KEY!);
    this.model = genAI.getGenerativeModel({ model: 'gemini-pro' });
  }

  async generateResponse(
    messages: Array<{ role: string; content: string }>,
    cleanContent: string
  ): Promise<LLMResponse>;
}

• Model: gemini-pro

• Chat support without function calling

• Role-based message history (assistant → model)

• Custom error handling

• Google AI SDK integration

Grok Provider

class GrokProvider implements LLMProvider {
  private client: OpenAI;

  constructor() {
    this.client = new OpenAI({
      apiKey: process.env.XAI_API_KEY,
      baseURL: 'https://api.x.ai/v1',
    });
  }

  async generateResponse(
    messages: Array<{ role: string; content: string }>,
    cleanContent: string
  ): Promise<LLMResponse>;
}

• Model: grok-2-latest

• OpenAI-compatible API

• Function calling support

• Response normalization

• Tool call format standardization

DeepSeek Provider

class DeepSeekProvider implements LLMProvider {
  async generateResponse(
    messages: Array<{ role: string; content: string }>,
    cleanContent: string
  ): Promise<LLMResponse>;
}

• Model: deepseek-chat

• Direct API integration

• Function calling support

• Custom message transformation

• Direct HTTP request handling

Function Calling

Supported tools across compatible providers:

Core Tools

const tools = [{
  type: "function",
  function: {
    name: "generate_image",
    description: "Generate an image based on the user's request",
    parameters: {
      type: "object",
      properties: {
        prompt: {
          type: "string",
          description: "The description of the image to generate"
        }
      },
      required: ["prompt"]
    }
  }
},
{
  type: "function",
  function: {
    name: "get_weather",
    description: "Get current weather data for any major city",
    parameters: {
      type: "object",
      properties: {
        location: {
          type: "object",
          properties: {
            name: { type: "string" },
            country: { type: "string" },
            coordinates: {
              type: "object",
              properties: {
                lat: { type: "number" },
                lon: { type: "number" }
              },
              required: ["lat", "lon"]
            }
          },
          required: ["name", "country", "coordinates"]
        },
        type: {
          type: "string",
          enum: ["current", "forecast"]
        }
      },
      required: ["location"]
    }
  }
},
{
  type: "function",
  function: {
    name: "get_market_data",
    description: "Get cryptocurrency market data",
    parameters: {
      type: "object",
      properties: {
        type: {
          type: "string",
          enum: ["token", "trending", "top", "latest", "boosted"]
        },
        query: {
          type: "string",
          description: "Token symbol or address"
        }
      },
      required: ["type"]
    }
  }
},
{
  type: "function",
  function: {
    name: "get_time",
    description: "Gets the current time for a location",
    parameters: {
      type: "object",
      properties: {
        location: {
          type: "object",
          properties: {
            name: { type: "string" },
            country: { type: "string" },
            coordinates: {
              type: "object",
              properties: {
                lat: { type: "number" },
                lon: { type: "number" }
              }
            }
          }
        }
      },
      required: ["location"]
    }
  }
}]

Usage

Basic Usage

const factory = LLMFactory.getInstance();
const response = await factory.generateResponse(messages, cleanContent);

Model Selection

const factory = LLMFactory.getInstance();
const success = factory.setModel('GEMINI');
if (success) {
  const response = await factory.generateResponse(messages, cleanContent);
}

Error Handling

try {
  const response = await factory.generateResponse(messages, cleanContent);
} catch (error) {
  console.error('LLM Error:', error);
  // Handle provider-specific errors
}

Best Practices

Provider Selection

• Use OpenAI for complex tasks requiring function calling

• Use Gemini for basic chat interactions

• Use Grok for OpenAI-compatible function calling

• Use DeepSeek for specialized tasks

• Consider rate limits and costs

• Match provider capabilities to task requirements

Error Handling

• Implement provider-specific error handling

• Handle rate limits and quotas

• Validate responses

• Log errors appropriately

• Implement retries where appropriate

• Handle network failures

• Normalize error formats

Performance

• Reuse factory instance (singleton pattern)

• Monitor token usage

• Implement caching where appropriate

• Handle streaming responses efficiently

• Optimize message history

• Clean content processing

• Validate tool calls

Related Documentation

• Model Preferences

• Provider Configuration

• Function Calling