Skip to content

17.02 LangChain Hub

One of the biggest bottlenecks when building LLM applications is managing the "Secret Sauce": The Prompt.

Hardcoding massive, multi-paragraph prompts directly into your Python files makes your codebase messy, hard to version control, and difficult for non-engineers to tweak. LangChain Hub (a feature integrated within LangSmith) is the solution.

What is LangChain Hub?

Think of LangChain Hub as a public registry—like NPM or Docker Hub—but specifically designed for Prompts, Chains, and Agents.

It serves two primary purposes: 1. Discovery: Browse and download highly-optimized prompts created by the community or framework creators. 2. Version Control: Host your own prompts remotely so you can update them without deploying new code.

Exploring the Hub

When you visit the Hub, you will find prompts categorized by use case.

Filtering by Use Case

You can easily filter for specific tasks such as: - RAG (Retrieval-Augmented Generation) - Autonomous Agents (e.g., ReAct prompts) - Information Extraction - Text Classification - SQL Generation

Model Optimization

A prompt that works brilliantly on OpenAI's gpt-4o might degrade in performance when run on Google's gemini-pro. The Hub helps identify prompts optimized for specific model architectures and vendors. You can sort by popularity, downloads, and likes to find battle-tested permutations.

Using Hub Prompts Logically

Instead of polluting your application with strings of text, you can pull a prompt dynamically at runtime using the hub.pull() method.

from langchain import hub
from langchain.chains import RetrievalQA

# Pull the highly-popular RAG prompt from the Hub
prompt = hub.pull("rlm/rag-prompt")

# Inject the remote prompt directly into your chain
qa_chain = RetrievalQA.from_chain_type(
    llm=llm,
    retriever=vectorstore.as_retriever(),
    chain_type_kwargs={"prompt": prompt}
)

[!NOTE] Every prompt on the Hub is fully versioned. You can view the commit history of a prompt to see exactly how the instructions evolved over time.

The Playground Integration

Perhaps the most powerful feature of the Hub is its integration with the LangSmith Playground.

If you find a popular "ReAct Agent" prompt on the Hub, you don't need to write a Python script to test it. You can open it directly in the web UI Playground, plug in variables, tweak the temperature, swap the underlying LLM provider, and test the output behavior instantly.