Skip to main content
Aerospike 8 introduces a new era of real-time transaction processingLearn more
Loading

Create a RAG application using LangChain

This guide describes how to create an example Retrieval Augmented Generation (RAG) application using LangChain templates.

Prerequisites​

You need the following components to set up this application:

Create a LangChain application​

  1. Install the LangChain CLI, which gives you access to a set of templates for building applications, with the following command.

    pip install -U langchain-cli

  2. Run langchain app new APP_NAME to create the necessary Python files and install dependencies. In this example, the name is my-app.

    langchain app new my-app && cd ./my-app

  3. While still in your my-app application folder, run the following command to apply the Aerospike RAG application template:

    langchain app add --repo="aerospike/rag-aerospike/" --branch="main"

  4. Edit app/server.py in two places to set the route for the RAG application. First, add the following to your import statements:

    from rag_aerospike import chain as rag_aerospike_chain

  5. Then replace add_routes(app, NotImplemented) with your import and path to your repo:

    add_routes(app, rag_aerospike_chain, path="/rag-aerospike")

  6. Save and close app/server.py.

  7. Run the following commands to export your OpenAI API key and any other relevant environment variables:

    export OPENAI_API_KEY=<your-OpenAI-api-key>
    export AVS_HOST=<your-AVS-host> #default localhost
    export AVS_PORT=<your-AVS-port> #default 5000
    export AVS_NAMESPACE=<your-AVS-namespace> #default test
    export DATASOURCE=<your-datasource-url> #default https://aerospike.com/files/ebooks/aerospike-up-and-running-early-release3.pdf

    If you have a paid LangSmith account, you can set your LangChain credentials as environment variables to enable application monitoring. These are not required for this example application.

    export LANGCHAIN_TRACING_V2=true
    export LANGCHAIN_API_KEY=<your-api-key>
    export LANGCHAIN_PROJECT=<your-project>  # if not specified, defaults to "default"

  8. Serve your LangChain application with the following command:

    langchain serve

    It takes some time to tokenize and embed your datasource. When this is finished, you should see the following log line:

    INFO:     Application startup complete.

Test your application​

When your application is serving, navigate to http://localhost:8000/rag-aerospike/playground/ and ask a context-relevant question.

image