Universal Agent

The Universal Agent provides a flexible interface for interacting with various AI models across different providers. It supports multiple model providers, custom instructions, and tool integration while maintaining a consistent interface for all interactions.

Universal Agent workflow and architecture

Configuration Parameters

Model Provider Options

Amazon Bedrock
- Claude
- Titan
- Jurassic-2
Anthropic
- Claude 3 Opus
- Claude 3 Sonnet
- Claude 3 Haiku
Azure OpenAI
- GPT-4
- GPT-3.5-Turbo
Groq
- LLaMA2 70B
- Mixtral 8x7B
OpenAI
- GPT-4
- GPT-3.5-Turbo
Custom
- User-defined model endpoints
- Custom model implementations

Required Input Parameters

model_provider: Selected AI model provider
model_name: Specific model to use
input: User query or instruction

Optional Configuration

agent_instructions: Custom behavior instructions
tools: Array of available tools
- name: Tool identifier
- description: Tool functionality
- parameters: Required inputs
- handler: Function implementation
model_params: Model-specific parameters
- temperature: Response creativity
- max_tokens: Response length
- top_p: Nucleus sampling

Output Format

{
  "response": {
    "content": string,
    "role": string,
    "tool_calls": [
      {
        "tool": string,
        "input": object,
        "output": any
      }
    ]
  },
  "metadata": {
    "model_info": {
      "provider": string,
      "model": string,
      "version": string
    },
    "usage": {
      "prompt_tokens": number,
      "completion_tokens": number,
      "total_tokens": number
    },
    "latency": number
  }
}

Features

Multi-provider support
Custom tool integration
Flexible instruction setting
Error handling and retries
Usage tracking
Performance monitoring
Model fallback options

Note: Each model provider may have specific authentication requirements and rate limits. Ensure proper API keys and credentials are configured for your selected provider.

Tip: Use clear and specific agent instructions to optimize model behavior. Consider implementing model fallback strategies for improved reliability.

Example Usage

const agent = new Agent({
  model_provider: "anthropic",
  model_name: "claude-3-opus",
  agent_instructions: "You are a helpful assistant...",
  tools: [
    {
      name: "calculator",
      description: "Perform mathematical calculations",
      parameters: {
        expression: "string"
      }
    }
  ]
});

const response = await agent.run({
  input: "Calculate the compound interest on $1000 at 5% for 3 years"
});