Building a Semantic Search Powered App with Atlas

Tutorial and template for a semantic search app powered by the Atlas Embedding Database and FastAPI. Optional integrations include the OpenAI Embedding API and Langchain.

Getting started

The typical process for building a semantic search app includes:

1. Gathering a dataset of text, images or other content you want your app to search through.
2. Generating embeddings of your data. This is usually done using an Embedding API such as Cohere or OpenAI.
3. Writing code that performs vector similarity search over your embeddings and returns back your content.
4. Integrating all of this into a back-end which you can hit over a REST API.

This app template simplifies the process by providing a ready to deploy FastAPI and React app that interfaces with the Atlas Embedding Database.

Atlas allow you to upload your content (text documents, embeddings) and instantly access semantic search over your data. Importantly, Atlas comes built with a visual debugger allowing you to see the types of results your semantic search app will generate for any query.

Tutorial

Starting the app

First build the FastAPI docker image. You only have to do this on initial build or when you add new dependencies to the requirements.txt file:

DOCKER_BUILDKIT=1 docker build -t backend_api --progress plain -f backend/Dockerfile.buildkit .

Then, start the backend with:

docker compose up --build

You will need to insert your Nomic API Key (giving your app access to your Atlas Embedding Database instance) into the settings file or the docker compose environment variables. You can find this API Key by making an Atlas account.

Uploading and indexing your content in Atlas

We will use demo news data in this tutorial. To upload the demo data go through the Content Upload Tutorial.

This tutorial stores your data in an Atlas Embedding DB project called '10k News Articles'.

You should paste this project name into the settings file under the environment variable atlas_project_name. Additionally, paste in your OpenAI API key into the settings file.

Spinning up your app

Run docker compose up to spin up the backend. Monitor the logs for errors in-case you forgot to set an environment variable above.

Testing your vector search functionality

Once the FastAPI ap is started you can access its documentation and test the search endpoint by going to:

localhost:80/docs

Serving the front end

The front-end is not implemented yet for this tutorial. Make a PR that builds a React App and we will merge it!