langchain-sourcey 0.1.0

LangChain retriever for Sourcey-generated documentation sites.

langchain-sourcey

langchain-sourcey is a LangChain retriever for Sourcey-generated documentation sites.

It works against Sourcey's public build artefacts instead of a private hosted API:

• search-index.json for candidate discovery
• llms-full.txt for full-page content hydration
• canonical page URLs for citations

Install

pip install langchain-sourcey

Usage

from langchain_sourcey import SourceyRetriever

retriever = SourceyRetriever(
    site_url="https://docs.example.com/reference",
    top_k=4,
)

docs = retriever.invoke("How does search work?")

for doc in docs:
    print(doc.metadata["source"])
    print(doc.page_content[:160])

The site_url should point at the root of a published Sourcey docs build. The retriever fetches search-index.json and llms-full.txt from that root.

Output requirements

For best results, the Sourcey site should:

• publish search-index.json
• publish llms-full.txt
• set siteUrl in sourcey.config.ts so citations are canonical

If llms-full.txt is not available, the retriever falls back to extracting plain text from the matched HTML page.

Scope

This package currently ships SourceyRetriever only. A document loader is intentionally out of scope until the retriever proves its usage.