docs/tutorials/llm_chain/

[giscus[bot]]

docs/tutorials/llm_chain/

In this quickstart we'll show you how to build a simple LLM application with LangChain. This application will translate text from English into another language. This is a relatively simple LLM application - it's just a single LLM call plus some prompting. Still, this is a great way to get started with LangChain - a lot of features can be built with just some prompting and an LLM call!

https://python.langchain.com/docs/tutorials/llm_chain/

[whao]

Every clear and straight forward tutorial! Thanks for your brilliant work!

[gzozulin]

Great tutorial!

[deepika-3]

Hi
very detailed explanation. Thank you

[SosthenesMS]

Congratulations, simple, objective and functional tutorial! Thank you

[charileyou823]

cool man

[iownthegame]

awesome!

[iownthegame]

at some point I got the output as Sei addestrato su dati fino a ottobre 2023. instead of Ciao :think

[Bayzon88]

i had the same issue, by any chance were you using gpt-4-mini? i changed my model to gpt-4 and started working properly.

[masamallow]

I got the same message with gpt-4o-mini.

Here is the issue that someone have raised on the OpenAI community site.
Language Translation Is Broken - API / Bugs - OpenAI Developer Forum

When using gpt-4o-mini, I suspect that a prompt for a too simple translation is recognized as an invalid prompt and this is happening.

[shark22k]

cool

[ksuchoi216]

awesome!!!

[SuperMissKuang]

short and sweet!

[parjun585]

Easy to understand, short and crisp

[liaoxianfu]

In version 0.3, model.invoke(prompt) returns a str, which does not have a content attribute.

[syntaxtrash]

The content attribute is still present. Maybe you included StrOutputParser()?

[Bhargav2525]

I have configured aws in my system. How to configure aws credentials in jupyter notebook

[Bhargav2525]

got it

[nizomjon9909]

Hey , could you please your codes as I am keep getting errors

[Bhargav2525]

AWS_ACCESS_KEY_ID=
AWS_SECRET_ACCESS_KEY=
AWS_SESSION_TOKEN=
give these values in .env file

load env to your notebook file
from dotenv import load_dotenv
load_dotenv()

[sujaydutta]

Great tutorial!

FYI I did have issues with the API key limits for OpenAI and Anthropic, but was successful with Groq.

[xiaomanwong]

I have a question about lcel,
in v0.2 ,the document use lcel to implement, but v0.3 not.

so, what about the lcel position in langchain?

[syntaxtrash]

In the v0.2, they included an output parser to demo LCEL, but in the v0.3, they removed it and just called the LLM directly with the messages.

Here's a document to help you decide whether to use LCEL: : https://python.langchain.com/docs/concepts/lcel/#should-i-use-lcel

[yusuphwasiu]

What a simple and great way to get started. A scratch in the surface indeed!

[nag9s]

I am seeing in older versions, tutorials use export LANGCHAIN_API_KEY="..." but in the newer documents it is replaced with export LANGSMITH_API_KEY="...", what should be used

[syntaxtrash]

use the new one.

LANGSMITH_TRACING=true
LANGSMITH_ENDPOINT=https://api.smith.langchain.com
LANGSMITH_API_KEY=...
LANGSMITH_PROJECT=...

[whillas-yabble]

No mention of LCEL? Isn't this the preferred way to chains?

[pnaguno]

for free API Key use Gemini Flash. I used the sample code with Gemini Free API key.

[Prasannajnaeyulu]

Did you create google cloud project and service account and this piece of code to connect with google vertext AI??

from google.oauth2 import service_account
from vertexai import init

credentials = service_account.Credentials.from_service_account_file(
"/path/to/your-service-account.json"
)

init(project="your-gcp-project-id", location="us-central1", credentials=credentials)

from langchain.chat_models import init_chat_model
model = init_chat_model("gemini-2.0-flash-001", model_provider="google_vertexai")

response = model.invoke([{"role": "user", "content": "Hello!"}])
print(response)

I couldn't get this working any inputs?

[ghermeztorob]

Is it a tutorial or an advertisement of LangSmith?!
If we want LangSmith, we can read about it. You don't need to indicate it in each sentence! Let it go; we don't care.

[Mo7amedDev]

I noticed an inconsistency in how ChatPromptTemplate handles variable interpolation depending on how the messages are defined.

These two approaches work as expected — variables are correctly interpolated:

✅ Using from_messages:

from langchain_core.prompts import ChatPromptTemplate

prompt_template = ChatPromptTemplate.from_messages([
("system", "Translate the following from English into {language}"),
("user", "{text}")
])

prompt = prompt_template.invoke({
'language': 'french',
'text': 'welcome to London'
})

✅ Using message dictionaries:

from langchain_core.prompts import ChatPromptTemplate

prompt_template = ChatPromptTemplate([
{"role": "system", "content": "Translate the following from English into {language}"},
{"role": "user", "content": "{user_input}"}
])

prompt = prompt_template.invoke({
'language': 'french',
'user_input': 'welcome to London'
})

However, this version using SystemMessage and HumanMessage doesn't interpolate the variables:

🚫 Using SystemMessage and HumanMessage directly:

from langchain_core.prompts import ChatPromptTemplate
from langchain_core.messages import SystemMessage, HumanMessage

prompt_template = ChatPromptTemplate([
SystemMessage("Translate the following from English into {language}"),
HumanMessage("{user_input}")
])

prompt = prompt_template.invoke({
'language': 'french',
'user_input': 'welcome to London'
})

In this case, {language} and {user_input} remain as-is in the final output.

It seems that ChatPromptTemplate only interpolates variables when using strings or dicts, but not when passing message objects. This could lead to confusion, especially for developers expecting uniform behavior across all formats.

Would be great if this could be unified, or at least mentioned in the docs for clarity 🙏