List of Agents

Agent Base Class

textarena.core.Agent

The Agent is a generic base class that defines the basic structure of an agent. It is designed to process observations and generate actions, providing a foundation for building custom agents.

Available Agent Classes

textarena.agents.HumanAgent

The HumanAgent allows manual user input for actions. It is primarily useful for debugging or exploratory interactions.

Example:

import textarena as ta
agent = ta.agents.HumanAgent()

textarena.agents.AnthropicAgent

The AnthropicAgent integrates with Anthropic's Claude models to generate responses. It supports customizable system prompts and model selection.

Example:

import textarena as ta
agent = ta.agents.AnthropicAgent(model_name="claude-3-5-sonnet-20241022")

textarena.agents.AWSBedrockAgent

The AWSBedrockAgent interacts with a range of model families via AWS Bedrock Runtime API, providing scalable cloud-based inference.

Example:

import textarena as ta
agent = ta.agents.AWSBedrockAgent(model_id="us.meta.llama3-2-11b-instruct-v1:0")

textarena.agents.CerebrasAgent

The CerebrasAgent uses the Cerebras cloud SDK to interact with large-scale models. It is ideal for tasks requiring high efficiency and scalability.

Example:

import textarena as ta
agent = ta.agents.CerebrasAgent(model_name="cerebras-llm-13b")

textarena.agents.GeminiAgent

The GeminiAgent uses the Google Gemini API to generate responses. It allows model selection and system prompt customization.

Example:

import textarena as ta
agent = ta.agents.GeminiAgent(model_name="gemini-pro")

textarena.agents.HFLocalAgent

The HFLocalAgent integrates with Hugging Face Transformers for local inference. It supports optional 8-bit quantization for efficient model execution.

Example:

import textarena as ta
agent = ta.agents.HFLocalAgent(
    model_name="gpt2",
    quantize=False  # default
)

textarena.agents.OpenAIAgent

The OpenAIAgent uses OpenAI's API to generate responses. It supports configurable system prompts and model customization.

Example:

import textarena as ta
agent = ta.agents.OpenAIAgent(model_name="gpt-4")

textarena.agents.OpenRouterAgent

The OpenRouterAgent uses OpenRouter's API to communicate with language models. It supports customizable system prompts and model selection.

Example:

import textarena as ta

# with the default system prompt
agent = ta.agents.OpenRouterAgent(model_name="GPT-4o-mini")

# with a custom system prompt
agent = ta.agents.OpenRouterAgent(
    model_name="GPT-4o-mini",
    system_prompt="Your life depends on whether you win [...]"
)