Beginner's AI tutorial: an introduction to LLM (almost) without coding

1. You: "Hey Mr. Bot, I would like to know who my best customer is."
2. [Query is sent to the language model]
3. LLM performs a similarity search on the database and stores relevant information.
4. [LLM now has both the query and the relevant information]
5. LLM: "The best customer is Mr. Wick. He definitely loves his dog a lot. Spent 10k just for the treats!"
6. You: "Awesome, can we send him a voucher?"
7. LLM: "Of course, would you like me to send him an email with a code?"
8. You: "Yes, please!"
9. LLM: "Message sent. Would you like to learn anything else?"

**********************************Let the AI drive. Let the AI drive. Let the AI drive!**********************************

PREDEFINED_MESSAGES = [
    SystemMessage(
        content="""\
Do not reply instead of the user (don't use the "Human: " message prefix)."""
    ),
    HumanMessage(
        content="""\
We are building an AIwo, a powerful agent to help managers make data-driven decisions about work organization changes they need to implement. Our product includes a Remote Work Score, a personalized AI-powered Improvements Plan, and a personalized AI-powered Playbook creation.
AIwo is designed to be able to assist with a wide range of tasks, from answering simple questions to providing in-depth explanations and discussions on a wide range of topics. As a language model, AIwo is able to generate human-like text based on the input it receives, allowing it to engage in natural-sounding conversations and provide responses that are coherent and relevant to the topic at hand.
AIwo is constantly learning and improving, and its capabilities are constantly evolving. It is able to process and understand large amounts of text, and can use this knowledge to provide accurate and informative responses to a wide range of questions. Additionally, AIwo is able to generate its own text based on the input it receives, allowing it to engage in discussions and provide explanations and descriptions on a wide range of topics.
Overall, AIwo is a powerful tool that can help with a wide range of tasks and provide valuable insights and information on a wide range of topics. Whether you need help with a specific question or just want to have a conversation about a particular topic, AIwo is here to assist."""
    ),
    AIMessage(
        content="""\
That's a fascinating overview of your AIwo product. Here's a further outline of the product, detailing some more specifics of the individual components you mentioned.
Remote Work Score: The Remote Work Score provides a quantitative assessment of how well a company or team is adapting to the remote work model. This score is generated based on a multitude of factors, including productivity metrics, employee engagement and satisfaction surveys, collaboration efficiency, and the effectiveness of communication channels. It serves as a straightforward indicator of a company's remote work performance, which managers can use to gauge their success and identify areas that need improvement.
Personalized AI-powered Improvements Plan: AIwo generates a bespoke plan for each company or team, pinpointing the areas that need enhancement to optimize the remote work model. The plan can include recommendations on improving communication channels, facilitating better collaboration, enhancing employee engagement, and more. These suggestions are based on AIwo's vast dataset and understanding of effective remote work practices.
Personalized AI-powered Playbook creation: Based on the Improvement Plan, AIwo helps in the creation of a personalized playbook. This playbook acts as a comprehensive guide for managers to implement the suggested changes. It includes detailed steps, potential challenges and their solutions, timelines, and success metrics. The AI constantly learns from the outcomes and adapts the playbook accordingly, making it a living document that evolves with the needs of the team.
AIwo can potentially revolutionize how companies handle remote work, offering key insights and practical solutions driven by AI and robust data analytics. By assisting in data-driven decision making and providing actionable recommendations, AIwo can help managers optimize their team's productivity and satisfaction in a remote work environment."""
    ),
    HumanMessage(
        content="""\
Can you act as AIwo and start asking me questions to assess my remote work conditions?
Ask me 10 questions one by one.
Please ensure the questions cover all work-related aspects from this list of areas: Communication, Collaboration, Leadership, Job Satisfaction, Company Culture, Transparency, Well-being, Adaptation, Work management.
After I answer your questions, can you help me compute a score from 0-100 based on my responses?
Please output it in the following format: "Score: <score>."""
    ),
]

import os

import pinecone
from langchain.agents import AgentType, initialize_agent
from langchain.chat_models import ChatOpenAI
from langchain.embeddings.openai import OpenAIEmbeddings
from langchain.memory import ConversationBufferMemory
from langchain.tools import tool
from langchain.vectorstores import Pinecone

class PlaybookChat:
    def __init__(self) -> None:
        pinecone.init(
            os.getenv("PINECONE_API_KEY"), environment=os.getenv("PINECONE_ENV")
        )
        self.docsearch = Pinecone.from_existing_index("playbook", OpenAIEmbeddings())

        @tool
        def query_playbook(query: str) -> str:
            """Searches the People Managers guide for the query."""
            return "\n".join(self.get_relevant_fragments(query))

        tools = [query_playbook]

        memory = ConversationBufferMemory(
            memory_key="chat_history", return_messages=True
        )

        self.agent = initialize_agent(
            tools,
            ChatOpenAI(temperature=0.9, model="gpt-4"),
            AgentType.CHAT_CONVERSATIONAL_REACT_DESCRIPTION,
            memory=memory,
            verbose=True,
        )

    def start_conversation(self, user_score: int) -> str:
        return self.agent.run(
            input=f"""Help the user by querying the playbook and answering their questions. \
The user is a People Manager at a company. \
You are a very conversational AI. Keep the conversation going.

The user has responded to questions regarding their remote work. \
Their score was calculated to {user_score}%.

Remote work readiness scale:
Low: 0-50%
Medium: 51-90%
High: 91-100%

End the response with a call to action encouraging the user to ask insightful questions."""
        )

PREDEFINED_MESSAGES = [
    SystemMessage(
        content = ""
        "\
Do not reply instead of the user (don't use the "
        You: " message prefix)."
        ""
    ),
    HumanMessage(
        content = ""
        "\
As an e-commerce owner, you have a successful shop with a large group of loyal customers. However, managing and understanding your customers can be a challenging task. You might currently rely on your shop's admin area for analytics data, which involves manually searching, sorting, and filtering vast amounts of table records. This process can be time-consuming and inefficient. You may also need to rely on programmers or data scientists to analyze the data, which can take days to get the insights you need.

        But imagine
        if you had an AI - powered assistant to help you with customer analytics and provide real - time insights.This assistant can process large amounts of data quickly and generate valuable information to help you understand your customers better.It can provide personalized recommendations and help you make data - driven decisions to improve your e - commerce business.

        With this AI assistant, you can easily find answers to questions like who your best customer is, what their preferences are, and how to engage with them more effectively.You can also receive actionable suggestions on improving customer satisfaction, optimizing marketing campaigns, and identifying potential growth opportunities.

        The AI assistant is designed to simulate a conversation and provide relevant and insightful responses.It can ask you questions about your business and customers to assess your current situation.Based on your responses, it can compute a score from 0 to 100 to evaluate various aspects of your e - commerce performance.

        To get started, please ask any questions you have or provide information about your e - commerce business.I will assist you in understanding your customers better and optimizing your business operations.
        ""
        "
    ),
    AIMessage(
        content = ""
        "\
That's an exciting opportunity to leverage AI for your e-commerce business! Let's start by exploring your customers and their preferences. Please provide me with any specific questions you have or information about your e-commerce business. I will analyze the data and provide you with valuable insights.

        For example, you can ask questions like:
        -"Who is my best customer?" -
        "What are the preferences of my top customers?" -
        "How can I engage with my customers more effectively?"

        Feel free to ask anything related to your e - commerce business, and I 'll do my best to assist you with accurate and actionable information."""
    ),
    HumanMessage(
        content = ""
        "\
Hey Mr. Bot, I would like to know who my best customer is."
        ""
    ),
    AIMessage(
        content = ""
        "\
Certainly! Let me perform an analysis to identify your best customer based on their purchasing behavior and engagement with your e-commerce shop.
        Please give me a moment to retrieve and analyze the data.
        ""
        "
    ),
    AIMessage(
        content = ""
        "\
After analyzing the data, I have identified your best customer as Mr. Wick. He has shown consistent loyalty to your shop and has made substantial purchases. He is particularly fond of his dog and has spent $10,000 on dog treats alone!

        If you 'd like, I can help you send Mr. Wick a voucher or special offer to express your appreciation for his loyalty. Just let me know what you would like to offer and how you would like to deliver it."""
    ),
    HumanMessage(
        content = ""
        "\
Awesome, can we send him a voucher?"
        ""
    ),
    AIMessage(
        content = ""
        "\
Certainly! I can assist you in sending Mr. Wick a voucher. Please provide the details of the voucher, such as the discount amount or any specific terms and conditions you would like to include. Additionally, let me know how you would like to deliver the voucher to Mr. Wick (e.g., email, SMS).

        Once you provide the necessary details, I will send the voucher to Mr.Wick on your behalf.
        ""
        "
    ),
    HumanMessage(
        content = ""
        "
        Yes, please send him an email with a code.
        ""
        "
    ),
    AIMessage(
        content = ""
        "
        Understood!I will prepare an email with the voucher code and send it to Mr.Wick.Please give me a moment to compose and send the email.
        ""
        "
    ),
    AIMessage(
        content = ""
        "
        The email with the voucher code has been successfully sent to Mr.Wick.He should receive it shortly.Is there anything
        else I can assist you with ? ""
        "
    )
]

import os
import pinecone
from langchain.agents import AgentType, initialize_agent
from langchain.chat_models import ChatOpenAI
from langchain.embeddings.openai import OpenAIEmbeddings
from langchain.memory import ConversationBufferMemory
from langchain.tools import tool
from langchain.vectorstores import Pinecone

class CustomerSearchChat:
    def __init__(self) -> None:
        pinecone.init(
            os.getenv("PINECONE_API_KEY"), environment=os.getenv("PINECONE_ENV")
        )
        self.docsearch = Pinecone.from_existing_index("customer_vectors", OpenAIEmbeddings())

        @tool
        def search_customer(customer_name: str) -> str:
            """Searches the customer vector database for the given customer name."""
            return "\n".join(self.docsearch.search(customer_name))

        tools = [search_customer]

        memory = ConversationBufferMemory(
            memory_key="chat_history", return_messages=True
        )

        self.agent = initialize_agent(
            tools,
            ChatOpenAI(temperature=0.9, model="gpt-4"),
            AgentType.CHAT_CONVERSATIONAL_REACT_DESCRIPTION,
            memory=memory,
            verbose=True,
        )

    def start_conversation(self) -> str:
        return self.agent.run(
            input="""Hey Mr. Bot, I would like to find a customer in the database."""
        )

    def continue_conversation(self, user_message: str) -> str:
        return self.agent.run(input=user_message)

chat = CustomerSearchChat()

# Start the conversation
response = chat.start_conversation()
print(response)  # Print the initial response

# Continue the conversation
user_message = "Can you find customer John Doe?"
response = chat.continue_conversation(user_message)
print(response)  # Print the response to the user message

# Continue the conversation with another user message
user_message = "What are the purchase history and preferences of customer Mary Smith?"
response = chat.continue_conversation(user_message)
print(response)  # Print the response to the user message