AI Providers

AI Providers implement the LLMProvider interface from @stratix/core, enabling your AI agents to communicate with different Large Language Model services.

Available Providers

Anthropic (Claude)

Package: @stratix/ai-anthropic

Anthropic's Claude models, known for their strong reasoning and long context windows.

Installation

npm install @stratix/ai-anthropic

Or using the CLI:

stratix add ai-anthropic

Supported Models

• claude-3-opus-20240229 - Most capable, best for complex tasks
• claude-3-sonnet-20240229 - Balanced performance and cost
• claude-3-haiku-20240307 - Fastest, most cost-effective
• claude-3-5-sonnet-20241022 - Latest Sonnet with improved capabilities

Usage

import { AnthropicProvider } from '@stratix/ai-anthropic';
import { AIAgent } from '@stratix/core';

// Create provider
const provider = new AnthropicProvider({
  apiKey: process.env.ANTHROPIC_API_KEY!
});

// Use with an agent
const agent = new AIAgent({
  id: 'assistant',
  name: 'Assistant',
  description: 'Helpful assistant',
  llmProvider: provider,
  model: 'claude-3-5-sonnet-20241022',
  systemPrompt: 'You are a helpful assistant.',
});

// Execute agent
const result = await agent.execute({
  messages: [
    { role: 'user', content: 'Hello!', timestamp: new Date() }
  ]
});

Features

• ✅ Chat completion
• ✅ Streaming responses
• ✅ Tool/function calling
• ✅ Cost tracking
• ❌ Embeddings (not supported by Anthropic)

---

OpenAI

Package: @stratix/ai-openai

OpenAI's GPT models, including GPT-4 and GPT-3.5.

Installation

npm install @stratix/ai-openai

Or using the CLI:

stratix add ai-openai

Supported Models

• gpt-4-turbo - Latest GPT-4 with vision
• gpt-4 - Original GPT-4
• gpt-3.5-turbo - Fast and cost-effective
• text-embedding-3-small - Embeddings model
• text-embedding-3-large - Higher quality embeddings

Usage

import { OpenAIProvider } from '@stratix/ai-openai';
import { AIAgent } from '@stratix/core';

// Create provider
const provider = new OpenAIProvider({
  apiKey: process.env.OPENAI_API_KEY!
});

// Use with an agent
const agent = new AIAgent({
  id: 'assistant',
  name: 'Assistant',
  description: 'Helpful assistant',
  llmProvider: provider,
  model: 'gpt-4-turbo',
  systemPrompt: 'You are a helpful assistant.',
});

Features

• ✅ Chat completion
• ✅ Streaming responses
• ✅ Tool/function calling
• ✅ Embeddings
• ✅ Cost tracking
• ✅ Vision (GPT-4 Turbo)

---

Provider Interface

All AI providers implement the LLMProvider interface:

interface LLMProvider {
  readonly name: string;
  readonly models: string[];

  chat(params: ChatParams): Promise<ChatResponse>;
  streamChat(params: ChatParams): AsyncIterable<ChatChunk>;
  embeddings(params: EmbeddingParams): Promise<EmbeddingResponse>;
}

Choosing a Provider

Use Case	Recommended Provider	Model
Complex reasoning	Anthropic	claude-3-opus-20240229
Balanced performance	Anthropic	claude-3-5-sonnet-20241022
Fast responses	Anthropic	claude-3-haiku-20240307
Vision tasks	OpenAI	gpt-4-turbo
Embeddings	OpenAI	text-embedding-3-large
Cost-effective	OpenAI	gpt-3.5-turbo

Creating a Custom Provider

To create your own LLM provider:

import type {
  LLMProvider,
  ChatParams,
  ChatResponse,
  ChatChunk,
  EmbeddingParams,
  EmbeddingResponse
} from '@stratix/core';

export class CustomProvider implements LLMProvider {
  readonly name = 'custom';
  readonly models = ['model-1', 'model-2'];

  constructor(private config: { apiKey: string }) {}

  async chat(params: ChatParams): Promise<ChatResponse> {
    // Call your LLM API
    const response = await fetch('https://api.example.com/chat', {
      method: 'POST',
      headers: {
        'Authorization': `Bearer ${this.config.apiKey}`,
        'Content-Type': 'application/json'
      },
      body: JSON.stringify({
        model: params.model,
        messages: params.messages,
        temperature: params.temperature
      })
    });

    const data = await response.json();

    return {
      content: data.content,
      usage: {
        promptTokens: data.usage.prompt_tokens,
        completionTokens: data.usage.completion_tokens,
        totalTokens: data.usage.total_tokens
      },
      finishReason: 'stop'
    };
  }

  async *streamChat(params: ChatParams): AsyncIterable<ChatChunk> {
    // Implement streaming
  }

  async embeddings(params: EmbeddingParams): Promise<EmbeddingResponse> {
    // Implement embeddings
  }
}

Next Steps

• Creating AI Agents - Build agents using these providers
• Agent Tools - Add tool calling to your agents
• Agent Testing - Test agents with MockLLMProvider