LangChain is an advanced framework that allows developers to create language model-powered applications. It provides a set of tools, components, and interfaces that make building LLM-based applications easier. With LangChain, managing interactions with language models, chaining together various components, and integrating resources like APIs and databases is a breeze. The platform includes a set of APIs that can be integrated into applications, allowing developers to add language processing capabilities without having to start from scratch.

1. 1/22
Build an LLM-powered application using LangChain
leewayhertz.com/build-llm-powered-apps-with-langchain
In the ever-evolving AI landscape, language models have taken center stage, redefining how
we interact with machines. With ChatGPT gaining widespread recognition and tech giants
like Google coming up with their own ChatGPT-like solutions, language models, especially
LLMs, have become a major talking point in the tech space. LLMs represent a significant
leap forward in AI’s ability to understand, interpret and generate human language. These
models are trained on vast amounts of text data, enabling them to grasp intricate linguistic
patterns and semantic nuances. With unprecedented language processing capabilities, they
enable users to generate high-quality content with remarkable accuracy and efficiency.
Applications based on LLMs excel in tasks like text generation, sentiment analysis, language
translation and conversational interfaces.
LangChain, a framework built around LLMs, opens up a world full of possibilities in natural
language processing, enabling the creation of various applications, including chatbots and
question-answering tools.
This article provides a comprehensive overview of LangChain, covering its conceptual
basics, use cases, and the best practices to be followed while building LLM-based
applications. Whether you are an experienced developer looking to integrate a language
model into your project or a novice interested in exploring the capabilities of LLMs, this guide
is designed to help you get started.

2. 2/22
Introduction to LangChain and LLM-powered applications
LangChain: Its components and working
Different types of models that are used in LangChain
Setting up a LangChain project: Building LLM-powered applications
LangChain’s applications & use cases
Best practices for building LLM-powered applications with LangChain
Introduction to LangChain and LLM-powered applications
LangChain is an advanced framework that allows developers to create language model-
powered applications. It provides a set of tools, components, and interfaces that make
building LLM-based applications easier. With LangChain, managing interactions with
language models, chaining together various components, and integrating resources like APIs
and databases is a breeze. The platform includes a set of APIs that can be integrated into
applications, allowing developers to add language processing capabilities without having to
start from scratch. Hence, LangChain simplifies and streamlines the process of developing
LLM-powered apps, making it appropriate for developers of all skill levels.
Chatbots, virtual assistants, language translation tools and sentiment analysis tools are all
examples of LLM-powered apps. Developers utilize LangChain to build custom language
model-based apps tailored to specific use cases.
As natural language processing becomes more advanced and widely used, the possible
applications for this technology could become endless.
LangChain: Its components and working
One of the unique aspects of LangChain is its focus on flexibility and modularity. By breaking
down the natural language processing pipeline into individual components, developers can
easily mix and match these building blocks to create custom workflows that meet their
specific needs. This makes LangChain a highly adaptable framework that can be used to
build conversational AI applications for a wide range of use cases and industries.
LangChain’s key features include components and chains, prompt templates and values,
example selectors, output parsers, indexes and retrievers, chat message history, document
loaders, text splitters, agents, and toolkits. All these enable developers to build end-to-end
conversational AI applications that can deliver personalized and engaging user experiences.
Components and chains: In LangChain, a “component” refers to a block of code or a
module that performs a specific function within the natural language processing
pipeline. Components can be combined into “chains” to create custom workflows for
specific use cases. For example, a chain for a customer service chatbot might include
components for sentiment analysis, intent recognition, and response generation.

3. 3/22
Prompt templates and values: Prompt templates are predefined prompts that can be
reused across multiple chains. Values can be inserted into prompt templates to make
them more dynamic and adaptable to specific use cases. For example, a prompt
template might ask the user for their name, and a value could be inserted into the
template to personalize the response. Prompt templates are useful when you need to
generate prompts based on dynamic resources.
from langchain.llms import OpenAI
from langchain import PromptTemplate
llm = OpenAI(model_name="text-davinci-003", openai_api_key="YourAPIKey") # Notice
"location" below, that is a placeholder for another value later
template = """ I really want to travel to {location}. What should I do there? Respond in one
short sentence """
prompt = PromptTemplate(
input_variables=["location"],
template=template,
)
final_prompt = prompt.format(location="Rome")
print(f"Final Prompt: {final_prompt}")
print("-----------")
print(f"LLM Output: {llm(final_prompt)}")
In this example, the final prompt is generated by replacing the {location} placeholder with
“Rome.” The AI then responds with a short sentence: “When in Rome, don’t miss a visit to
the iconic Colosseum.”
Final Prompt:
I really want to travel to Rome. What should I do there?
Respond in one short sentence
-----------
LLM Output:

4. 4/22
Visit the Colosseum, the Pantheon, and the Trevi Fountain for a taste of Rome's ancient and
modern culture.
Example selectors: Example selectors are used to identify relevant examples from the
model’s training data. This helps the model to generate more accurate and relevant
responses. Example selectors can be customized to prioritize specific types of
examples or to exclude irrelevant ones. Example selectors allow you to tailor the AI’s
response based on the user input.
from langchain.prompts.example_selector import SemanticSimilarityExampleSelector
from langchain.vectorstores import FAISS
from langchain.embeddings import OpenAIEmbeddings
from langchain.prompts import FewShotPromptTemplate, PromptTemplate
from langchain.llms import OpenAI
llm = OpenAI(model_name="text-davinci-003", openai_api_key="YourAPIKey")
example_prompt = PromptTemplate(
input_variables=["input", "output"],
template="Example Input: {input}nExample Output: {output}",
)
# Examples of locations that nouns are found
examples = [
{"input": "pirate", "output": "ship"},
{"input": "pilot", "output": "plane"},
{"input": "driver", "output": "car"},
{"input": "tree", "output": "ground"},
{"input": "bird", "output": "nest"},
]
# SemanticSimilarityExampleSelector will select examples that are similar to your input by
semantic meaning

5. 5/22
example_selector = SemanticSimilarityExampleSelector.from_examples(
# This is the list of examples available to select from.
examples,
# This is the embedding class used to produce embeddings which are used to measure
semantic similarity.
OpenAIEmbeddings(openai_api_key=openai_api_key),
# This is the VectorStore class that is used to store the embeddings and do a similarity
search over.
FAISS,
# This is the number of examples to produce.
k=2,
)
similar_prompt = FewShotPromptTemplate(
# The object that will help select examples
example_selector=example_selector,
# Your prompt
example_prompt=example_prompt,
# Customizations that will be added to the top and bottom of your prompt
prefix="Give the location an item is usually found in",
suffix="Input: {noun}nOutput:",
# What inputs your prompt will receive
input_variables=["noun"],
)
For instance, given the noun “student,” the example selector might find “driver” and “pilot” to
be the most similar examples.
# Select a noun!

6. 6/22
my_noun = "student"
print(similar_prompt.format(noun=my_noun))
Output:
Give the location an item is usually found in
Example Input: driver
Example Output: car
Example Input: pilot
Example Output: plane
Input: student
Output:
Output Parsers: Output parsers are used to process and filter the model’s responses.
They can be used to remove unwanted content, format the output in a specific way, or
add additional information to the response. Output parsers enable you to obtain
structured output, such as JSON objects, from the language model’s responses.
from langchain.output_parsers import StructuredOutputParser, ResponseSchema
from langchain.prompts import ChatPromptTemplate, HumanMessagePromptTemplate
from langchain.llms import OpenAI
llm = OpenAI(model_name="text-davinci-003", openai_api_key="YourAPIKey")
# How you would like your reponse structured. This is basically a fancy prompt template
response_schemas = [
ResponseSchema(name="bad_string", description="This a poorly formatted user input
string"),
ResponseSchema(name="good_string", description="This is your response, a reformatted
response"),
]
# How you would like to parse your output
output_parser = StructuredOutputParser.from_response_schemas(response_schemas)

7. 7/22
After defining the response schema, create an output parser to read the schema and parse
it. Next, create the formatting instructions using the get_format_instructions() the method
from the output parser.
# See the prompt template you created for formatting
format_instructions = output_parser.get_format_instructions()
print(format_instructions)
Now, create a prompt template with placeholder variables for the formatting instructions and
user input. This template will also include a section for the AI-generated response.
json
{ "bad_string": string // This a poorly formatted user input string "good_string": string // This is
your response, a reformatted response
}
Once you’ve prepared the prompt template, send it to the language model and obtain the
response. You can then parse the response using the output parser to get a structured JSON
object (or a dictionary in Python).
template = ( """ You will be given a poorly formatted string from a user. Reformat it and make
sure all the words are spelled correctly {format_instructions} % USER INPUT: {user_input}
YOUR RESPONSE: """
)
prompt = PromptTemplate(input_variables=["user_input"], partial_variables=
{"format_instructions": format_instructions}, template=template)
promptValue = prompt.format(user_input="welcom to califonya!")
print(promptValue)
Output:
jsonn{nt"bad_string": "welcom to califonya!",nt"good_string": "Welcome to California!"n}n
Indexes and retrievers: Indexes are databases that store information and metadata
about the model’s training data. Whereas retrievers can quickly search the index for
specific information or examples. This helps the model generate more accurate and
relevant responses by providing them with context and relevant information.
from langchain.document_loaders import TextLoader

8. 8/22
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.vectorstores import FAISS
from langchain.embeddings import OpenAIEmbeddings
loader = TextLoader("data/PaulGrahamEssays/worked.txt")
documents = loader.load()
# Get your splitter ready
text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=50)
# Split your docs into texts
texts = text_splitter.split_documents(documents)
# Get embedding engine ready
embeddings = OpenAIEmbeddings(openai_api_key="YourAPIKey")
# Embedd your texts
db = FAISS.from_documents(texts, embeddings)
# Init your retriever. Asking for just 1 document back
retriever = db.as_retriever()
By initializing the retriever with a document store, you can easily find relevant documents
based on your query. In the example below, we search for documents related to building
things.
VectorStoreRetriever(vectorstore=,
search_type='similarity', search_kwargs={})
The retriever will convert the query into a vector and compare it to the vectors in the
document store. It will then return the most similar documents.
docs = retriever.get_relevant_documents("what types of things did the author want to build?")
print("nn".join([x.page_content[:200] for x in docs[:2]]))
Output:
standards; what was the point? No one else wanted one either, so

9. 9/22
off they went. That was what happened to systems work.I wanted not just to build things, but
to build things that would
last.In this di
much of it in grad school.Computer Science is an uneasy alliance between two halves,
theory
and systems. The theory people prove things, and the systems people
build things. I wanted to build things.
Chat message history: Chat message history refers to the log of messages
exchanged between the user and the model. This can be used to improve the model’s
performance over time by providing it with feedback on its responses and identifying
areas where it can improve.
from langchain.memory import ChatMessageHistory
from langchain.chat_models import ChatOpenAI
chat = ChatOpenAI(temperature=0, openai_api_key="YourAPIKey")
history = ChatMessageHistory()
history.add_ai_message("hi!")
history.add_user_message("what is the capital of france?")
After adding messages to the history, you can pass this history to the language model to
generate context-aware responses:
ai_response = chat(history.messages)
Output:
AIMessage(content='The capital of France is Paris.', additional_kwargs={})
You can then add the AI-generated response to the history:
history.add_ai_message(ai_response.content)
Document loaders: Document loaders enable you to load data from various sources
in a structured format. By using document loaders, you can quickly and easily load data
from various sources and make it available for use in your language models.
from langchain.document_loaders import HNLoader

10. 10/22
loader = HNLoader("https://news.ycombinator.com/item?id=34422627")
data = loader.load()
print(f"Found {len(data)} comments")
print(f"Here's a sample:nn{''.join([x.page_content[:150] for x in data[:2]])}")
By using document loaders, you can quickly and easily load data from various sources and
make it available for use in your language models.
Found 76 comments
Here's a sample:
dang 69 days ago | next [–]
Related ongoing thread:GPT-3.5 and Wolfram Alpha via LangChain -
https://news.ycombinator.com/item?id=344Ozzie_osman 69 days ago | prev | next [–]
LangChain is awesome. For people not sure what it's doing, large language models (LLMs)
are
Text splitters: Text splitters allow you to split a document into smaller chunks. It helps
the model to process the content more effectively. Text splitters are used to split vast
amount of data and divide them into measurable parts.
from langchain.text_splitter import RecursiveCharacterTextSplitter
# This is a long document we can split up.
with open("data/PaulGrahamEssays/worked.txt") as f:
pg_work = f.read()
print(f"You have {len([pg_work])} document")
text_splitter = RecursiveCharacterTextSplitter(
# Set a really small chunk size, just to show.
chunk_size=150,
chunk_overlap=20,
)
texts = text_splitter.create_documents([pg_work])

11. 11/22
print(f"You have {len(texts)} documents")
print("Preview:")
print(texts[0].page_content, "n")
print(texts[1].page_content)
Output:
Preview:
February 2021Before college the two main things I worked on, outside of school,
were writing and programming. I didn't write essays. I wrote what
beginning writers were supposed to write then, and probably still
are: short stories. My stories were awful. They had hardly any plot,
VectorStores: VectorStores are used to store and search the information through
embeddings, they are mainly used to analyze the numerical representations of the
semantic meaning of documents. The VectorStore works as the storage house of these
embeddings to make them easily searchable. It acts as a database for storing the
semantic meaning of your documents, allowing for quick and efficient searching based
on semantic similarity.
from langchain.document_loaders import TextLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.vectorstores import FAISS
from langchain.embeddings import OpenAIEmbeddings
loader = TextLoader("data/PaulGrahamEssays/worked.txt")
documents = loader.load()
# Get your splitter ready
text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=50)
# Split your docs into texts
texts = text_splitter.split_documents(documents)
# Get embedding engine ready

12. 12/22
embeddings = OpenAIEmbeddings(openai_api_key="YourAPIKey")
print(f"You have {len(texts)} documents")
In this example, we create embeddings based on the split documents. The number of
embeddings should match the number of documents created.
You have 78 documents
The VectorStore will store these embeddings and make them easily searchable. It acts as a
database for storing the semantic meaning of your documents, allowing for quick and
efficient searching based on semantic similarity.
embedding_list = embeddings.embed_documents([text.page_content for text in texts])
print(f"You have {len(embedding_list)} embeddings")
print(f"Here's a sample of one: {embedding_list[0][:3]}...")
Output:
You have 78 embeddings
Here's a sample of one: [-0.0011257503647357225, -0.01111479103565216,
-0.012860921211540699]...
Agents: Agents are individual instances of LangChain that are deployed to interact
with users, each agent has unique prompts, memory, and chain tailored to a specific
use case or application. Agents can be deployed on various platforms, including web
apps, mobile apps, and chatbots, making them accessible to a broad audience. Agents
also enable the language model to dynamically decide which tools to use to best
respond to a given query.
from langchain.agents import load_tools
from langchain.agents import initialize_agent
from langchain.llms import OpenAI
import json
llm = OpenAI(temperature=0, openai_api_key="YourAPIKey")
serpapi_api_key = "..."
toolkit = load_tools(["serpapi"], llm=llm, serpapi_api_key=serpapi_api_key)

13. 13/22
agent = initialize_agent(toolkit, llm, agent="zero-shot-react-description", verbose=True,
return_intermediate_steps=True)
response = agent({"input": "what was the first album of the" "band that Natalie Bergman is a
part of?"})
The final response provides a clear and accurate answer to the query.
> Entering new AgentExecutor chain...
I should try to find out what band Natalie Bergman is a part of.
Action: Search
Action Input: "Natalie Bergman band"
Observation: Natalie Bergman is an American singer-songwriter. She is one half of the duo
Wild Belle, along with her brother Elliot Bergman. Her debut solo album, Mercy, was released
on Third Man Records on May 7, 2021. She is based in Los Angeles.
Thought: I should search for the debut album of Wild Belle.
Action: Search
Action Input: "Wild Belle debut album"
Observation: Isles
Thought: I now know the final answer.
Final Answer: Isles is the debut album of Wild Belle, the band that Natalie Bergman is a part
of.
> Finished chain.
Toolkits: LangChain includes toolkits that provide a set of prebuilt components and
chains that can be customized for specific use cases. Toolkits can help developers get
started quickly and provide a solid foundation for building more complex workflows.
Different types of models that are used in LangChain
There are different types of models accessible through the LangChain framework.
LangChain is designed to be modular and flexible, allowing developers to choose the right
model for their specific use case.
Three primary types of models are used in LangChain:

14. 14/22
1. Language models: LLMs are the foundation of many language model applications and
are designed to take in a text string and produce a text string. They are trained on vast
amounts of text data and can generate high-quality responses to natural language input.
from langchain.llms import OpenAI
llm = OpenAI(model_name="text-ada-001", openai_api_key="YourAPIKey")
llm("What day comes after Friday?")
The model would generate the output “Saturday.”
2. Chat models: Chat models, on the other hand, have a more formal API and are designed
to handle conversation history and context. They take in as input a list of chat messages and
there comes another list of chat messages as output, making it easy to manage the
conversational flow and keep track of user interactions.
from langchain.chat_models import ChatOpenAI
from langchain.schema import HumanMessage, SystemMessage, AI Message
chat = ChatOpenAI(temperature=1, openai_api_key="YourAPIKey")
chat([SystemMessage(content="You are an unhelpful AI bot that makes a joke at whatever
the user says"), HumanMessage(content="I would like to go to New York; how should I do
this?")])
In this example, the model responds humorously and unhelpfully to the user’s query about
traveling to New York, as instructed by the system message.
AI Message(content="You could try walking, but I don't recommend it unless you have a lot
of time on your hands. Maybe try flapping your arms really hard and see if you can fly
there?", additional_kwargs={})
3. Text embedding models: They are designed to accept the text as input and produce a
list of items that represent the embeddings of the text. These embeddings can be used for a
wide range of applications, including document retrieval, grouping, and similarity
comparisons.
from langchain.embeddings import OpenAIEmbeddings
embeddings = OpenAIEmbeddings(openai_api_key="YourAPIKey")
text = "Hi! It's time for the beach"
text_embedding = embeddings.embed_query(text)

15. 15/22
print( f"Your embedding is length {len(text_embedding)}")
print( f"Here's a sample: {text_embedding[:5]}...")
The generated embeddings are a one-dimensional array (or a list) of numbers that
semantically represent the text’s meaning.
By selecting the right LangChain model for their use cases, developers can leverage these
powerful tools to build sophisticated conversational AI applications. LangChain’s modular
and flexible design makes it easy to mix and match different components and models,
allowing developers to create custom workflows that meet their specific needs.
Setting up a LangChain project: Building LLM-powered applications
The simplest approach to going further into LangChain is to begin developing actual
applications, which we will accomplish in the this section. Here we have set up the
Langchain project with python.
Let’s begin by making a new project folder:
mkdir langchain-app
cd langchain-app
Next, create a new Python virtual environment:
python3 -m venv env
Here’s a breakdown of the command:
python3: This instructs the command to use Python 3 as its interpreter.
-m venv: This value specifies that the command should utilise the built-in venv module to
create virtual environments.
env: This is the name of the virtual environment you want to create. In this case, the virtual
environment will be named env.
A virtual environment is an isolated python environment that allows you to install project-
specific packages and dependencies without interfering with your system-wide Python
installation or other projects. This isolation aids in maintaining consistency and avoiding
potential conflicts between project requirements.
Once the virtual environment has been created, use the following command to activate it:
source env/bin/activate

16. 16/22
With the virtual environment activated, we can begin installing the project’s dependencies. To
begin, we will use the following command to install LangChain:
pip install langchain
The output on the console should then look like the following:
Let’s continue installing the openai package:
pip install openai
This package is required in order tomake use of OpenAI’s Large Language Models (LLM) in
LangChain.
To be able to access OpenAI models with LangChain, you must first obtain an API key from
OpenAI. Take the following steps:
Go to the OpenAI website: https://www.openai.com/
Click on “Get Started” or “Sign in” if you already have an account.
Create an account or sign in to your existing account.
After signing in, you’ll be directed to the OpenAI Dashboard.
Navigate to the API section by clicking “API” in the left sidebar menu or by visiting:
https://platform.openai.com/signup
Follow the instructions to access or sign up for the API. If you’re eligible, you’ll be provided
with an API key.
The API key should look like a long alphanumeric string (e.g., “sk-
12345abcdeABCDEfghijKLMNOP”).
To set the OpenAI Key for our environment you can use the command line:
export OPENAI_API_KEY="..."
Or you can include the following two lines of Python code in your script:
import os
os.environ["OPENAI_API_KEY"] = "..."
Using LLMs in LangChain

17. 17/22
LangChain provides an LLM class designed for interfacing with various language model
providers, such as OpenAI, Cohere, and Hugging Face. This class provides a common
interface for all LLM kinds. In this section we’ll walk you through integrating LLMs with
LangChain using an OpenAI LLM wrapper and the functionalities highlighted here are
applicable to all LLM types.
Import the LLM wrapper:
To start, import the desired LLM wrapper. In this example, we’ll use the OpenAI wrapper from
LangChain:
from langchain.llms import OpenAI
llm = OpenAI(model_name="text-ada-001", n=2, best_of=2)
Generate text:
The most basic functionality of an LLM is generating text. To do this, simply call the LLM
instance and pass in a string as the prompt:
llm("Tell me a joke")
Let’s combine everything to a complete Python script in file langchain-llm-01.py:
import os
from langchain.llms import OpenAI
os.environ["OPENAI_API_KEY"] = ""
llm = OpenAI(model_name="text-ada-001", n=2, best_of=2)
result = llm("Tell me a joke")
print(result)
Let’s run this script with:
python langchain-llm-01.py
You should then be able to see the output on the console:
Generate more detailed output:
You can also call the LLM instance with a list of inputs, obtaining a more complete response
that includes multiple top responses and provider-specific information:
llm_result = llm.generate(["Tell me a joke", "Tell me a poem"]*15)

18. 18/22
len(llm_result.generations)
The code provided has two main components:
llm_result = llm.generate(["Tell me a joke", "Tell me a poem"]*15)
This line of code calls the generate() method of the llm instance, which is an instance of the
LangChain LLM class. The generate() method takes a list of prompts as input. In this case,
the list consists of two prompts: “Tell me a joke” and “Tell me a poem”. The *15 operation
repeats this list 15 times, resulting in a list with a total of 30 prompts.
When the generate() method is called with this list of prompts, the LLM generates responses
for each prompt. The method returns a result object that contains the generated responses
as well as additional information.
len(llm_result.generations)
This line of code retrieves the number of generations (generated responses) in the llm_result
object. The generations attribute of the llm_result object is a list containing the generated
responses for each input prompt. In this case, since there are 30 input prompts, the length of
the llm_result.generations list will be 30, indicating that the LLM has generated 30 responses
corresponding to the 30 input prompts.
Access the top responses and provider-specific information:
ou can access the generations provided by the LLM by accessing the generations array:
llm_result.generations[0]
You can also retrieve more information about the output by using property llm_output:
llm_result.llm_output
The structure available via llm_output is specific to the LLM you’re using. In case of OpenAI
LLM’s it should contain information about the token usage, e.g.:
{'token_usage': {'completion_tokens': 3903,
'total_tokens': 4023,
'prompt_tokens': 120}}
This is a dictionary representation of the LLM provider-specific information contained in the
llm_output attribute. In this case, it shows information about token usage:
completion_tokens: The number of tokens generated by the LLM as responses for
the input prompts. In this case, the LLM generated 3,903 tokens as completion text.

19. 19/22
total_tokens: The total number of tokens used during the generation process,
including both the input prompts and the generated responses. In this case, the total
token count is 4,023.
prompt_tokens: The number of tokens in the input prompts. In this case, the prompts
contain 120 tokens in total.
These token counts can be useful when working with LLMs, as they help you understand the
token usage and costs associated with generating text using the LLM.
Let’s combine it all into a new Python script:
import os
from langchain.llms import OpenAI
os.environ["OPENAI_API_KEY"] = "sk-
uagF4hvjOZfS7icdEm2nT3BlbkFJHS4fShHjYTf36XG3ykor"
llm = OpenAI(model_name="text-ada-001", n=2, best_of=2)
llm_result = llm.generate(["Tell me a joke", "Tell me a poem"]*15)
print("Number of Genrations provided: ")
print(len(llm_result.generations))
print("n")
print("Generation Index 0: ")
print(llm_result.generations[0])
print("n")
print("Addtional information:")
print(llm_result.llm_output)
Starting the script again will lead to the following results:
Estimate token count:
Estimating the number of tokens in a piece of text is useful because models have a context
length (and cost more for more tokens), so it’s important to be aware of the text length you’re
passing in. By default, the tokens are estimated using tiktoken.
This means you need to install the corresponding package as well:

20. 20/22
pip install tiktoken
You can then use the following line of code to get top token estimate:
llm.get_num_tokens("what a joke")
LangChain applications and use cases
LangChain allows developers to access a vast corpus of text data, use pre-trained models
for language processing, and build robust applications that can be customized as per user’s
interests. Here are some use cases and applications of LangChain:
Language translation: LangChain allows connecting a language model to other sources of
data. In translation, this means that It enhances translation capabilities by integrating
additional data sources and allowing customization for more powerful and differentiated
translation applications.
Chatbots and virtual assistants: LangChain can be used to build chatbots and virtual
assistants that can understand and respond to natural language queries. Developers can
use LangChain to train models to understand different languages and use them to power
chatbots and virtual assistants.
Sentiment analysis: LangChain can be used to build sentiment analysis applications that
can analyze large volumes of text data and identify positive, negative, and neutral sentiment.
This can be used for market research, social media monitoring, and other use cases.
Language learning: LangChain can be used to build language learning applications that
can help people learn new languages. Developers can use pre-trained models to build
applications that adapt to users’ comprehension level and provide them with grammar and
vocabulary exercises accordingly.
Content moderation: LangChain can be used to build content moderation applications that
can automatically identify and flag inappropriate or offensive content. This can be used on
social media platforms, online forums, and other websites to protect users from harmful
content.
Best practices for building LLM-powered applications with
LangChain
While building LLM-powered applications with LangChain, there are several best practices
that developers should follow to ensure that their applications are robust, scalable, and
secure. Here are some of the best practices for building LLM-powered applications with
LangChain:

21. 21/22
Choose the right model: The success of an LLM-powered application depends on the
accuracy and performance of the machine learning model that powers it. Developers should
choose the right model for their application based on the type of data they are working with,
the size of their dataset, and the specific task they are trying to accomplish.
Preprocess your data: Machine learning models require clean, structured, and labeled data
to function accurately. Developers should preprocess their data before training their models
to ensure that it is formatted correctly, contains no errors, and is labeled accurately.
Fine-tune your model: Pre-trained models can be used as a starting point for LLM-powered
applications, but they may not be optimal for specific use cases. Developers should fine-tune
their models to improve their accuracy and performance on specific tasks and datasets.
Storage systems: Storage systems are used to store data and models for LLM-powered
applications. This ensures that data is accessible to all users of the application and that the
application is not reliant on a centralized server.
Ensure security: Security is crucial for any application that processes user data. Developers
should ensure that their LLM-powered applications are secure by encrypting sensitive data,
implementing access controls, and using secure communication protocols.
Test your application: Developers should thoroughly test their LLM-powered applications
before deploying them to ensure that they function as intended and are free of bugs. Testing
should include unit tests, integration tests, and user acceptance testing.
Conclusion
LangChain provides developers with a powerful toolkit to build language-based applications.
It serves as the structural foundation that empowers developers to create language-based
applications to understand and process human language with high accuracy and efficiency.
By leveraging the power of machine learning algorithms, LangChain provides a secure and
transparent platform for building innovative applications that can reinvent the way we
communicate and interact with technology.
This beginner’s guide provides an overview of LangChain, its core concept, use cases and
how to get started with building LLM-powered applications with Python. By following the best
practices outlined in this guide, developers can ensure that their applications are robust,
scalable, secure and ready to provide a seamless user experience. As the field of natural
language processing continues to evolve, LangChain is poised to play a significant role in the
development of language-based applications in a wide range of industries, from healthcare
and finance to social media and education. By empowering developers to build applications
that can understand and process human language, LangChain is helping to bridge the gap
between humans and technology, making it easier for people to communicate and interact
with machines.

22. 22/22
Ready to build your own LLM-powered application with LangChain? Contact LeewayHertz
for your consultancy and development needs.