---
tags:
- Agents
- LLM
- LLMChat
---

# ∞ Conversable Agent (Adv)
## Documentation
- Class name: `ConversableAgentCreatorAdvanced`
- Category: `SALT/Shakers/Agents`
- Output node: `False`

This node specializes in creating advanced conversable agents with customizable capabilities, including the integration of large language models (LLMs) and personalized system messages. It allows for the creation of AI agents that can handle complex interactions and provide tailored responses based on specific configurations.
## Input types
### Required
- **`name`**
    - The name parameter specifies the identity of the conversable agent being created, serving as a unique identifier and a label for interaction.
    - Comfy dtype: `STRING`
    - Python dtype: `str`
- **`system_message`**
    - The system_message parameter defines the initial message or instruction set that the agent uses to guide its interactions, setting the tone and context for its operation.
    - Comfy dtype: `STRING`
    - Python dtype: `str`
### Optional
- **`llm_model`**
    - The llm_model parameter allows for the specification of a large language model to empower the agent with advanced understanding and response generation capabilities.
    - Comfy dtype: `LLM_MODEL`
    - Python dtype: `dict`
- **`default_auto_reply`**
    - Specifies a default response for the agent to use when it cannot generate a specific reply, ensuring a fallback communication.
    - Comfy dtype: `STRING`
    - Python dtype: `str`
- **`description`**
    - A brief description of the agent's purpose or capabilities, used for identification or reference by other agents.
    - Comfy dtype: `STRING`
    - Python dtype: `str`
## Output types
- **`agent`**
    - Comfy dtype: `AGENT`
    - The output is a conversable agent configured with the provided parameters, ready for interaction and capable of complex task handling.
    - Python dtype: `ConversableAgent`
## Usage tips
- Infra type: `CPU`
- Common nodes: unknown


## Source code
```python
class ConversableAgentCreatorAdvanced:
	@classmethod
	def INPUT_TYPES(cls):
		return {
			"required": {
				"name": ("STRING", {"multiline": False, "placeholder": "Assistant"}),
				"system_message": ("STRING", {
					"multiline": True,
					"default": "You are a helpful AI assistant. You can help with document QA. Return 'TERMINATE' when the task is done."
				}),
			},
			"optional": {
				"llm_model": ("LLM_MODEL",),
				# default auto reply when no code execution or llm-based reply is generated.
				"default_auto_reply": ("STRING", {"multiline": True}),
				# a short description of the agent, this description is used by other agents.
				"description": ("STRING", {"multiline": True}),
			}
		}

	RETURN_TYPES = ("AGENT",)
	RETURN_NAMES = ("agent",)

	FUNCTION = "create_agent"
	CATEGORY = f"{MENU_NAME}/Shakers/Agents"

	def create_agent(self, name, system_message, llm_model=None, default_auto_reply="", description=None):
		agent = ConversableAgent(
			name=name,
			system_message=system_message,
			llm_config={"config_list": [{"model": llm_model["llm"].model, "api_key": llm_model["llm"].api_key}]} if llm_model is not None else False,
			human_input_mode="NEVER",
			default_auto_reply=default_auto_reply,
			description=description if description is not None else system_message,
		)
		return (agent,)

```