Basic Usage LangChain

3 Basic Things to know

Before we jump into the deep of the LangChain Framework, there are 3 basic things to know:

  • Prompt Template
  • Chat Models
  • Chains

Prompt Template

Prompt Templates are predefined structures used to guide the responses of language models.

Why Prompt Templates?

  • Consistency

    By using a set structure, prompt templates ensure consistent responses from the language model.

  • Relevance

    Templates can be designed to elicit specific types of responses, ensuring the output is relevant to the user’s needs.

Example Prompt Template

Example 1

Prompt Template for a Movie Recommendation

"Based on your preference for {genre} movies and your interest in {actor}, I would recommend 5 Films:"

In this template: * {genre} would be replaced by the user’s preferred movie genre (e.g., action, comedy, drama). * {actor} would be replaced by the user’s favorite actor.

The final prompt would look something like this:

"Based on your preference for comedy movies and your interest in Will Ferrell, I would recommend 5 Film:"
Example 2

Prompt Template for a Restaurant Recommendation

"Considering your love for {cuisine_type} and your location in {city}, I suggest you try the following restaurant:"

In this template: * {cuisine_type} would be replaced by the user’s preferred cuisine (e.g., Italian, Chinese, Mexican). * {city} would be replaced by the user’s current city.

The final prompt would look something like this:

"Considering your love for ramen and your location in Bandung, I suggest you try the following restaurant:"

Creating a Prompt Template

Prompt templates are created using the PromptTemplate from the langchain library.

from langchain import PromptTemplate

template_movie_recomendation = "Based on your preference for {genre} movies and your interest in {actor}, I would recommend {movie_title}."

prompt_template_movie_recomendation = PromptTemplate(input_variables=["genre", "actor", "movie_title"], template=template_movie_recomendation)

Chat Models

Chat Models are specialized language models designed for conversational interactions.

Why Chat Models?

  • Natural Interaction

Chat models are trained to respond in a conversational manner, providing a more engaging and natural user experience.

  • Context Awareness

Unlike traditional models, chat models can maintain context over a series of exchanges, allowing for more coherent and meaningful conversations.

Note: In this course, we will be using chat models from OpenAI’s

Declaring a Chat Model

Chat model is declared using the ChatOpenAI from the langchain.chat_models.

from langchain.chat_models import ChatOpenAI

chat = ChatOpenAI(openai_api_key="...", model_name="gpt-3.5-turbo", temperature=0)

Chains

Chains in LangChain allow for sequential interactions with the language model, enabling more complex and contextual conversations.

Why Chains?

  • Multiple Steps

    Chains allow for multi-step interactions, where each step can influence the next, creating a dynamic conversation flow.

  • Context Maintenance

    Chains keep track of context over multiple exchanges, ensuring continuity in conversations.

Image

Declaring & Running a Chain

Chains are declared using the LLMChain from the langchain.chains library.

from langchain.chains import LLMChain

chain = LLMChain(llm=llm, prompt=prompt_template_movie_recomendation)

print(chain.run(genre="action", actor="Tom Cruise"))

We already created a basic and simple application using the LangChain Framework. Let’s dive more with the LangChain Framework in the next section.

Back to top