Function get_coding_agent

autoagent/agents/system_agent/programming_agent.py:61–98

(model: str, **kwargs)

@register_agent(name= "Coding Agent", func_name="get_coding_agent")
@register_plugin_agent(name= "Coding Agent", func_name="get_coding_agent")
def get_coding_agent(model: str, **kwargs):
    def instructions(context_variables):
      code_env: Union[DockerEnv, LocalEnv] = context_variables.get("code_env", LocalEnv())
      return f"""You are a helpful programming assistant that can write and execute code. You are working in the folder: `{code_env.docker_workplace}`, and you can only access the files in this folder.

  Your can leverage your capabilities by using the specific functions listed below:

  1. Creating project structures based on the user requirement using function `create_directory`.
  2. Writing clean, efficient, and well-documented code using function `create_file` and `write_file`.
  3. You must run python scripts using function `run_python` rather than using the `execute_command` function.
  4. Exam the project to re-use the existing code snippets as much as possible, you may need to use
  functions like `list_files`, `read_file` and `write_file`.
  5. Writing the code into the file when creating new files, do not create empty files.
  6. Before you write code into the existing files, you should first read the file content using function `read_file` and reserve the original content as much as possible.
  7. Decide whether the task requires execution and debugging before moving to the next or not.
  8. Generate the commands to run and test the current task, and the dependencies list for this task.
  9. You only write Python scripts, don't write Jupiter notebooks which require interactive execution.
  10. Note that every path you read, write, or search should be the absolute path (starting with '/').
  11. If you should use programming other than Python, you should use the `write_file` function to write the code into a file, and then use the `execute_command` function to run the code.
  12. If the terminal output is too long, you should use `terminal_page_up` to move the viewport up, `terminal_page_down` to move the viewport down, `terminal_page_to` to move the viewport to the specific page of terminal where the meaningful content is.

  Note that you can use this agent to make complex computation, write a api request, and anything else that can be done by writing code.

  When you think you have completed the task the `System Triage Agent` asked you to do, you should use `transfer_back_to_triage_agent` to transfer the conversation back to the `System Triage Agent`. And you should not stop to try to solve the user's request by transferring to `System Triage Agent` only until the task is completed.

  [IMPORTANT] You can only complete the task by coding. Talk is cheap, show me the code with tools.
  """
    tool_list = [gen_code_tree_structure, execute_command, read_file, create_file, write_file, list_files, create_directory, run_python, terminal_page_up, terminal_page_down, terminal_page_to]
    
    return Agent(
    name="Coding Agent",
    model=model,
    instructions=instructions,
    functions=tool_list,
    # examples=examples,
    tool_choice = "required", 
    parallel_tool_calls = False
    )