Langchain Outline Document Loader

Overview

langchain-outline provides a DocumentLoader for Langchain that allows you to load documents directly from an Outline knowledge base instance.

This loader uses the Outline API to fetch all documents within your instance.

Installation

pip install langchain-outline

Configuration

The loader requires the URL of your Outline instance and an API key for authentication.

Usage

import os
from langchain_outline.outline import OutlineLoader

# Option 1: Using environment variables (ensure they are set)
# os.environ["OUTLINE_INSTANCE_URL"] = "YOUR_OUTLINE_URL"
# os.environ["OUTLINE_API_KEY"] = "YOUR_API_KEY"
# loader = OutlineLoader()

# Option 2: Passing parameters directly
loader = OutlineLoader(
    outline_base_url="YOUR_OUTLINE_URL",
    outline_api_key="YOUR_API_KEY"
)

# Load documents (iteratively)
docs_iterator = loader.lazy_load()
for doc in docs_iterator:
    print(f"Loaded document: {doc.metadata['title']}")
    # Process the document...

# Or load all documents into a list (might consume more memory for large instances)
all_docs = loader.load()
print(f"Loaded {len(all_docs)} documents.")

Replace "YOUR_OUTLINE_URL" and "YOUR_API_KEY" with your actual Outline instance URL and API key.

Async usage

OutlineLoader also supports non-blocking async loading, backed by httpx.AsyncClient, for use in async applications:

import asyncio
from langchain_outline.outline import OutlineLoader

async def main():
    loader = OutlineLoader(
        outline_base_url="YOUR_OUTLINE_URL",
        outline_api_key="YOUR_API_KEY"
    )

    # Load documents (iteratively)
    async for doc in loader.alazy_load():
        print(f"Loaded document: {doc.metadata['title']}")
        # Process the document...

    # Or load all documents into a list (might consume more memory for large instances)
    all_docs = await loader.aload()
    print(f"Loaded {len(all_docs)} documents.")

asyncio.run(main())

httpx is a direct dependency of this package used for the async loading path.

For maintainers

To build and test publishing to TestPyPi

To publish a new version of the package you can use poetry (or twine, etc). First test using TestPyPi first.

1. Make sure you have TestPyPi as a repo:

poetry config repositories.testpypi https://test.pypi.org/legacy/

2. Obtain API token from TestPyPi and add it to poetry config:

poetry config pypi-token.testpypi <<token from testpypi>>

3. Publish the package to TestPypi (make sure version defined in pyproject.toml follows PEP440):

poetry publish --build --repository testpypi

To publish to PyPi

To publish a new version of the package you can use poetry (or twine, etc). First test using TestPyPi first.

1. Obtain API token from PyPi and add it to poetry config:

poetry config pypi-token.pypi <<token from pypi>>

3. Publish the package to PyPi (make sure version defined in pyproject.toml follows PEP440):

poetry publish --build