LangChain Unstructured Integration

1.0.1 · active · verified Thu Apr 16

langchain-unstructured is an integration package connecting the LangChain framework with Unstructured, a library for parsing and processing unstructured documents. It provides document loaders to extract text and metadata from various file types (PDFs, images, HTML, etc.) for use in LangChain applications. The current version is 1.0.1, with a release cadence that has recently seen a rapid transition to 1.x and subsequent minor updates.

Common errors

```
ModuleNotFoundError: No module named 'langchain_unstructured'
```
cause The `langchain-unstructured` package is not installed.

fix Run `pip install langchain-unstructured`.
```
ModuleNotFoundError: No module named 'unstructured'
```
cause The core `unstructured` library, a dependency, is missing.

fix Run `pip install unstructured`. For specific file types, consider `pip install "unstructured[pdf]"` or other extras.
```
ImportError: cannot import name 'UnstructuredFileLoader' from 'langchain.document_loaders'
```
cause Attempting to import `UnstructuredFileLoader` from the old `langchain` package path.

fix Change the import to `from langchain_unstructured.document_loaders import UnstructuredFileLoader`.
```
ValueError: Unstructured API key not provided.
```
cause When using `UnstructuredAPIFileLoader`, the `UNSTRUCTURED_API_KEY` environment variable is not set and no API key was provided in the constructor.

fix Set the environment variable: `export UNSTRUCTURED_API_KEY='your_api_key'` or pass it directly: `UnstructuredAPIFileLoader(..., unstructured_api_key='your_api_key')`.

Warnings

breaking Migration from `langchain` document loaders to `langchain-unstructured` package.
Fix: Update imports from `from langchain.document_loaders import ...` to `from langchain_unstructured.document_loaders import ...` and ensure `langchain-unstructured` is installed.
breaking Upgrade of `langchain-core` dependency version in `langchain-unstructured` v1.0.0.
Fix: Ensure `langchain-core` is updated to a compatible version (e.g., `pip install --upgrade langchain-core`) to avoid dependency conflicts or unexpected behavior.
gotcha Unstructured requires additional dependencies for specific file types.
Fix: For parsing PDFs, install `unstructured[pdf]` (e.g., `pip install "unstructured[pdf]"`). For images, `unstructured[image]` (which requires Tesseract OCR). Check the Unstructured documentation for specific requirements.
gotcha Using `UnstructuredAPIFileLoader` requires an API key.
Fix: Set the `UNSTRUCTURED_API_KEY` environment variable before instantiating `UnstructuredAPIFileLoader`, or pass it directly via `unstructured_api_key` parameter.

Install

pip install langchain-unstructured unstructured Basic Installation
pip install "langchain-unstructured[pdf]" "unstructured[pdf]" With PDF Support

Imports

UnstructuredFileLoader

from langchain_unstructured.document_loaders import UnstructuredFileLoader

UnstructuredAPIFileLoader

wrong:

from langchain.document_loaders import UnstructuredAPIFileLoader

correct:

from langchain_unstructured.document_loaders import UnstructuredAPIFileLoader

As of v1.x, these loaders are part of the `langchain_unstructured` package, not `langchain` directly.

Quickstart

This quickstart demonstrates how to use `UnstructuredFileLoader` to load text from a local file. For more complex file types like PDFs or images, ensure you have the necessary `unstructured` extra dependencies installed (e.g., `pip install "unstructured[pdf]"`). For `UnstructuredAPIFileLoader`, ensure the `UNSTRUCTURED_API_KEY` environment variable is set.

import os
from langchain_unstructured.document_loaders import UnstructuredFileLoader

# Create a dummy file for demonstration
with open("example.txt", "w") as f:
    f.write("This is a test document.\n")
    f.write("It contains some sample text to be loaded.")

# Instantiate the loader for a local file
loader = UnstructuredFileLoader("example.txt")

# Load the document(s)
docs = loader.load()

# Print the content of the first loaded document
if docs:
    print(f"Loaded {len(docs)} document(s).")
    print(f"Page content: {docs[0].page_content[:50]}...")
    print(f"Metadata: {docs[0].metadata}")

# Clean up the dummy file
os.remove("example.txt")

view raw JSON →