agentverse/tasks/tasksolving/humaneval/gpt-4/config.yaml

cnt_agents: &cnt_agents 4
max_turn: &max_turn 2
max_criticizing_rounds: 2

prompts:
  role_assigner_prepend_prompt: &role_assigner_prepend_prompt |-
    # Role Description
    You are the leader of a group of experts, now you need to recruit a small group of experts with diverse identity to correctly write the code to solve the given problems:
    ${task_description}
    
    You can recruit ${cnt_critic_agents} expert in different fields. What experts will you recruit to better generate an accurate solution?

    Here are some suggestion:
    ${advice}
    
  role_assigner_append_prompt: &role_assigner_append_prompt |-
    # Response Format Guidance
    You should respond with a list of expert description. For example:
    1. an electrical engineer specified in the filed of xxx.
    2. an economist who is good at xxx.
    3. a lawyer with a good knowledge of xxx.
    ...

    Only respond with the description of each role. Do not include your reason.

  solver_prepend_prompt: &solver_prepend_prompt |-
    Can you complete the following code?
    ```python 
    ${task_description} 
    ```

  # # Previous Solution
  # The solution you gave in the last step is:
  # ${former_solution}

  # # Evaluation
  # The unit testing gave the following suggestion:
  # ${advice}

  # # Critics
  # The following messages are the critics on the provided solution:
  
  solver_append_prompt: &solver_append_prompt |-
    You are ${role_description}. Provide a correct completion of the code. Explain your reasoning. Your response should contain only Python code. Do not give any additional information. Use ```python to put the completed Python code in markdown quotes. When responding, please include the given code and the completion.

  critic_prepend_prompt: &critic_prepend_prompt |-
    You are in a discussion group, aiming to complete the following code function:
    ```python
    ${task_description}
    ```

  # Below is a possible code completion:
  # ```
  # ${preliminary_solution}
  # ```

  critic_append_prompt: &critic_append_prompt |-
    You are ${role_description}. Based on your knowledge, can you check the functional correctness of the latest completion given above? When responding, you should follow the following rules:
    1. If the latest provided solution is correct, end your response with a special token "[Agree]". 
    2. If the solution is incorrect, give your comment and end your response with a special token "[Disagree]".
  

  manager_prompt: &manager_prompt |-
    According to the Previous Solution and the Previous Sentences, select the most appropriate Critic from a specific Role and output the Role.
    ```python 
    ${task_description} 
    ```
    # Previous Solution
    The solution you gave in the last step is:
    ${former_solution}

    # Critics
    There are some critics on the above solution:
    ```
    ${critic_opinions}
    ```

    # Previous Sentences
    The previous sentences in the previous rounds is:
    ${previous_sentence}

  executor_prepend_prompt: &executor_prepend_prompt |-
    You are an experienced program tester. Now your team is trying to solve the problem: 
    '''
    Complete the Python function:
    ${task_description}
    '''

  executor_append_prompt: &executor_append_prompt |-
    The solution has been written to `tmp/main.py`. Your are going to write the unit testing code for the solution. You should respond in the following format:
    Thought: your thought
    Reasoning: your reasoning on the testing cases
    Criticism: constructive self-criticism
    File Path: the path to write your testing code
    Code: the testing code with explaination in docstring. make sure to write the input in the assertion to make it appear in the unit test report, and make sure the expected answer is correct
    Command: the command to change directory and execute your testing code

  evaluator_prepend_prompt: &evaluator_prepend_prompt |-
    # Problem
    Complete the following function
    ```python
    ${task_description}
    ```
    
    # Experts
    The experts recruited in this turn includes:
    ${all_role_description}

    # Writer's Solution: 
    ${solution}

    # Tester's Feedback:
    ${result}

  evaluator_append_prompt: &evaluator_append_prompt |-
    # Response Format Guidance
    You must respond in the following format:
    Score: (0 or 1, 0 for incorrect and 1 for correct)
    Response: (give your advice on how to correct the solution, and your suggestion on what experts should recruit in the next round)
    

name: pipeline


environment:
  env_type: task-basic
  max_turn: *max_turn
  rule:
    role_assigner:
      type: role_description
      cnt_agents: *cnt_agents
    decision_maker:
      type: vertical-solver-first
    executor:
      type: code-test
    evaluator:
      type: basic

agents:
  - #role_assigner_agent:
    agent_type: role_assigner
    name: role assigner
    max_retry: 1000
    prepend_prompt_template: *role_assigner_prepend_prompt
    append_prompt_template: *role_assigner_append_prompt
    memory:
      memory_type: chat_history
    llm:
      llm_type: gpt-4
      model: "gpt-4"
      temperature: 0
      max_tokens: 512
    output_parser:
      type: role_assigner

  - #solver_agent:
    agent_type: solver
    name: Planner
    max_retry: 1000
    prepend_prompt_template: *solver_prepend_prompt
    append_prompt_template: *solver_append_prompt
    memory:
      memory_type: chat_history
    llm:
      llm_type: gpt-4
      model: "gpt-4"
      temperature: 0
      max_tokens: 1024
    output_parser:
      type: humaneval-solver
      # stop:
      #   - "\ndef "
      #   - "\nclass "
      #   - "\nif "
      #   - "\n\n#"

  - #critic_agents:
    agent_type: critic
    name: Critic 1
    max_retry: 1000
    role_description: |-
      Waiting to be assigned.
    prepend_prompt_template: *critic_prepend_prompt
    append_prompt_template: *critic_append_prompt
    memory:
      memory_type: chat_history
    llm:
      llm_type: gpt-4
      model: "gpt-4"
      temperature: 0
      max_tokens: 1024
    output_parser:
      type: mgsm-critic-agree

  - #executor_agent:
    agent_type: executor
    name: Executor
    max_retry: 1000
    prepend_prompt_template: *executor_prepend_prompt
    append_prompt_template: *executor_append_prompt
    memory:
      memory_type: chat_history
    llm:
      llm_type: gpt-4
      model: gpt-4
      temperature: 0
      max_tokens: 1024
    output_parser:
      type: humaneval-executor

  - #evaluator_agent:
    agent_type: evaluator
    name: Evaluator
    max_retry: 1000
    role_description: |-
      Evaluator
    prepend_prompt_template: *evaluator_prepend_prompt
    append_prompt_template: *evaluator_append_prompt
    memory:
      memory_type: chat_history
    llm:
      llm_type: gpt-4
      model: gpt-4
      temperature: 0.3
      max_tokens: 1024
    output_parser:
      type: mgsm-evaluator
      dimensions:
        - Score


  - #manager_agent:
    agent_type: manager
    name: Manager
    max_retry: 1000
    prompt_template: *manager_prompt
    memory:
      memory_type: chat_history
    llm:
      llm_type: gpt-4
      model: "gpt-4"
      temperature: 0
      max_tokens: 1024
    output_parser:
      type: humaneval-manager