Build agents and RAG pipelines across the NVIDIA ecosystem with one LangChain code path. These samples wire chat, embedding, and reranking models into agents and chains with LangChain, orchestrate multi-step flows with LangGraph, and trace and evaluate runs with LangSmith, using langchain-nvidia-ai-endpoints. The same code reaches models hosted on the NVIDIA API Catalog or self-hosted as NIM microservices on your own GPUs, with NeMo Retriever for embedding and reranking.
| Example | What it shows | Stack |
|---|---|---|
| openshell-deepagent | General-purpose coding agent in LangGraph Studio, powered by Nemotron 3 Super, with code execution in OpenShell and persistent local memory | Deep Agents + OpenShell + NVIDIA Nemotron + LangGraph Studio |
| nemotron-openshell-incident-agent | Deep Agent that analyzes incident logs by running Python in a policy-governed OpenShell sandbox, on NVIDIA Nemotron Ultra via NIM, traced in LangSmith | Deep Agents + langchain-nvidia-ai-endpoints + LangSmith |
| nemotron-openshell-contained-incident-agent | Parent Deep Agent, subagents, tools, and incident-analysis executable all running inside one policy-governed OpenShell sandbox | Deep Agents + OpenShell + NVIDIA Nemotron + LangSmith |
Clone the repo and change into an example:
git clone https://github.com/langchain-samples/langchain-nvidia-samples.git
cd langchain-nvidia-samples/examples/<example-name>From there, follow that example's own README for setup, the NVIDIA and LangSmith keys or NIM endpoints it needs, and how to run it.
This project is licensed under the MIT License.