Zhipu AI

author: Jael

Description 描述

This operator is implemented with ChatGLM services from Zhipu AI. It directly returns the original response in dictionary without parsing. Please note you will need API Key to access the service.

LLM/ZhipuAI 利用了来自智谱AI开放平台的大语言模型服务。该算子以字典的形式直接返回原始的模型回复。请注意，您需要API Key才能访问该服务。

Code Example 代码示例

Write a pipeline with explicit inputs/outputs name specifications:

from towhee import pipe, ops

p = (
    pipe.input('messages')
        .map('messages', 'response', ops.LLM.ZhipuAI(
            api_key=ZHIPUAI_API_KEY,
            model_name='chatglm_130b',  # or 'chatglm_6b'
            temperature=0.5,
            max_tokens=50,
            ))
        .output('response')
)

messages=[
    {'system': '你是一个资深的软件工程师，善于回答关于科技项目的问题。'},
    {'question': 'Zilliz Cloud 是什么？', 'answer': 'Zilliz Cloud 是一种全托管的向量检索服务。'},
    {'question': '它和 Milvus 的关系是什么？'}
    ]
response = p(messages).get()[0]
answer = response['choices'][0]['content']
token_usage = response['usage']

Factory Constructor 接口说明

Create the operator via the following factory method:

LLM.ZhipuAI(api_key: str, model_name: str, **kwargs)

Parameters:

api_key: str=None

The Zhipu AI API key in string, defaults to None. If None, it will use the environment variable ZHIPUAI_API_KEY.

model_name: str='chatglm_130b'

The model used in Zhipu AI service, defaults to 'chatglm_130b'. Visit Zhipu AI documentation for supported models.

**kwargs

Other ChatGLM parameters such as temperature, etc.

Interface 使用说明

The operator takes a piece of text in string as input. It returns answer in json.

__call__(txt)

Parameters:

messages: list

​ A list of messages to set up chat. Must be a list of dictionaries with key value from "system", "question", "answer". For example, [{"question": "a past question?", "answer": "a past answer."}, {"question": "current question?"}]. It also accepts the orignal ChatGLM message format like [{"role": "user", "content": "a question?"}, {"role": "assistant", "content": "an answer."}]

Returns:

response: dict

The original llm response in dictionary, including next answer and token usage.