LangChain

What it is

LangChain is a framework, not an app: a Python (and JS/TS) library for building LLM apps and agents, with a standard interface for models, embeddings, vector stores and retrievers. You write code to build a RAG pipeline with it; there is no chat-with-your-files window out of the box. LangSmith (a separate, mostly hosted product) adds tracing and evals.

New to local RAG and don't want to code? Start with a desktop app instead. See LangChain RAG vs. local RAG.

License

MIT

Latest release

langchain-core 1.7.3

Release date

2026-10-29

Hardware

Framework (Python library; LangChain.js for JS/TS). Runs wherever your code runs

Document types

Any via document loaders and integrations unverified

OCR / tables / retrieval

Pluggable: vector stores (FAISS, Chroma, PGVector, Pinecone, Weaviate, Elasticsearch...), hybrid search where the store supports it, rerankers (e.g. Cohere Rerank)

Cloud calls

Self-hostable code, but the official quickstart defaults to OpenAI (cloud). Fully local only if you wire in local models and a local vector store

Fully offline

Needs config

GitHub stars

148,778

Last commit

2026-10-31 (last push, GitHub API 2026-10-31)

Retrieval

Vector search Hybrid (vector + keyword) Agentic / iterative

Based on: docs.langchain.com retrieval docs (2026-10-31): 2-step, agentic and hybrid RAG architectures; pluggable vector stores; hybrid search via stores like Weaviate; rerankers (Cohere) (checked 2026-10-31). Vector vs. vectorless RAG explained →

NVIDIA DGX Spark

DGX Spark: Unknown

No DGX Spark report found yet. We only mark a tool as working when there is a published source.

DGX Spark guide →