CAMEL Agents

Communicative Agents for "Mind" Exploration

中文文档 | English

Architecture • Installation • Quick Start • API Reference

A Python implementation of CAMEL, a framework for building communicative agents that collaborate through role-playing.

Architecture

CAMEL Architecture
│
├── CAMELSociety (Orchestrator)
│   ├── Manages multi-agent collaboration
│   ├── Handles turn-taking between agents
│   └── Tracks conversation progress
│
├── RolePlayingAgent (Agents)
│   ├── Role-based behavior
│   ├── Contextual responses
│   └── Goal-oriented communication
│
├── Conversation (Dialogue Manager)
│   ├── Message history
│   ├── Context formatting
│   └── OpenAI format export
│
└── Types (Data Structures)
    ├── Role - Agent role definition
    ├── Task - Collaborative task
    ├── Message - Single utterance
    └── TaskSpec - Task specification

Core Workflow

┌─────────────────────────────────────────────────────────────────┐
│                    CAMEL Collaboration Flow                      │
├─────────────────────────────────────────────────────────────────┤
│                                                                  │
│   Task ──▶ Agent 1 (Role A) ◀──▶ Conversation ◀──▶ Agent 2     │
│              │                    │                    │         │
│              │                    │                    │         │
│              ▼                    ▼                    ▼         │
│           Respond              Update              Respond       │
│              │                    │                    │         │
│              └────────────────────┴────────────────────┘         │
│                         (Iterate until done)                     │
│                                                                  │
└─────────────────────────────────────────────────────────────────┘

Installation

pip install camel-agents

Quick Start

from camel import CAMELSociety, Role, AgentType

# Create society
society = CAMELSociety(verbose=True, max_turns=10)

# Define roles
programmer = Role(
    name="Python Programmer",
    description="Expert Python developer who writes clean code",
    agent_type=AgentType.ASSISTANT
)

reviewer = Role(
    name="Code Reviewer",
    description="Meticulous reviewer ensuring quality",
    agent_type=AgentType.USER
)

# Run collaboration
result = society.collaborate(
    task="Write a function to sort a list",
    agents=[programmer, reviewer]
)

print(f"Success: {result.success}")
print(f"Output: {result.final_output}")
print(result.to_transcript())

Predefined Roles

CLI Usage

# Quick collaboration with predefined roles
camel run "Write a sorting algorithm" --role1 programmer --role2 reviewer

# Interactive mode
camel interactive --role1 teacher --role2 student

# List available roles
camel roles

# Export result
camel run "Task" --export result.json

API Reference

CAMELSociety

RolePlayingAgent

Academic Reference

Implementation of the CAMEL paper:

CAMEL: Communicative Agents for "Mind" Exploration of Large Language Models

Guohao Li, Hasan Hammoud, Hani Itani, Dmitrii Khizbullin, Bernard Ghanem

NeurIPS 2023

Paper: https://arxiv.org/abs/2303.17760

@inproceedings{li2023camel,
  title={CAMEL: Communicative Agents for "Mind" Exploration of Large Language Models},
  author={Li, Guohao and Hammoud, Hasan and Itani, Hani and Khizbullin, Dmitrii and Ghanem, Bernard},
  booktitle={Advances in Neural Information Processing Systems (NeurIPS)},
  year={2023}
}

License

MIT License

Made with ❤️ by AI Agent Research Team