I Tried AISuite by AndrewNg, and It’s GREAT!

Andrew Ng not too long ago launched AISuite, an open-source Python bundle designed to streamline the usage of giant language fashions (LLMs) throughout a number of suppliers. This modern device simplifies the complexities of working with numerous LLMs by permitting seamless switching between fashions with a easy “supplier:mannequin” string. By considerably decreasing integration overhead, AISuite enhances flexibility and accelerates software growth, making it a useful useful resource for builders navigating the dynamic panorama of AI. On this article, we’ll see how efficient it’s.

What’s AISuite?

AISuite is an open-source venture led by Andrew Ng, designed to make working with a number of giant language mannequin (LLM) suppliers simpler and extra environment friendly. Out there on GitHub, it gives a easy, unified interface that enables seamless switching between LLMs utilizing HTTP endpoints or SDKs, following OpenAI’s interface. This device is good for college kids, educators, and builders, providing constant and hassle-free interactions throughout numerous suppliers.

Supported by a group of open-source contributors, AISuite bridges the hole between completely different LLM frameworks. It allows customers to combine and evaluate fashions from suppliers like OpenAI, Anthropic, and Meta’s Llama with ease. The device simplifies duties reminiscent of producing textual content, conducting analyses, and constructing interactive techniques. With options like streamlined API key administration, customizable consumer configurations, and an intuitive setup, AISuite helps each easy purposes and sophisticated LLM-based initiatives.

Implementation of AISuite

1. Set up Essential Libraries

!pip set up openai
!pip set up aisuite[all]

• !pip set up openai: Installs the OpenAI Python library, which is required to work together with OpenAI’s GPT fashions.
• !pip set up aisuite[all]: Installs AISuite together with non-compulsory dependencies wanted to help a number of LLM suppliers.

2. Set API Keys for Authentication

os.environ['OPENAI_API_KEY'] = getpass('Enter your OPENAI API key: ')
os.environ['ANTHROPIC_API_KEY'] = getpass('Enter your ANTHROPIC API key: ')

• os.environ: Units setting variables to securely retailer the API keys required to entry LLM providers.
• getpass(): Prompts the consumer to enter their OpenAI and Anthropic API keys securely (with out displaying the enter).
• These keys authenticate your requests to the respective platforms.

Additionally learn: Find out how to Generate Your Personal OpenAI API Key and Add Credit?

3. Initialize the AISuite Shopper

consumer = ai.Shopper()

This initializes an occasion of the AISuite consumer, permitting interplay with a number of LLMs in a standardized approach.

4. Outline the Immediate (Messages)

messages = [
   {"role": "system", "content": "Talk using Pirate English."},
   {"role": "user", "content": "Tell a joke in 1 line."}
]

• The messages checklist defines a dialog enter:
  • position: “system”: Supplies directions to the mannequin (e.g., “Discuss utilizing Pirate English”).
  • position: “consumer”: Represents the consumer’s question (e.g., “Inform a joke in 1 line”).
• This immediate ensures the responses comply with a pirate theme and embrace a one-line joke.

5. Question the OpenAI Mannequin

response = consumer.chat.completions.create(mannequin="openai:gpt-4o", messages=messages, temperature=0.75)
print(response.selections[0].message.content material)

• mannequin=”openai:gpt-4o”: Specifies the OpenAI GPT-4o mannequin.
• messages=messages: Sends the immediate outlined earlier to the mannequin.
• temperature=0.75: Controls the randomness of the response. A better temperature leads to extra inventive outputs, whereas decrease values produce extra deterministic responses.
• response.selections[0].message.content material: Extracts the textual content content material of the mannequin’s response.

6. Question the Anthropic Mannequin

response = consumer.chat.completions.create(mannequin="anthropic:claude-3-5-sonnet-20241022", messages=messages, temperature=0.75)
print(response.selections[0].message.content material)

• mannequin=”anthropic:claude-3-5-sonnet-20241022″: Specifies the Anthropic Claude-3-5 mannequin.
• The remaining parameters are an identical to the OpenAI question. This demonstrates how AISuite allows straightforward switching between suppliers by altering the mannequin parameter.

7. Question the Ollama Mannequin

response = consumer.chat.completions.create(mannequin="ollama:llama3.1:8b", messages=messages, temperature=0.75)
print(response.selections[0].message.content material)

• mannequin=”ollama:llama3.1:8b”: Specifies the Ollama Llama3.1 mannequin.
• Once more, the parameters and logic are constant, showcasing how AISuite gives a unified interface throughout suppliers.

Output

Why did the pirate go to highschool? To enhance his "arrrrrrr-ticulation"!
Arrr, why do not pirates take a bathe earlier than they stroll the plank? As a result of
they will simply wash up on shore later! 🏴‍☠️
Why did the scurvy canine's parrot go to the physician? As a result of it had a fowl
mood, savvy?

Create a Chat Completion

!pip set up openai
!pip set up aisuite[all]
os.environ['OPENAI_API_KEY'] = getpass('Enter your OPENAI API key: ')
from getpass import getpass
import aisuite as ai
consumer = ai.Shopper()
supplier = "openai"
model_id = "gpt-4o"
messages = [
   {"role": "system", "content": "You are a helpful assistant."},
   {"role": "user", "content": "Give me a tabular comparison of RAG and AGENTIC RAG"},
]
response = consumer.chat.completions.create(
   mannequin=f"{supplier}:{model_id}",
   messages=messages,
)
print(response.selections[0].message.content material)

Output

Actually! Under is a tabular comparability of Retrieval-Augmented Era
(RAG) and Agentic RAG.
| Characteristic                | RAG                                             |
Agentic RAG                                        |
|------------------------|-------------------------------------------------|-
---------------------------------------------------|
| Definition             | A framework that mixes retrieval from exterior
paperwork with era. | An extension of RAG that includes actions
based mostly on exterior interactions and dynamic decision-making. |
| Elements             | - Retrieval System (e.g., a search engine or
doc database) 
- Generator (e.g., a language mannequin) | - Retrieval
System 
- Generator 
- Agentic Layer (action-taking and interplay
controller) |
| Performance          | Retrieves related paperwork and generates
responses based mostly on prompted inputs mixed with the retrieved data.
| Provides the aptitude to take actions based mostly on interactions, reminiscent of
interacting with APIs, controlling gadgets, or dynamically gathering extra
data. |
| Use Circumstances              | - Information-based query answering 
-
Content material summarization 
- Open-domain dialogue techniques | - Autonomous
brokers 
- Interactive techniques 
- Determination-making purposes 
-
Methods requiring context-based actions |
| Interplay            | Restricted to the enter retrieval and output
era cycle. | Can work together with exterior techniques or interfaces to
collect information, execute duties, and alter the setting based mostly on goal
features. |
| Complexity             | Usually less complicated because it combines retrieval with
era with out taking actions past producing textual content. | Extra advanced due
to its means to work together with and modify the state of exterior
environments. |
| Instance of Software | Answering advanced questions by retrieving elements of
paperwork and synthesizing them into coherent solutions. | Implementing a
digital assistant able to performing duties like scheduling appointments
by accessing calendars, or a chatbot that manages customer support queries
via actions. |
| Flexibility            | Restricted to the accessible retrieval corpus and
era mannequin capabilities. | Extra versatile attributable to action-oriented
interactions that may adapt to dynamic environments and situations. |
| Determination-Making Means| Restricted decision-making based mostly on static retrieval
and era. | Enhanced decision-making via dynamic interplay and
adaptive habits. |
This comparability outlines the foundational variations and capabilities
between conventional RAG techniques and the extra superior, interaction-capable
Agentic RAG frameworks.

Every mannequin Makes use of a Totally different Supplier

1. Putting in and Importing Libraries

!pip set up aisuite[all]
from pprint import pprint as pp

• Installs the aisuite library with all non-compulsory dependencies.
• Imports a pretty-printing operate (pprint) to format output for higher readability. A customized pprint operate is outlined to permit a customized width.

2. Setting Up API Keys

import os
from getpass import getpass
os.environ['GROQ_API_KEY'] = getpass('Enter your GROQ API key: ')

Prompts the consumer to enter their GROQ API key, which is saved within the setting variable GROQ_API_KEY.

3. Initializing the AI Shopper

import aisuite as ai
consumer = ai.Shopper()

Initializes an AI consumer utilizing the aisuite library to work together with completely different fashions.

4. Chat Completions

messages = [
    {"role": "system", "content": "You are a helpful agent, who answers with brevity."},
    {"role": "user", "content": 'Hi'},
]
response = consumer.chat.completions.create(mannequin="groq:llama-3.2-3b-preview", messages=messages)
print(response.selections[0].message.content material)

Output

How can I help you?

• Defines a chat with two messages:
  • A system message that units the tone or habits of the AI (concise responses).
  • A consumer message as enter.
• Sends the messages to the AI mannequin groq:llama-3.2-3b-preview and prints the mannequin’s response.

5. Operate to Ship Queries

def ask(message, sys_message="You're a useful agent.",
         mannequin="groq:llama-3.2-3b-preview"):
    consumer = ai.Shopper()
    messages = [
        {"role": "system", "content": sys_message},
        {"role": "user", "content": message}
    ]
    response = consumer.chat.completions.create(mannequin=mannequin, messages=messages)
    return response.selections[0].message.content material
ask("Hello. what's capital of Japan?")

Output

'Hiya. The capital of Japan is Tokyo.'

• ask is a reusable operate to ship queries to the mannequin.
• Accepts:
  • message: The consumer’s question.
  • sys_message: Elective system instruction.
  • mannequin: Specifies the AI mannequin.
• Sends the enter and returns the AI’s response.

6. Utilizing A number of APIs

os.environ['OPENAI_API_KEY'] = getpass('Enter your OPENAI API key: ')
os.environ['ANTHROPIC_API_KEY'] = getpass('Enter your ANTHROPIC API key: ')

print(ask("Who's your creator?"))
print(ask('Who's your creator?', mannequin="anthropic:claude-3-5-sonnet-20240620"))
print(ask('Who's your creator?', mannequin="openai:gpt-4o"))

Output

I used to be created by Meta AI, a number one synthetic intelligence analysis
group. My data was developed from a big corpus of textual content, which
I exploit to generate human-like responses to consumer queries.
I used to be created by Anthropic.
I used to be developed by OpenAI, a corporation that focuses on synthetic
intelligence analysis and deployment.

• Prompts the consumer for OpenAI and Anthropic API keys.
• Sends a question (“Who’s your creator?”) to completely different fashions:
  • groq:llama-3.2-3b-preview
  • anthropic:claude-3-5-sonnet-20240620
  • openai:gpt-4o
• Prints the response from every mannequin, exhibiting how completely different techniques interpret the identical question.

7. Querying A number of Fashions

fashions = [
    'llama-3.1-8b-instant',
    'llama-3.2-1b-preview',
    'llama-3.2-3b-preview',
    'llama3-70b-8192',
    'llama3-8b-8192'
]
ret = []
for x in fashions:
    ret.append(ask('Write a brief one sentence clarification of the origins of AI?', mannequin=f'groq:{x}'))

• A listing of various mannequin identifiers (fashions) is outlined.
• Loops via every mannequin and queries it with:
  • Write a brief one sentence clarification of the origins of AI?
• Shops responses within the checklist ret.

8. Displaying Mannequin Responses

for idx, x in enumerate(ret):
    pprint(fashions[idx] + ': n ' + x + ' ')

• Loops via the saved responses.
• Codecs and prints the mannequin’s identify together with its response, making it straightforward to match outputs.

Output

('llama-3.1-8b-instant: n'
 ' The origins of Synthetic Intelligence (AI) date again to the 1956 Dartmouth '
 'Summer time Analysis Challenge on Synthetic Intelligence, the place a gaggle of '
 'laptop scientists, led by John McCarthy, Marvin Minsky, Nathaniel '
 'Rochester, and Claude Shannon, coined the time period and laid the inspiration for '
 'the event of AI as a definite area of research. ')
('llama-3.2-1b-preview: n'
 ' The origins of Synthetic Intelligence (AI) date again to the mid-Twentieth '
 'century, when the primary laptop packages, which mimicked human-like '
 'intelligence via algorithms and rule-based techniques, had been developed by '
 'famend mathematicians and laptop scientists, together with Alan Turing, '
 'Marvin Minsky, and John McCarthy within the Fifties. ')
('llama-3.2-3b-preview: n'
 ' The origins of Synthetic Intelligence (AI) date again to the Fifties, with '
 'the Dartmouth Summer time Analysis Challenge on Synthetic Intelligence, led by '
 'laptop scientists John McCarthy, Marvin Minsky, and Nathaniel Rochester, '
 'marking the delivery of AI as a proper area of analysis. ')
('llama3-70b-8192: n'
 ' The origins of Synthetic Intelligence (AI) might be traced again to the Fifties '
 'when laptop scientist Alan Turing proposed the Turing Check, a technique for '
 'figuring out whether or not a machine might exhibit clever habits equal '
 'to, or indistinguishable from, that of a human. ')
('llama3-8b-8192: n'
 ' The origins of Synthetic Intelligence (AI) might be traced again to the '
 'Fifties, when laptop scientists DARPA funded the event of the primary AI '
 'packages, such because the Logical Theorist, which aimed to simulate human '
 'problem-solving talents and study from expertise. ')

Fashions present diversified responses to the question concerning the origins of AI, reflecting their coaching and reasoning capabilities. As an illustration:

• Some fashions reference the Dartmouth Summer time Analysis Challenge on AI.
• Others point out Alan Turing or early DARPA-funded AI packages.

Key Options and Takeaways

• Modularity: The script makes use of reusable features (ask) to make querying environment friendly and customisable.
• Multi-Mannequin Interplay: Showcases the power to work together with numerous AI techniques, together with GROQ, OpenAI, and Anthropic.
• Comparative Evaluation: Facilitates comparability of responses throughout fashions for insights into their strengths and biases.
• Actual-Time Inputs: Helps dynamic enter for API keys, making certain safe integration.

This script is a wonderful start line for exploring completely different AI mannequin capabilities and understanding their distinctive behaviours.

Conclusion

AISuite is a necessary device for anybody navigating the world of enormous language fashions. It empowers customers to harness the most effective of a number of AI suppliers whereas simplifying growth and fostering innovation. Its open-source nature and considerate design underscore its potential as a contemporary AI software growth cornerstone.

It accelerates growth and enhances flexibility by enabling seamless switching between fashions like OpenAI, Anthropic, and Meta with minimal integration effort. Perfect for each easy and sophisticated purposes, AISuite helps modular workflows, API key administration, and real-time multi-model comparisons. Its ease of use, scalability, and skill to streamline cross-provider interactions make it a useful useful resource for builders, researchers, and educators, empowering environment friendly and modern utilisation of numerous LLMs in an evolving AI panorama.

If you’re in search of generative AI course on-line then discover: GenAI Pinnacle Program

Often Requested Questions

Hello, I’m Pankaj Singh Negi – Senior Content material Editor | Obsessed with storytelling and crafting compelling narratives that rework concepts into impactful content material. I really like studying about know-how revolutionizing our life-style.