update camel version to 0.2.84

2026-03-22 05:57:17 +08:00 · 2026-01-24 16:58:28 +08:00
parent 7006768991
commit 5e80b97fb7
25 changed files with 192 additions and 107 deletions
--- a/README.md
+++ b/README.md
@@ -1,14 +1,3 @@
-<div align="center" style="background-color: #fff3cd; padding: 15px; border-radius: 10px; border: 3px solid #ff6b6b; margin: 20px 0;">
-  <h3 style="color: #d63031; margin: 0;">
-    ‼️ <b>IMPORTANT NOTICE</b> ‼️
-  </h3>
-  <p style="font-size: 1.1em; margin: 10px 0; color: #2d3436;">
-    <b>The current version of OWL does not use the up-to-date version of <a href="https://github.com/camel-ai/camel">CAMEL</a>.</b><br>
-    If you want to build the best performing agents powered by workforce, please check out <a href="https://github.com/camel-ai/camel/blob/master/examples/workforce/eigent.py"><b>eigent.py</b></a>.<br>
-    We are also working on updating CAMEL to the latest version.
-  </p>
-</div>
-
 <div align="center">

 </div>
--- a/README_ja.md
+++ b/README_ja.md
@@ -1,14 +1,3 @@
-<div align="center" style="background-color: #fff3cd; padding: 15px; border-radius: 10px; border: 3px solid #ff6b6b; margin: 20px 0;">
-  <h3 style="color: #d63031; margin: 0;">
-    ‼️ <b>重要なお知らせ</b> ‼️
-  </h3>
-  <p style="font-size: 1.1em; margin: 10px 0; color: #2d3436;">
-    <b>現在のOWLバージョンは、最新版の<a href="https://github.com/camel-ai/camel">CAMEL</a>を使用していません。</b><br>
-    workforceを活用した最高性能のエージェントを構築したい場合は、<a href="https://github.com/camel-ai/camel/blob/master/examples/workforce/eigent.py"><b>eigent.py</b></a>をご確認ください。<br>
-    CAMELを最新版に更新する作業も進行中です。
-  </p>
-</div>
-
 <h1 align="center">
 	🦉 OWL: 労働力学習の最適化による、現実世界のタスク自動化における一般的なマルチエージェント支援
 </h1>
@@ -75,12 +64,42 @@
 - [📋 目次](#-目次)
 - [🔥 ニュース](#-ニュース)
 - [🎬 デモビデオ](#-デモビデオ)
- [✨️ コア機能](#-コア機能)
+- [✨️ コア機能](#️-コア機能)
 - [🛠️ インストール](#️-インストール)
+  - [オプション1：uvを使用する（推奨）](#オプション1uvを使用する推奨)
+  - [オプション2：venvとpipを使用する](#オプション2venvとpipを使用する)
+  - [オプション3：condaを使用する](#オプション3condaを使用する)
+  - [**環境変数の設定**](#環境変数の設定)
+    - [オプション1：`.env`ファイルを使用する（推奨）](#オプション1envファイルを使用する推奨)
+    - [オプション2：環境変数を直接設定](#オプション2環境変数を直接設定)
+  - [**Dockerでの実行**](#dockerでの実行)
+    - [**セットアップ手順**](#セットアップ手順)
+    - [**デプロイメントオプション**](#デプロイメントオプション)
+      - [**オプション1：事前構築されたイメージを使用する（推奨）**](#オプション1事前構築されたイメージを使用する推奨)
+      - [**オプション2：ローカルでイメージを構築する**](#オプション2ローカルでイメージを構築する)
+      - [**オプション3：便利なスクリプトを使用する**](#オプション3便利なスクリプトを使用する)
+    - [**MCPデスクトップコマンダーのセットアップ**](#mcpデスクトップコマンダーのセットアップ)
 - [🚀 クイックスタート](#-クイックスタート)
+  - [基本的な使用法](#基本的な使用法)
+  - [異なるモデルでの実行](#異なるモデルでの実行)
+    - [モデルの要件](#モデルの要件)
+      - [サポートされているモデル](#サポートされているモデル)
+    - [例のタスク](#例のタスク)
 - [🧰 ツールキットと機能](#-ツールキットと機能)
-  - [モデルコンテキストプロトコル (MCP)](#モデルコンテキストプロトコル-mcp)
+  - [モデルコンテキストプロトコル（MCP）](#モデルコンテキストプロトコルmcp)
+    - [**Node.jsのインストール**](#nodejsのインストール)
+    - [Windows](#windows)
+    - [Linux](#linux)
+    - [Mac](#mac)
+    - [**Playwright MCPサービスのインストール**](#playwright-mcpサービスのインストール)
+  - [利用可能なツールキット](#利用可能なツールキット)
+  - [利用可能なツールキット](#利用可能なツールキット-1)
+    - [マルチモーダルツールキット（マルチモーダルモデル機能が必要）](#マルチモーダルツールキットマルチモーダルモデル機能が必要)
+    - [テキストベースのツールキット](#テキストベースのツールキット)
+  - [カスタマイズ設定](#カスタマイズ設定)
 - [🌐 ウェブインターフェース](#-ウェブインターフェース)
+  - [ウェブUIの起動](#ウェブuiの起動)
+  - [機能](#機能)
 - [🧪 実験](#-実験)
 - [⏱️ 将来の計画](#️-将来の計画)
 - [📄 ライセンス](#-ライセンス)
@@ -89,6 +108,7 @@
 - [🔥 コミュニティ](#-コミュニティ)
 - [❓ FAQ](#-faq)
 - [📚 CAMEL依存関係の探索](#-camel依存関係の探索)
+  - [CAMELソースコードへのアクセス](#camelソースコードへのアクセス)
 - [⭐ Star History](#-star-history)

 # 🔥 ニュース
--- a/README_zh.md
+++ b/README_zh.md
@@ -1,14 +1,3 @@
-<div align="center" style="background-color: #fff3cd; padding: 15px; border-radius: 10px; border: 3px solid #ff6b6b; margin: 20px 0;">
-  <h3 style="color: #d63031; margin: 0;">
-    ‼️ <b>重要提示</b> ‼️
-  </h3>
-  <p style="font-size: 1.1em; margin: 10px 0; color: #2d3436;">
-    <b>当前版本的 OWL 未使用最新版本的 <a href="https://github.com/camel-ai/camel">CAMEL</a>。</b><br>
-    如果您想构建基于 workforce 的最佳性能智能体，请查看 <a href="https://github.com/camel-ai/camel/blob/master/examples/workforce/eigent.py"><b>eigent.py</b></a>。<br>
-    我们也正在努力将 CAMEL 更新到最新版本。
-  </p>
-</div>
-
 <h1 align="center">
 	🦉 OWL: Optimized Workforce Learning for General Multi-Agent Assistance in Real-World Task Automation
 	🦉 OWL: 优化劳动力学习的通用智能体，用于处理现实世界的自动化任务
@@ -73,11 +62,38 @@
 - [📋 目录](#-目录)
 - [🔥 新闻](#-新闻)
 - [🎬 演示视频](#-演示视频)
- [✨️ 核心功能](#-核心功能)
+- [✨️ 核心功能](#️-核心功能)
 - [🛠️ 安装](#️-安装)
+  - [选项1：使用 uv（推荐）](#选项1使用-uv推荐)
+  - [选项2：使用 venv 和 pip](#选项2使用-venv-和-pip)
+  - [选项3：使用 conda](#选项3使用-conda)
+  - [**设置环境变量**](#设置环境变量)
+    - [选项 1：使用 `.env` 文件（推荐）](#选项-1使用-env-文件推荐)
+    - [选项 2：直接设置环境变量](#选项-2直接设置环境变量)
+  - [**使用Docker运行**](#使用docker运行)
+    - [**设置说明**](#设置说明)
+    - [**部署选项**](#部署选项)
+      - [**选项1：使用预构建镜像（推荐）**](#选项1使用预构建镜像推荐)
+      - [**选项2：本地构建镜像**](#选项2本地构建镜像)
+      - [**选项3：使用便捷脚本**](#选项3使用便捷脚本)
+    - [**MCP Desktop Commander设置**](#mcp-desktop-commander设置)
 - [🚀 快速开始](#-快速开始)
+  - [基本用法](#基本用法)
+  - [使用不同的模型](#使用不同的模型)
+    - [模型要求](#模型要求)
+      - [支持的模型](#支持的模型)
 - [🧰 工具包与功能](#-工具包与功能)
-  - [模型上下文协议 (MCP)](#模型上下文协议-mcp)
+  - [模型上下文协议（MCP）](#模型上下文协议mcp)
+    - [**安装 Node.js**](#安装-nodejs)
+    - [Windows](#windows)
+    - [Linux](#linux)
+    - [Mac](#mac)
+    - [**安装 MCP 服务**](#安装-mcp-服务)
+  - [可用工具包](#可用工具包)
+  - [主要工具包](#主要工具包)
+    - [多模态工具包（需要模型具备多模态能力）](#多模态工具包需要模型具备多模态能力)
+    - [基于文本的工具包](#基于文本的工具包)
+  - [自定义配置](#自定义配置)
 - [🌐 网页界面](#-网页界面)
 - [🧪 实验](#-实验)
 - [⏱️ 未来计划](#️-未来计划)
@@ -87,6 +103,7 @@
 - [🔥 社区](#-社区)
 - [❓ 常见问题](#-常见问题)
 - [📚 探索 CAMEL 依赖](#-探索-camel-依赖)
+  - [访问 CAMEL 源代码](#访问-camel-源代码)
 - [⭐ Star History](#-star-history)

 </div>
--- a/examples/run.py
+++ b/examples/run.py
@@ -23,7 +23,7 @@ from camel.toolkits import (
    SearchToolkit,
    VideoAnalysisToolkit,
    BrowserToolkit,
-    FileWriteToolkit,
+    FileToolkit,
 )
 from camel.types import ModelPlatformType, ModelType
 from camel.logger import set_log_level
@@ -52,37 +52,37 @@ def construct_society(question: str) -> RolePlaying:
    models = {
        "user": ModelFactory.create(
            model_platform=ModelPlatformType.OPENAI,
-            model_type=ModelType.GPT_4O,
+            model_type=ModelType.GPT_5_1,
            model_config_dict={"temperature": 0},
        ),
        "assistant": ModelFactory.create(
            model_platform=ModelPlatformType.OPENAI,
-            model_type=ModelType.GPT_4O,
+            model_type=ModelType.GPT_5_1,
            model_config_dict={"temperature": 0},
        ),
        "browsing": ModelFactory.create(
            model_platform=ModelPlatformType.OPENAI,
-            model_type=ModelType.GPT_4O,
+            model_type=ModelType.GPT_5_1,
            model_config_dict={"temperature": 0},
        ),
        "planning": ModelFactory.create(
            model_platform=ModelPlatformType.OPENAI,
-            model_type=ModelType.GPT_4O,
+            model_type=ModelType.GPT_5_1,
            model_config_dict={"temperature": 0},
        ),
        "video": ModelFactory.create(
            model_platform=ModelPlatformType.OPENAI,
-            model_type=ModelType.GPT_4O,
+            model_type=ModelType.GPT_5_1,
            model_config_dict={"temperature": 0},
        ),
        "image": ModelFactory.create(
            model_platform=ModelPlatformType.OPENAI,
-            model_type=ModelType.GPT_4O,
+            model_type=ModelType.GPT_5_1,
            model_config_dict={"temperature": 0},
        ),
        "document": ModelFactory.create(
            model_platform=ModelPlatformType.OPENAI,
-            model_type=ModelType.GPT_4O,
+            model_type=ModelType.GPT_5_1,
            model_config_dict={"temperature": 0},
        ),
    }
@@ -103,7 +103,7 @@ def construct_society(question: str) -> RolePlaying:
        SearchToolkit().search_wiki,
        *ExcelToolkit().get_tools(),
        *DocumentProcessingToolkit(model=models["document"]).get_tools(),
-        *FileWriteToolkit(output_dir="./").get_tools(),
+        *FileToolkit().get_tools(),
    ]

    # Configure agent roles and parameters
--- a/examples/run_azure_openai.py
+++ b/examples/run_azure_openai.py
@@ -22,7 +22,7 @@ from camel.toolkits import (
    ImageAnalysisToolkit,
    SearchToolkit,
    BrowserToolkit,
-    FileWriteToolkit,
+    FileToolkit,
 )
 from camel.types import ModelPlatformType

@@ -77,7 +77,7 @@ def construct_society(question: str) -> OwlRolePlaying:
        SearchToolkit().search_google,  # Comment this out if you don't have google search
        SearchToolkit().search_wiki,
        *ExcelToolkit().get_tools(),
-        *FileWriteToolkit(output_dir="./").get_tools(),
+        *FileToolkit().get_tools(),
    ]

    # Configure agent roles and parameters
--- a/examples/run_claude.py
+++ b/examples/run_claude.py
@@ -23,7 +23,7 @@ from camel.toolkits import (
    SearchToolkit,
    VideoAnalysisToolkit,
    BrowserToolkit,
-    FileWriteToolkit,
+    FileToolkit,
 )
 from camel.types import ModelPlatformType, ModelType
 from camel.logger import set_log_level
@@ -101,7 +101,7 @@ def construct_society(question: str) -> RolePlaying:
        SearchToolkit().search_wiki,
        *ExcelToolkit().get_tools(),
        *DocumentProcessingToolkit(model=models["document"]).get_tools(),
-        *FileWriteToolkit(output_dir="./").get_tools(),
+        *FileToolkit().get_tools(),
    ]

    # Configure agent roles and parameters
--- a/examples/run_cli.py
+++ b/examples/run_cli.py
@@ -17,7 +17,7 @@ from camel.models import ModelFactory
 from camel.toolkits import (
    ExcelToolkit,
    SearchToolkit,
-    FileWriteToolkit,
+    FileToolkit,
    CodeExecutionToolkit,
    BrowserToolkit,
    VideoAnalysisToolkit,
@@ -154,7 +154,7 @@ def construct_society() -> RolePlaying:
        SearchToolkit().search_bing,
        *ExcelToolkit().get_tools(),
        *DocumentProcessingToolkit(model=models["document"]).get_tools(),
-        *FileWriteToolkit(output_dir="./").get_tools(),
+        *FileToolkit().get_tools(),
    ]

    # Configure agent roles and parameters
--- a/examples/run_deepseek_zh.py
+++ b/examples/run_deepseek_zh.py
@@ -24,7 +24,7 @@ from camel.models import ModelFactory
 from camel.toolkits import (
    ExcelToolkit,
    SearchToolkit,
-    FileWriteToolkit,
+    FileToolkit,
    CodeExecutionToolkit,
 )
 from camel.types import ModelPlatformType, ModelType
@@ -73,7 +73,7 @@ def construct_society(question: str) -> RolePlaying:
        SearchToolkit().search_wiki,
        SearchToolkit().search_baidu,
        *ExcelToolkit().get_tools(),
-        *FileWriteToolkit(output_dir="./").get_tools(),
+        *FileToolkit().get_tools(),
    ]

    # Configure agent roles and parameters
--- a/examples/run_gaia_roleplaying.py
+++ b/examples/run_gaia_roleplaying.py
@@ -27,7 +27,7 @@ from camel.toolkits import (
    SearchToolkit,
    VideoAnalysisToolkit,
    BrowserToolkit,
-    FileWriteToolkit,
+    FileToolkit,
 )
 from camel.types import ModelPlatformType, ModelType
 from camel.configs import ChatGPTConfig
@@ -108,7 +108,7 @@ def main():
        *ImageAnalysisToolkit(model=models["image"]).get_tools(),
        *SearchToolkit().get_tools(),
        *ExcelToolkit().get_tools(),
-        *FileWriteToolkit(output_dir="./").get_tools(),
+        *FileToolkit().get_tools(),
    ]

    # Configure agent roles and parameters
--- a/examples/run_gemini.py
+++ b/examples/run_gemini.py
@@ -21,7 +21,7 @@ from camel.toolkits import (
    ImageAnalysisToolkit,
    SearchToolkit,
    BrowserToolkit,
-    FileWriteToolkit,
+    FileToolkit,
 )
 from camel.types import ModelPlatformType, ModelType
 from camel.logger import set_log_level
@@ -99,7 +99,7 @@ def construct_society(question: str) -> RolePlaying:
        SearchToolkit().search_wiki,
        *ExcelToolkit().get_tools(),
        *DocumentProcessingToolkit(model=models["document"]).get_tools(),
-        *FileWriteToolkit(output_dir="./").get_tools(),
+        *FileToolkit().get_tools(),
    ]

    # Configure agent roles and parameters
--- a/examples/run_groq.py
+++ b/examples/run_groq.py
@@ -37,7 +37,7 @@ from camel.toolkits import (
    SearchToolkit,
    VideoAnalysisToolkit,
    BrowserToolkit,
-    FileWriteToolkit,
+    FileToolkit,
 )
 from camel.types import ModelPlatformType, ModelType
 from camel.logger import set_log_level
@@ -114,7 +114,7 @@ def construct_society(question: str) -> OwlRolePlaying:
        SearchToolkit().search_wiki,
        *ExcelToolkit().get_tools(),
        *DocumentProcessingToolkit(model=models["document"]).get_tools(),
-        *FileWriteToolkit(output_dir="./").get_tools(),
+        *FileToolkit().get_tools(),
    ]

    # Configure agent roles and parameters
--- a/examples/run_mini.py
+++ b/examples/run_mini.py
@@ -18,7 +18,7 @@ from camel.models import ModelFactory
 from camel.toolkits import (
    SearchToolkit,
    BrowserToolkit,
-    FileWriteToolkit,
+    FileToolkit,
    CodeExecutionToolkit,
 )
 from camel.types import ModelPlatformType, ModelType
@@ -82,7 +82,7 @@ def construct_society(question: str) -> RolePlaying:
        *CodeExecutionToolkit(sandbox="subprocess", verbose=True).get_tools(),
        SearchToolkit().search_duckduckgo,
        SearchToolkit().search_wiki,
-        *FileWriteToolkit(output_dir="./").get_tools(),
+        *FileToolkit().get_tools(),
    ]

    # Configure agent roles and parameters
--- a/examples/run_mistral.py
+++ b/examples/run_mistral.py
@@ -82,7 +82,7 @@ from camel.models import ModelFactory
 from camel.toolkits import FunctionTool
 from camel.types import ModelPlatformType, ModelType
 from camel.logger import set_log_level
-from camel.toolkits import MCPToolkit, FileWriteToolkit, CodeExecutionToolkit
+from camel.toolkits import MCPToolkit, FileToolkit, CodeExecutionToolkit
 from camel.societies import RolePlaying

 from owl.utils.enhanced_role_playing import arun_society
@@ -161,7 +161,7 @@ async def main():
        # Connect to toolkits
        tools = [
            *mcp_toolkit.get_tools(),
-            *FileWriteToolkit().get_tools(),
+            *FileToolkit().get_tools(),
            *CodeExecutionToolkit().get_tools(),
        ]
        society = await construct_society(task, tools)
--- a/examples/run_novita_ai.py
+++ b/examples/run_novita_ai.py
@@ -20,7 +20,7 @@ from camel.toolkits import (
    ExcelToolkit,
    ImageAnalysisToolkit,
    BrowserToolkit,
-    FileWriteToolkit,
+    FileToolkit,
 )
 from camel.types import ModelPlatformType, ModelType
 from camel.logger import set_log_level
@@ -90,7 +90,7 @@ def construct_society(question: str) -> RolePlaying:
        *ImageAnalysisToolkit(model=models["image"]).get_tools(),
        *ExcelToolkit().get_tools(),
        *DocumentProcessingToolkit(model=models["document"]).get_tools(),
-        *FileWriteToolkit(output_dir="./").get_tools(),
+        *FileToolkit().get_tools(),
    ]

    # Configure agent roles and parameters
--- a/examples/run_ollama.py
+++ b/examples/run_ollama.py
@@ -22,7 +22,7 @@ from camel.toolkits import (
    ImageAnalysisToolkit,
    SearchToolkit,
    BrowserToolkit,
-    FileWriteToolkit,
+    FileToolkit,
 )
 from camel.types import ModelPlatformType

@@ -98,7 +98,7 @@ def construct_society(question: str) -> RolePlaying:
        # SearchToolkit().search_google,  # Comment this out if you don't have google search
        SearchToolkit().search_wiki,
        *ExcelToolkit().get_tools(),
-        *FileWriteToolkit(output_dir="./").get_tools(),
+        *FileToolkit().get_tools(),
    ]

    # Configure agent roles and parameters
--- a/examples/run_openai_compatible_model.py
+++ b/examples/run_openai_compatible_model.py
@@ -22,7 +22,7 @@ from camel.toolkits import (
    ImageAnalysisToolkit,
    SearchToolkit,
    BrowserToolkit,
-    FileWriteToolkit,
+    FileToolkit,
 )
 from camel.types import ModelPlatformType

@@ -101,7 +101,7 @@ def construct_society(question: str) -> RolePlaying:
        SearchToolkit().search_google,  # Comment this out if you don't have google search
        SearchToolkit().search_wiki,
        *ExcelToolkit().get_tools(),
-        *FileWriteToolkit(output_dir="./").get_tools(),
+        *FileToolkit().get_tools(),
    ]

    # Configure agent roles and parameters
--- a/examples/run_ppio.py
+++ b/examples/run_ppio.py
@@ -24,7 +24,7 @@ from camel.models import ModelFactory
 from camel.toolkits import (
    ExcelToolkit,
    SearchToolkit,
-    FileWriteToolkit,
+    FileToolkit,
    CodeExecutionToolkit,
 )
 from camel.types import ModelPlatformType, ModelType
@@ -73,7 +73,7 @@ def construct_society(question: str) -> RolePlaying:
        SearchToolkit().search_wiki,
        SearchToolkit().search_baidu,
        *ExcelToolkit().get_tools(),
-        *FileWriteToolkit(output_dir="./").get_tools(),
+        *FileToolkit().get_tools(),
    ]

    # Configure agent roles and parameters
--- a/examples/run_qwen_mini_zh.py
+++ b/examples/run_qwen_mini_zh.py
@@ -22,7 +22,7 @@ from camel.models import ModelFactory
 from camel.toolkits import (
    SearchToolkit,
    BrowserToolkit,
-    FileWriteToolkit,
+    FileToolkit,
 )
 from camel.types import ModelPlatformType, ModelType
 from camel.societies import RolePlaying
@@ -100,7 +100,7 @@ def construct_society(question: str) -> RolePlaying:
            output_language="Chinese",
        ).get_tools(),
        SearchToolkit().search_baidu,
-        *FileWriteToolkit(output_dir="./").get_tools(),
+        *FileToolkit().get_tools(),
    ]

    # Configure agent roles and parameters
--- a/examples/run_qwen_zh.py
+++ b/examples/run_qwen_zh.py
@@ -26,7 +26,7 @@ from camel.toolkits import (
    SearchToolkit,
    VideoAnalysisToolkit,
    BrowserToolkit,
-    FileWriteToolkit,
+    FileToolkit,
 )
 from camel.types import ModelPlatformType, ModelType
 from camel.societies import RolePlaying
@@ -112,7 +112,7 @@ def construct_society(question: str) -> RolePlaying:
        SearchToolkit().search_baidu,
        *ExcelToolkit().get_tools(),
        *DocumentProcessingToolkit(model=models["document"]).get_tools(),
-        *FileWriteToolkit(output_dir="./").get_tools(),
+        *FileToolkit().get_tools(),
    ]

    # Configure agent roles and parameters
--- a/examples/run_terminal.py
+++ b/examples/run_terminal.py
@@ -18,7 +18,7 @@ from camel.models import ModelFactory
 from camel.toolkits import (
    SearchToolkit,
    BrowserToolkit,
-    FileWriteToolkit,
+    FileToolkit,
    TerminalToolkit,
 )
 from camel.types import ModelPlatformType, ModelType
@@ -73,14 +73,14 @@ def construct_society(question: str) -> RolePlaying:

    # 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(),
+        # *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(),
        SearchToolkit().search_duckduckgo,
        SearchToolkit().search_wiki,
-        *FileWriteToolkit(output_dir="./").get_tools(),
+        *FileToolkit().get_tools(),
        *TerminalToolkit().get_tools(),
    ]

--- a/examples/run_terminal_zh.py
+++ b/examples/run_terminal_zh.py
@@ -18,7 +18,7 @@ from camel.models import ModelFactory
 from camel.toolkits import (
    SearchToolkit,
    BrowserToolkit,
-    FileWriteToolkit,
+    FileToolkit,
    TerminalToolkit,
 )
 from camel.types import ModelPlatformType, ModelType
@@ -80,7 +80,7 @@ def construct_society(question: str) -> RolePlaying:
        ).get_tools(),
        SearchToolkit().search_duckduckgo,
        SearchToolkit().search_wiki,
-        *FileWriteToolkit(output_dir="./").get_tools(),
+        *FileToolkit().get_tools(),
        *TerminalToolkit().get_tools(),
    ]

--- a/examples/run_together_ai.py
+++ b/examples/run_together_ai.py
@@ -20,7 +20,7 @@ from camel.toolkits import (
    ExcelToolkit,
    ImageAnalysisToolkit,
    BrowserToolkit,
-    FileWriteToolkit,
+    FileToolkit,
 )
 from camel.types import ModelPlatformType, ModelType
 from camel.logger import set_log_level
@@ -90,7 +90,7 @@ def construct_society(question: str) -> RolePlaying:
        *ImageAnalysisToolkit(model=models["image"]).get_tools(),
        *ExcelToolkit().get_tools(),
        *DocumentProcessingToolkit(model=models["document"]).get_tools(),
-        *FileWriteToolkit(output_dir="./").get_tools(),
+        *FileToolkit().get_tools(),
    ]

    # Configure agent roles and parameters
--- a/owl/utils/enhanced_role_playing.py
+++ b/owl/utils/enhanced_role_playing.py
@@ -67,13 +67,15 @@ class OwlRolePlaying(RolePlaying):

    def _init_agents(
        self,
-        init_assistant_sys_msg: BaseMessage,
-        init_user_sys_msg: BaseMessage,
+        init_assistant_sys_msg: Optional[BaseMessage],
+        init_user_sys_msg: Optional[BaseMessage],
        assistant_agent_kwargs: Optional[Dict] = None,
        user_agent_kwargs: Optional[Dict] = None,
        output_language: Optional[str] = None,
        is_reasoning_task: bool = False,
        stop_event: Optional[threading.Event] = None,
+        assistant_agent: Optional[ChatAgent] = None,
+        user_agent: Optional[ChatAgent] = None,
    ) -> None:
        r"""Initialize assistant and user agents with their system messages.

@@ -105,20 +107,60 @@ class OwlRolePlaying(RolePlaying):
        #         model_platform=ModelPlatformType.OPENAI,
        #         model_type=ModelType.O3_MINI,
        #     )
+        if assistant_agent is not None:
+            # Ensure functionality consistent with our configuration
+            if (
+                hasattr(assistant_agent, 'output_language')
+                and output_language is not None
+            ):
+                assistant_agent.output_language = output_language
+            if hasattr(assistant_agent, 'stop_event'):
+                assistant_agent.stop_event = stop_event
+            self.assistant_agent = assistant_agent
+            # Handle potential None system_message - use provided or fallback
+            if assistant_agent.system_message is not None:
+                self.assistant_sys_msg = assistant_agent.system_message
+            elif init_assistant_sys_msg is not None:
+                self.assistant_sys_msg = init_assistant_sys_msg
+            else:
+                raise ValueError("Assistant system message cannot be None")
+        else:
+            if init_assistant_sys_msg is None:
+                raise ValueError("Assistant system message cannot be None")

-        self.assistant_agent = ChatAgent(
-            init_assistant_sys_msg,
-            output_language=output_language,
-            **(assistant_agent_kwargs or {}),
-        )
-        self.assistant_sys_msg = self.assistant_agent.system_message
+            self.assistant_agent = ChatAgent(
+                init_assistant_sys_msg,
+                output_language=output_language,
+                **(assistant_agent_kwargs or {}),
+            )
+            self.assistant_sys_msg = self.assistant_agent.system_message

-        self.user_agent = ChatAgent(
-            init_user_sys_msg,
-            output_language=output_language,
-            **(user_agent_kwargs or {}),
-        )
-        self.user_sys_msg = self.user_agent.system_message
+        if user_agent is not None:
+            # Ensure functionality consistent with our configuration
+            if (
+                hasattr(user_agent, 'output_language')
+                and output_language is not None
+            ):
+                user_agent.output_language = output_language
+            if hasattr(user_agent, 'stop_event'):
+                user_agent.stop_event = stop_event
+            self.user_agent = user_agent
+            # Handle potential None system_message - use provided or fallback
+            if user_agent.system_message is not None:
+                self.user_sys_msg = user_agent.system_message
+            elif init_user_sys_msg is not None:
+                self.user_sys_msg = init_user_sys_msg
+            else:
+                raise ValueError("User system message cannot be None")
+        else:
+            if init_user_sys_msg is None:
+                raise ValueError("User system message cannot be None")
+            self.user_agent = ChatAgent(
+                init_user_sys_msg,
+                output_language=output_language,
+                **(user_agent_kwargs or {}),
+            )
+            self.user_sys_msg = self.user_agent.system_message

    # def _judge_if_reasoning_task(self, question: str) -> bool:
    #     r"""Judge if the question is a reasoning task."""
--- a/owl/utils/gaia.py
+++ b/owl/utils/gaia.py
@@ -19,6 +19,7 @@ import json
 import random
 import re
 import string
+import pandas as pd
 from pathlib import Path
 from typing import Any, Dict, List, Literal, Optional, Union, Tuple

@@ -120,15 +121,31 @@ class GAIABenchmark(BaseBenchmark):
        # Load metadata for both validation and test datasets
        for path, label in zip([valid_dir, test_dir], ["valid", "test"]):
            self._data[label] = []
-            with open(path / "metadata.jsonl", "r") as f:
-                lines = f.readlines()
-                for line in lines:
-                    data = json.loads(line)
+            metadata_jsonl_path = path / "metadata.jsonl"
+            metadata_parquet_path = path / "metadata.parquet"
+            if metadata_parquet_path.exists():
+                raw_data = pd.read_parquet(metadata_parquet_path)
+                # convert to dict
+                raw_data = raw_data.to_dict(orient="records")
+                for data in raw_data:
                    if data["task_id"] == "0-0-0-0-0":
                        continue
                    if data["file_name"]:
                        data["file_name"] = path / data["file_name"]
                    self._data[label].append(data)
+            elif metadata_jsonl_path.exists():
+                with open(metadata_jsonl_path, "r") as f:
+                    lines = f.readlines()
+                    for line in lines:
+                        data = json.loads(line)
+                        if data["task_id"] == "0-0-0-0-0":
+                            continue
+                        if data["file_name"]:
+                            data["file_name"] = path / data["file_name"]
+                        self._data[label].append(data)
+            else:
+                raise FileNotFoundError(f"Metadata file not found: {metadata_parquet_path} or {metadata_jsonl_path}")
+
        return self

    @property
@@ -167,7 +184,7 @@ class GAIABenchmark(BaseBenchmark):
                f"Invalid value for `level`: {level}, expected 1, 2, 3 " "or 'all'."
            )
        logger.info(f"Running benchmark on {on} set at levels {levels}.")
-        datas = [data for data in self._data[on] if data["Level"] in levels]
+        datas = [data for data in self._data[on] if int(data["Level"]) in levels]
        # Shuffle and subset data if necessary
        if randomize:
            random.shuffle(datas)
--- a/requirements.txt
+++ b/requirements.txt
@@ -1,4 +1,4 @@
-camel-ai[owl]==0.2.57
+camel-ai[owl]==0.2.84
 docx2markdown>=0.1.1
 gradio>=3.50.2
 mcp-simple-arxiv==0.2.2