Ai

LangChain Text Splitters 1.1.2 Improves URL Safety and Fixes JSON Splitting Data Loss

LangChain Text Splitters 1.1.2 Improves URL Safety and Fixes JSON Splitting Data Loss

LangChain has released langchain-text-splitters 1.1.2, a focused update that improves security and reliability for developers using its text chunking utilities in AI pipelines. The release is headlined by an SSRF-safe transport change for URL-based text extraction, a fix to prevent silent data loss in RecursiveJsonSplitter, and a set of dependency, testing, and security-related maintenance updates.

What Changed

The most important fix in version 1.1.2 is the deprecation of the previous URL transport in favor of an SSRF-safe transport for split_text_from_url. This matters for teams that ingest remote content into LLM workflows, because server-side request forgery protections are increasingly critical in AI systems that fetch and process external data.

Another notable change addresses silent data loss in RecursiveJsonSplitter when handling empty dictionary values. That bug could lead to incomplete structured-data processing, so the fix should improve trustworthiness for applications splitting JSON documents before retrieval, indexing, or model input.

The release also adds support for spaCy tests on Python 3.14, helping keep the package aligned with newer Python environments. Alongside that, the update includes a broad set of maintenance changes: dependency bumps for langsmith, pytest, requests, nltk, orjson, and tornado, plus lockfile, CI, and linting improvements.

Security maintenance is also visible in the changelog, including alignment on pygments>=2.20.0 across packages in response to CVE-2026-4539. While not all of these changes alter runtime behavior directly for text splitting, they strengthen the package ecosystem around the release.

Why It Matters

This is a small but meaningful release for developers building AI ingestion and retrieval systems on LangChain. The SSRF-safe transport update reduces risk in workflows that fetch remote text, which is especially important for enterprise and production deployments handling untrusted URLs.

The JSON splitter fix is equally practical. Silent failures are among the hardest bugs to detect in document-processing pipelines, and preventing dropped empty values can improve downstream consistency for embedding, chunking, and structured context preparation.

Even the maintenance changes carry weight. Dependency refreshes, Python compatibility work, and security patch alignment help teams keep modern AI stacks stable without waiting for a larger feature release. For users already on 1.1.1, version 1.1.2 looks like a sensible upgrade focused on safer defaults and more dependable parsing behavior.

Official Source: https://github.com/langchain-ai/langchain/releases/tag/langchain-text-splitters%3D%3D1.1.2