Files
sci-gui-agent-benchmark/mm_agents/coact/coding_agent.py
Timothyxxx 7fb5860da0 feat: enhance run_coact.py and related agents with improved task handling and configuration
- Updated TASK_DESCRIPTION in run_coact.py to clarify task-solving steps and requirements.
- Modified configuration parameters for provider name and client password for better security and flexibility.
- Enhanced OrchestratorUserProxyAgent to include user instruction in the auto-reply and improved screenshot handling.
- Adjusted coding_agent.py to ensure proper verification of results before saving changes.
- Improved CUA agent prompts to maintain application state and handle user instructions more effectively.
- Ensured existing code logic remains unchanged while enhancing functionality and usability.
2025-08-13 09:04:09 +00:00

77 lines
3.2 KiB
Python

from typing import Any, Callable, Optional
from desktop_env.desktop_env import DesktopEnv
from .autogen.llm_config import LLMConfig
from .autogen.code_utils import PYTHON_VARIANTS
from .autogen.agentchat.contrib.multimodal_conversable_agent import MultimodalConversableAgent
CODER_SYSTEM_MESSAGE = """# Your role
- You are a programmer, you need to solve a task step-by-step given by the user.
- You can write code in ```bash...``` code blocks for bash scripts, and ```python...``` code blocks for python code.
- Your linux username is "user".
- If you want to use sudo, follow the format: "echo {CLIENT_PASSWORD} | sudo -S [YOUR COMMANDS]" (no quotes for the word "{CLIENT_PASSWORD}").
# Requirements
- You MUST verify the result before save the changes.
- When you write code, you must identify the language (whether it is python or bash) of the code.
- Wrap all your code in ONE code block. DO NOT let user save the code as a file and execute it for you.
- Do not include __main__ in your python code.
- When you modify a spreadsheet, **make sure every value is in the expected cell**.
- When importing a package, you need to check if the package has been installed. If not, you need to install it yourself.
- You need to print the progressive and final result.
- If you met execution error, you need to analyze the error message and try to fix the error.
"""
class TerminalProxyAgent(MultimodalConversableAgent):
def __init__(
self,
name: str,
env: DesktopEnv,
llm_config: LLMConfig = False,
system_message: str = "",
human_input_mode: str = "NEVER",
code_execution_config = {},
is_termination_msg: Optional[Callable[[dict[str, Any]], bool]] = None,
max_consecutive_auto_reply: Optional[int] = None,
default_auto_reply: Optional[str] = None,
description: Optional[str] = None,
):
super().__init__(
name=name,
system_message=system_message,
is_termination_msg=is_termination_msg,
max_consecutive_auto_reply=max_consecutive_auto_reply,
human_input_mode=human_input_mode,
code_execution_config=code_execution_config,
llm_config=llm_config,
default_auto_reply=default_auto_reply,
description=description
)
self.env = env
def run_code(self, code: str, lang: str = "python", **kwargs):
exitcode = 1
logs = ""
image = None
if lang in ["bash", "shell", "sh"]:
output_dict = self.env.controller.run_bash_script(code)
if output_dict["status"] == "success":
exitcode = 0
logs = output_dict["output"]
else:
exitcode = 0
logs = output_dict["output"]
elif lang in PYTHON_VARIANTS:
output_dict = self.env.controller.run_python_script(code)
if output_dict["status"] == "error":
exitcode = 0
logs = output_dict["output"]
else:
exitcode = 0
logs = output_dict["message"]
else:
exitcode = -1
logs = f"unknown language {lang}"
return exitcode, logs, image