Building a Responsive Chatbot with Llama 3.1, Ollama and LangChain

Introduction

In the fast-paced world of AI, crafting a smart, multilingual chatbot is now within reach. Picture a tool that understands and chats in various languages, helps with coding, and generates high-quality data effortlessly. Enter Meta’s Llama 3.1, a powerful language model that’s transforming AI and making it accessible to everyone. By combining Llama 3.1, Ollama, and LangChain, along with the user-friendly Streamlit, we’re set to create an intelligent and responsive chatbot that makes complex tasks feel simple.

Learning Outcomes

• Understand the key features and advancements of Meta’s Llama 3.1.
• Learn how to integrate Llama 3.1 with Ollama and LangChain.
• Gain hands-on experience in building a chatbot using Streamlit.
• Explore the benefits of open-source AI models in real-world applications.
• Develop skills to fine-tune and optimize AI models for various tasks.

This article was published as a part of the Data Science Blogathon.

What is Meta’s Llama 3.1?

Llama 3.1 represents the most recent update to Meta’s series of language models under the Llama line. In its version dated July 23, 2024, it comes with 8 billion, 70 billion, and—drum roll—a massive 405 billion parameters. These have been trained on a corpus of over 15 trillion tokens in this version, bigger than all the preceding versions put together; hence, improved performance and capabilities.

Open-Source Commitment

Meta maintains their commitment to open-source AI by making Llama 3.1 freely available to the community. This technique promotes innovation by allowing developers to create and improve models for a variety of applications. Llama 3.1’s open-source nature provides access to powerful AI, allowing more individuals to harness its capabilities without incurring large fees.

Ecosystem and Partnerships

In the Llama ecosystem are over 25 partners, including AWS, NVIDIA, Databricks, Groq, Dell, Azure, Google Cloud, Snowflake, and many more, who make their services available right on day one. Such collaborations enhance the accessibility and utility of llama3.1, easing integration into a number of platforms and workflows.

Security and Safety

Meta has introduced a number of new safety and security tools, including Llama Guard 3 and Prompt Guard, to make sure that it builds AI ethically. These ensure that Llama 3.1 is safe to be run, sans possible dangers accruing from the roll-out of Gen-AI.

Instruction Tuning and Fine-Tuning

• Instruction Tuning: Llama 3.1 has undergone extensive tuning on the instructions; it achieves an MMLU knowledge assessment score of 86.1, so it will be quite good at comprehending and following through with complicated instructions typical in advanced uses of AI.
• Fine-Tuning: The fine-tuning process involves several rounds of supervised fine-tuning, rejection sampling, and direct preference optimization. This iterative process ensures that Llama 3.1 generates high-quality synthetic data, improving its performance across different- different tasks.

Key Enhancements in Llama 3.1

• Expanded Parameters: Llama 3.1’s 405B model features 405 billion parameters, making it the most powerful open-source model available. This enhancement facilitates advanced tasks like multilingual translation, synthetic data generation, and complex coding assistance.
• Multilingual Support: The new models support multiple languages, broadening their applicability across diverse linguistic contexts. This makes Llama 3.1 suitable for global applications, offering robust performance in various languages.
• Extended Context Length: One of the main updates in this version is that this length increases to a maximum context length of 128K. That means the model can process longer inputs and outputs, making it suitable for any application that requires full-text understanding and generation.

Performance Metrics

Meta-evaluated Llama over over 150 benchmark datasets and across multiple languages, the results of which show this model to stand in good stead with the best in the field, which currently consists of GPT-4 and Claude 3.5 Sonnet, in various tasks, meaning Llama 3.1 stands right at the top tier in the firmament of AI.

Applications and Use Cases

• Synthetic Data Generation: Llama 3.1’s advanced capabilities make it suitable for generating synthetic data, aiding in the improvement and training of smaller models. This is particularly beneficial for developing new AI applications and enhancing existing ones.
• Coding Assistance: The model’s high performance in code generation tasks makes it a valuable tool for developers seeking AI-assisted coding solutions. Llama 3.1 can help write, debug, and optimize code, streamlining the development process.
• Multilingual Conversational Agents: With robust multilingual support, Llama 3.1 can power complex conversational agents capable of understanding and responding in multiple languages. This is ideal for global customer service applications.

Setting Up Your Environment

Let us now set up the environment.

Creating a Virtual Environment

 python -m venv env

Installing Dependencies

Install dependencies from requirements.txt file.

langchain
langchain-ollama
streamlit
langchain_experimental

pip install -r requirements.txt

Install Ollama

Click here to download Ollama.

Pull the Llama3.1 model

ollama pull llama3.1

You can use it Locally using cmd.

ollama run llama3.1

Running the Streamlit App

We’ll now walk through run a Streamlit app that leverages the powerful Llama 3.1 model for interactive Q&A. This app transforms user questions into thoughtful responses using the latest in natural language processing technology. With a clean interface and straightforward functionality, you can quickly see how to integrate and deploy a chatbot application.

Import Libraries and Initialize Streamlit

We set up the environment for our Streamlit app by importing the necessary libraries and initializing the app’s title.

from langchain_core.prompts import ChatPromptTemplate
from langchain_ollama.llms import OllamaLLM
import streamlit as st
st.title("LLama 3.1 ChatBot")

Style the Streamlit App

We customize the appearance of the Streamlit app to match our desired aesthetic by applying custom CSS styling.

# Styling
st.markdown("""
<style>
.main {
    background-color: #00000;
}
</style>
""", unsafe_allow_html=True)

Create the Sidebar

Now we will add a sidebar to provide additional information about the app and its functionalities.

# Sidebar for additional options or information
with st.sidebar:
    st.info("This app uses the Llama 3.1 model to answer your questions.")

Define the Chatbot Prompt Template and Model

Define the structure of the chatbot’s responses and initialize the language model that will generate the answers.

template = """Question: {question}
Answer: Let's think step by step."""
prompt = ChatPromptTemplate.from_template(template)
model = OllamaLLM(model="llama3.1")
chain = prompt | model

Create the Main Content Area

This section sets up the main interface of the app where users can input their questions and interact with the chatbot.

# Main content
col1, col2 = st.columns(2)
with col1:
    question = st.text_input("Enter your question here")

Process the User Input and Display the Answer

Now handling the user’s input, process it with the chatbot model, and display the generated answer or appropriate messages based on the input.

if question:
    with st.spinner('Thinking...'):
        answer = chain.invoke({"question": question})
        st.success("Done!")
    st.markdown(f"**Answer:** {answer}")
else:
    st.warning("Please enter a question to get an answer.")

Run the App

streamlit run app.py

or

python -m streamlit run app.py

Conclusion

Meta’s Llama 3.1 stands out as a groundbreaking model in the field of artificial intelligence. Its combination of scale, performance, and accessibility makes it a versatile tool for a wide range of applications. By maintaining an open-source approach, Meta not only promotes transparency and innovation but also empowers developers and organizations to harness the full potential of advanced AI. As the Llama 3.1 ecosystem continues to evolve, it is poised to drive significant advancements in how AI is applied across industries and disciplines. In this article we learned how we can build our own chatbot with Llama 3.1, Ollama and LangChain.

Key Takeaways

• Llama 3.1 packs up to 405 billion parameters, raising the computational muscle.
• Supports languages in many applications. Extended Context Length: Now supporting up to 128K tokens for full-text processing.
• Beating baselines, especially for reasoning, translation, and tool use.
• Very proficient in following through complex instructions.
• Openly accessible, free, and extendable for community innovation.
• Suitable for AI agents, Translation, Coding Assistance, Content Creation.
• Backed by major tech partnerships for seamless integration.
• Packs tools such as Llama Guard 3 and Prompt Guard for safe deployment.

Frequently Asked Questions

The media shown in this article is not owned by Analytics Vidhya and is used at the Author’s discretion.

Gourav Lohar

Hi I'm Gourav, a Data Science Enthusiast with a medium foundation in statistical analysis, machine learning, and data visualization. My journey into the world of data began with a curiosity to unravel insights from datasets.