Neutrino AI¶

Neutrino 能够智能地将查询路由至最适合当前提示词的 LLM，在优化成本和延迟的同时实现最佳性能。

访问我们的官网：neutrinoapp.com 文档中心：docs.neutrinoapp.com 创建 API 密钥：platform.neutrinoapp.com

In [ ]:

%pip install llama-index-llms-neutrino

%pip install llama-index-llms-neutrino

In [ ]:

!pip install llama-index

!pip install llama-index

设置 Neutrino API 密钥环境变量¶

您可以在以下地址创建 API 密钥：platform.neutrinoapp.com

In [ ]:

import os

os.environ["NEUTRINO_API_KEY"] = "<your-neutrino-api-key>"

import os os.environ["NEUTRINO_API_KEY"] = ""

使用路由器功能¶

路由器（router）是由多个大语言模型（LLM）组成的集合，您可以将查询请求路由至其中的模型。您可以通过 Neutrino 控制面板创建自定义路由器，或直接使用默认路由器（该路由器包含所有受支持的模型）。 您可以将路由器视为一个大语言模型来使用。

In [ ]:

from llama_index.llms.neutrino import Neutrino
from llama_index.core.llms import ChatMessage

llm = Neutrino(
    # api_key="<your-neutrino-api-key>",
    # router="<your-router-id>"  # (or 'default')
)

response = llm.complete("In short, a Neutrino is")
print(f"Optimal model: {response.raw['model']}")
print(response)

from llama_index.llms.neutrino import Neutrino from llama_index.core.llms import ChatMessage llm = Neutrino( # api_key="", # router="" # (or 'default') ) response = llm.complete("In short, a Neutrino is") print(f"Optimal model: {response.raw['model']}") print(response)

Optimal model: gpt-3.5-turbo
a subatomic particle that is electrically neutral and has a very small mass. It is one of the fundamental particles that make up the universe. Neutrinos are extremely difficult to detect because they interact very weakly with matter, making them able to pass through most materials without any interaction. They are produced in various natural processes, such as nuclear reactions in the Sun and other stars, as well as in particle interactions on Earth. Neutrinos have played a significant role in advancing our understanding of particle physics and the universe.

In [ ]:

message = ChatMessage(
    role="user",
    content="Explain the difference between statically typed and dynamically typed languages.",
)
resp = llm.chat([message])
print(f"Optimal model: {resp.raw['model']}")
print(resp)

message = ChatMessage( role="user", content="Explain the difference between statically typed and dynamically typed languages.", ) resp = llm.chat([message]) print(f"Optimal model: {resp.raw['model']}") print(resp)

Optimal model: mistralai/Mixtral-8x7B-Instruct-v0.1
assistant:  Statically typed languages and dynamically typed languages are categories of programming languages based on how they handle variable types.

In statically typed languages, the type of a variable is determined at compile-time, which means that the type is checked before the program is run. This ensures that variables are always used in a type-safe manner, preventing many types of errors from occurring at runtime. Examples of statically typed languages include Java, C, C++, and C#.

In dynamically typed languages, the type of a variable is determined at runtime, which means that the type is checked as the program is running. This provides more flexibility, as variables can be assigned values of different types at different times, but it also means that type errors may not be caught until the program is running, which can make debugging more difficult. Examples of dynamically typed languages include Python, Ruby, JavaScript, and PHP.

One key difference between statically typed and dynamically typed languages is that statically typed languages tend to be more verbose, requiring explicit type declarations for variables, while dynamically typed languages are more concise and allow for implicit type declarations. However, this also means that statically typed languages can catch type errors earlier in the

流式响应¶

In [ ]:

message = ChatMessage(
    role="user", content="What is the approximate population of Mexico?"
)
resp = llm.stream_chat([message])
for i, r in enumerate(resp):
    if i == 0:
        print(f"Optimal model: {r.raw['model']}")
    print(r.delta, end="")

message = ChatMessage( role="user", content="What is the approximate population of Mexico?" ) resp = llm.stream_chat([message]) for i, r in enumerate(resp): if i == 0: print(f"Optimal model: {r.raw['model']}") print(r.delta, end="")

Optimal model: anthropic.claude-instant-v1
 According to the latest UN estimates, the population of Mexico is approximately 128 million as of 2020. Mexico has the 10th largest population in the world.