Agents

Configuring and managing AI agents in Playgent

Agents

An Agent is your AI system configuration in Playgent. It includes your LLM provider, credentials, system prompt, and default evaluation metrics.

Quick Start

Python

from playgent import Playgent

client = Playgent(api_key="your-api-key")

agent = client.agents.create(
    name="Customer Support Agent",
    provider="openai",
    credentials={"OPENAI_API_KEY": "sk-xxx..."},
    system_prompt="You are a helpful customer support agent for Acme Corp...",
    default_scorers=["answer_relevancy", "faithfulness", "bias"]
)

Configuration Options

Supported Providers

Provider	Value
OpenAI	openai
Anthropic	anthropic
AWS Bedrock	bedrock
LangChain	langchain
Custom	custom

System Prompt

Base instructions for your agent. Used during test execution and prompt optimization.

Default Scorers

Automatic evaluation metrics for all test runs:

agent = client.agents.create(
    name="Support Agent",
    provider="openai",
    default_scorers=["answer_relevancy", "faithfulness", "bias"]
)

Custom Endpoints

For custom agent invocation:

agent = client.agents.create(
    name="Custom Agent",
    provider="custom",
    invocation={
        "url": "https://your-api.com/chat",
        "headers": {"X-API-Key": "your-key"}
    }
)

Next Steps

Test Cases

Create test scenarios for your agent

API Reference

Full API documentation