68 lines
2.0 KiB
Markdown
68 lines
2.0 KiB
Markdown
|
|
# Documentation
|
|
- Class name: GenerateReply
|
|
- Category: SALT/Language Toolkit/Agents
|
|
- Output node: False
|
|
|
|
GenerateReply节点用于生成对话代理之间的回复。它基于输入的消息和对话上下文生成回复，抽象了会话代理交互的复杂性，从而实现动态和响应性的对话创建。
|
|
|
|
# Input types
|
|
## Required
|
|
- recipient
|
|
- 指定预期接收消息的代理，在确定生成回复的上下文和内容方面起着关键作用。
|
|
- Comfy dtype: AGENT
|
|
- Python dtype: ConversableAgent
|
|
- message
|
|
- 接收代理预期回复的输入消息。这条消息为回复生成设定了上下文。
|
|
- Comfy dtype: STRING
|
|
- Python dtype: str
|
|
|
|
## Optional
|
|
- sender
|
|
- 可选参数，用于识别消息发送者，有助于在对话上下文中更准确地定制回复。
|
|
- Comfy dtype: AGENT
|
|
- Python dtype: ConversableAgent or None
|
|
|
|
# Output types
|
|
- reply
|
|
- 接收代理对输入消息生成的回复，体现了代理在持续对话中的响应。
|
|
- Comfy dtype: STRING
|
|
- Python dtype: str
|
|
|
|
|
|
## Usage tips
|
|
- Infra type: `CPU`
|
|
- Common nodes: unknown
|
|
|
|
|
|
## Source code
|
|
```python
|
|
class GenerateReply:
|
|
@classmethod
|
|
def INPUT_TYPES(cls):
|
|
return {
|
|
"required": {
|
|
"recipient": ("AGENT",),
|
|
"message": ("STRING", {"multiline": True, "dynamicPrompts": False}),
|
|
},
|
|
"optional": {
|
|
"sender": ("AGENT", {"default": None}),
|
|
}
|
|
}
|
|
|
|
RETURN_TYPES = ("STRING",)
|
|
RETURN_NAMES = ("reply",)
|
|
|
|
FUNCTION = "start_chat"
|
|
CATEGORY = f"{MENU_NAME}/{SUB_MENU_NAME}/Agents"
|
|
|
|
def start_chat(self, recipient, message, sender=None):
|
|
recipient = clone_conversable_agent(recipient)
|
|
if sender:
|
|
sender = clone_conversable_agent(sender)
|
|
message = recipient._oai_messages[sender] + [{"content": message, "role": "user"}]
|
|
reply = recipient.generate_reply(message, sender)
|
|
return (reply,)
|
|
|
|
```
|