docs/tutorials/extraction/

[giscus[bot]]

docs/tutorials/extraction/

This guide assumes familiarity with the following concepts:

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

[sanjeed5]

This was helpful! Good tutorial!

[OkYongChoi]

Gracias! Super useful!!

[dawngo]

This tutorial is very, very good！

[hakim89]

Any chance this tutorial can be ran using Ollama running mistal locally?
if I do

from langchain_ollama import OllamaLLM
llm = OllamaLLM(model="mistral")
structured_llm = llm.with_structured_output(schema=Person)

I get an error: NotImplementedError

appreciate any advice.

[efriis]

Hey there! You'll have to use ChatOllama instead of OllamaLLM to use structured output

[shuze295]

The error still occurs when using this ChatOllama.

[lcombita09]

how does one can combine this with a retriever, to perform data extraction for large documents? It is possible to pass to the retriever also the schema in order to extract the relevant documents?

[efriis]

Not sure exactly what you mean so making some assumptions, but it sounds like you want to do some kind of map-reduce on the retrieved documents, where you're determining if each document is relevant using an llm. would recommend using langgraph map reduce branches for this if that's the case! https://langchain-ai.github.io/langgraph/how-tos/map-reduce/

[lcombita09]

hey! thanks for the response. What I mean, is that here on the examples the extraction is done from a text, but what if I want to do it on a 100 pages pdf, there must be a way to pass to the llm just the relevant chunks to extract the information needed, that is what I meant

[richlysakowski]

To use this in Colab is a PAIN IN THE ARSE (PITA) if you don't want to expose your API Keys for LangSmith and OpenAI.

Google has its own proprietary Linux-based Secrets manager instead of the standard Jupyter NB conda environment.

You must replace the Jupyter NB code with the following code:

Get API Keys from Local User Environment

from google.colab import userdata

Get the API keys from User Secrets

OPENAI_API_KEY = userdata.get('OPENAI_API_KEY')
LANGCHAIN_API_KEY = userdata.get('LANGCHAIN_API_KEY')

Set the environment variables

import os
os.environ["OPENAI_API_KEY"] = OPENAI_API_KEY
os.environ["LANGCHAIN_API_KEY"] = LANGCHAIN_API_KEY

Now you can use the ChatOpenAI class as before:

from langchain_openai import ChatOpenAI
llm = ChatOpenAI(model="gpt-4o-mini")

[efriis]

hey there! As a heads up, when you "open in colab" it doesn't actually save your changes back to the repository, so you should be able to set things in plaintext while experimenting. Is that what you're referencing, or is the issue something else? Happy to take suggestions of how we can improve the docs!