langchain - 💡(How to fix) Fix Bug: path traversal in Chroma.add_images() via unsanitized URI [1 pull requests]

Package: langchain-chroma

Description

Chroma.add_images(uris=[...]) passes caller-supplied URIs directly to Path(uri).open("rb") with no validation. Any caller with access to the public API can read arbitrary files from the filesystem.

Reproducer

import chromadb
from langchain_chroma import Chroma

store = Chroma(client=chromadb.Client(), embedding_function=None)
store.add_images(uris=["../../../../etc/passwd"])  # reads arbitrary file, returns content base64-encoded

Expected behavior

ValueError raised before the file is opened.

Proposed fix

Add a _validate_image_uri instance method called before encode_image opens the file:

• Default (no config): reject absolute paths and any .. path component
• With allowed_image_dir (new keyword-only __init__ param): resolve path canonically and enforce containment within the allowed directory (symlinks followed via Path.resolve())

I have a working fix + unit tests ready. Happy to submit a PR once assigned.