This template performs RAG using Milvus Vector Store and NVIDIA Models (Embedding and Chat).
You should export your NVIDIA API Key as an environment variable. If you do not have an NVIDIA API Key, you can create one by following these steps:
- Create a free account with the NVIDIA GPU Cloud service, which hosts AI solution catalogs, containers, models, etc.
- Navigate to
Catalog > AI Foundation Models > (Model with API endpoint).
- Select the
APIoption and click
- Save the generated key as
NVIDIA_API_KEY. From there, you should have access to the endpoints.
For instructions on hosting the Milvus Vector Store, refer to the section at the bottom.
To use this package, you should first have the LangChain CLI installed:
pip install -U langchain-cli
To use the NVIDIA models, install the Langchain NVIDIA AI Endpoints package:
pip install -U langchain_nvidia_aiplay
To create a new LangChain project and install this as the only package, you can do:
langchain app new my-app --package nvidia-rag-canonical
If you want to add this to an existing project, you can just run:
langchain app add nvidia-rag-canonical
And add the following code to your
from nvidia_rag_canonical import chain as nvidia_rag_canonical_chain
add_routes(app, nvidia_rag_canonical_chain, path="/nvidia-rag-canonical")
If you want to set up an ingestion pipeline, you can add the following code to your
from nvidia_rag_canonical import ingest as nvidia_rag_ingest
add_routes(app, nvidia_rag_ingest, path="/nvidia-rag-ingest")
Note that for files ingested by the ingestion API, the server will need to be restarted for the newly ingested files to be accessible by the retriever.
(Optional) Let's now configure LangSmith. LangSmith will help us trace, monitor and debug LangChain applications. LangSmith is currently in private beta, you can sign up here. If you don't have access, you can skip this section
export LANGCHAIN_PROJECT=<your-project> # if not specified, defaults to "default"
If you DO NOT already have a Milvus Vector Store you want to connect to, see
Milvus Setup section below before proceeding.
If you DO have a Milvus Vector Store you want to connect to, edit the connection details in
If you are inside this directory, then you can spin up a LangServe instance directly by:
This will start the FastAPI app with a server is running locally at http://localhost:8000
We can see all templates at http://127.0.0.1:8000/docs We can access the playground at http://127.0.0.1:8000/nvidia-rag-canonical/playground
We can access the template from code with:
from langserve.client import RemoteRunnable
runnable = RemoteRunnable("http://localhost:8000/nvidia-rag-canonical")
Use this step if you need to create a Milvus Vector Store and ingest data. We will first follow the standard Milvus setup instructions here.
Download the Docker Compose YAML file.
wget https://github.com/milvus-io/milvus/releases/download/v2.3.3/milvus-standalone-docker-compose.yml -O docker-compose.yml
Start the Milvus Vector Store container
sudo docker compose up -d
Install the PyMilvus package to interact with the Milvus container.
pip install pymilvus
Let's now ingest some data! We can do that by moving into this directory and running the code in
Note that you can (and should!) change this to ingest data of your choice.