examples/run_groq.py

# ========= Copyright 2023-2024 @ CAMEL-AI.org. All Rights Reserved. =========
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ========= Copyright 2023-2024 @ CAMEL-AI.org. All Rights Reserved. =========

"""
This module provides integration with the Groq API platform for the OWL system.

It configures different agent roles with appropriate Groq models based on their requirements:
- Tool-intensive roles (assistant, web, planning, video, image) use GROQ_LLAMA_3_3_70B
- Document processing uses GROQ_MIXTRAL_8_7B
- Simple roles (user) use GROQ_LLAMA_3_1_8B

To use this module:
1. Set GROQ_API_KEY in your .env file
2. Set OPENAI_API_BASE_URL to "https://api.groq.com/openai/v1"
3. Run with: python -m examples.run_groq
"""

import sys
from dotenv import load_dotenv
from camel.models import ModelFactory
from camel.toolkits import (
    AudioAnalysisToolkit,
    CodeExecutionToolkit,
    ExcelToolkit,
    ImageAnalysisToolkit,
    SearchToolkit,
    VideoAnalysisToolkit,
    BrowserToolkit,
    FileWriteToolkit,
)
from camel.types import ModelPlatformType, ModelType
from camel.logger import set_log_level

from owl.utils import OwlRolePlaying, run_society, DocumentProcessingToolkit

load_dotenv()

set_log_level(level="DEBUG")


def construct_society(question: str) -> OwlRolePlaying:
    r"""Construct a society of agents based on the given question.

    Args:
        question (str): The task or question to be addressed by the society.

    Returns:
        OwlRolePlaying: A configured society of agents ready to address the question.
    """

    # Create models for different components
    models = {
        "user": ModelFactory.create(
            model_platform=ModelPlatformType.GROQ,
            model_type=ModelType.GROQ_LLAMA_3_1_8B,  # Simple role, can use 8B model
            model_config_dict={"temperature": 0},
        ),
        "assistant": ModelFactory.create(
            model_platform=ModelPlatformType.GROQ,
            model_type=ModelType.GROQ_LLAMA_3_3_70B,  # Main assistant needs tool capability
            model_config_dict={"temperature": 0},
        ),
        "browsing": ModelFactory.create(
            model_platform=ModelPlatformType.GROQ,
            model_type=ModelType.GROQ_LLAMA_3_3_70B,  # Web browsing requires tool usage
            model_config_dict={"temperature": 0},
        ),
        "planning": ModelFactory.create(
            model_platform=ModelPlatformType.GROQ,
            model_type=ModelType.GROQ_LLAMA_3_3_70B,  # Planning requires complex reasoning
            model_config_dict={"temperature": 0},
        ),
        "video": ModelFactory.create(
            model_platform=ModelPlatformType.GROQ,
            model_type=ModelType.GROQ_LLAMA_3_3_70B,  # Video analysis is multimodal
            model_config_dict={"temperature": 0},
        ),
        "image": ModelFactory.create(
            model_platform=ModelPlatformType.GROQ,
            model_type=ModelType.GROQ_LLAMA_3_3_70B,  # Image analysis is multimodal
            model_config_dict={"temperature": 0},
        ),
        "document": ModelFactory.create(
            model_platform=ModelPlatformType.GROQ,
            model_type=ModelType.GROQ_MIXTRAL_8_7B,  # Document processing can use Mixtral
            model_config_dict={"temperature": 0},
        ),
    }

    # Configure toolkits
    tools = [
        *BrowserToolkit(
            headless=False,  # Set to True for headless mode (e.g., on remote servers)
            web_agent_model=models["browsing"],
            planning_agent_model=models["planning"],
        ).get_tools(),
        *VideoAnalysisToolkit(model=models["video"]).get_tools(),
        *AudioAnalysisToolkit().get_tools(),  # This requires OpenAI Key
        *CodeExecutionToolkit(sandbox="subprocess", verbose=True).get_tools(),
        *ImageAnalysisToolkit(model=models["image"]).get_tools(),
        SearchToolkit().search_duckduckgo,
        SearchToolkit().search_google,  # Comment this out if you don't have google search
        SearchToolkit().search_wiki,
        *ExcelToolkit().get_tools(),
        *DocumentProcessingToolkit(model=models["document"]).get_tools(),
        *FileWriteToolkit(output_dir="./").get_tools(),
    ]

    # Configure agent roles and parameters
    user_agent_kwargs = {"model": models["user"]}
    assistant_agent_kwargs = {"model": models["assistant"], "tools": tools}

    # Configure task parameters
    task_kwargs = {
        "task_prompt": question,
        "with_task_specify": False,
    }

    # Create and return the society
    society = OwlRolePlaying(
        **task_kwargs,
        user_role_name="user",
        user_agent_kwargs=user_agent_kwargs,
        assistant_role_name="assistant",
        assistant_agent_kwargs=assistant_agent_kwargs,
    )

    return society


def main():
    r"""Main function to run the OWL system with an example question."""
    # Example research question
    default_task = "Navigate to Amazon.com and identify one product that is attractive to coders. Please provide me with the product name and price. No need to verify your answer."

    # Construct and run the society
    # Note: This configuration uses GROQ_LLAMA_3_3_70B for tool-intensive roles (assistant, web, planning, video, image)
    # and GROQ_MIXTRAL_8_7B for document processing. GROQ_LLAMA_3_1_8B is used only for the user role
    # which doesn't require tool usage capabilities.

    # Override default task if command line argument is provided
    task = sys.argv[1] if len(sys.argv) > 1 else default_task

    # Construct and run the society
    society = construct_society(task)
    answer, chat_history, token_count = run_society(society)

    # Output the result
    print(f"\033[94mAnswer: {answer}\033[0m")


if __name__ == "__main__":
    main()