Hosted Demo    Code    More AI Recipes    Blog Post    Redis Vector Search Documentation   

🔎 Redis arXiv Search

This repository is the official codebase for the arxiv paper search app hosted at: https://docsearch.redisvl.com

Redis is a highly performant, production-ready vector database, which can be used for many types of applications. Here we showcase Redis vector search applied to a document retrieval use case. Read more about AI-powered search in the technical blog post published by our partners, Data Science Dojo.

Dataset

The arXiv papers dataset was sourced from the the following Kaggle link. arXiv is commonly used for scientific research in a variety of fields. Exposing a semantic search layer enables natural human language to be used to discover relevant papers.

Application

This app was built as a Single Page Application (SPA) with the following components:

• Redis Stack for vector database
• RedisVL for Python vector db client
• FastAPI for Python API
• Pydantic for schema and validation
• React (with Typescript)
• Docker Compose for development
• MaterialUI for some UI elements/components
• React-Bootstrap for some UI elements
• Huggingface, OpenAI, and Cohere for vector embedding creation

Some inspiration was taken from this tiangolo/full-stack-fastapi-template and turned into a SPA application instead of a separate front-end server approach.

General Project Structure

/backend
    /arxivsearch
        /api
            /routes
                papers.py # primary paper search logic lives here
        /db
            load.py # seeds Redis DB
            redis_helpers.py # redis util
        /schema
            # pydantic models for serialization/validation from API
        /tests
        /utils
        config.py
        spa.py # logic for serving compiled react project
        main.py # entrypoint
/frontend
    /public
        # index, manifest, logos, etc.
    /src
        /config
        /styles
        /views
            # primary components live here

        api.ts # logic for connecting with BE
        App.tsx # project entry
        Routes.tsk # route definitions
        ...
/data
    # folder mounted as volume in Docker
    # load script auto populates initial data from S3

Embedding Providers

Embeddings represent the semantic properies of the raw text and enable vector similarity search. This applications supports HuggingFace, OpenAI, and Cohere embeddings out of the box.

Provider	Embedding Model	Required?
HuggingFace	sentence-transformers/all-mpnet-base-v2	Yes
OpenAI	text-embedding-ada-002	Yes
Cohere	embed-multilingual-v3.0	Yes

Interested in a different embedding provider? Feel free to open a PR and make a suggested addition.

Want to use a different model than the one listed? Set the following environment variables in your .env file (see below) to change:

• SENTENCE_TRANSFORMER_MODEL
• OPENAI_EMBEDDING_MODEL
• COHERE_EMBEDDING_MODEL

🚀 Running the App

1. Before running the app, install Docker Desktop.
2. Clone (and optionally fork) this Github repo to your machine.

   $ git clone https://github.com/RedisVentures/redis-arXiv-search.git

3. Make a copy of the .env.template file:

   $ cd redis-arXiv-search/
   $ cp .env.template .env

   • Add your OPENAI_API_KEY to the .env file. Need one? Get an API key
   • Add you COHERE_API_KEY to the .env file. Need one? Get an API key
4. Decide which Redis you plan to use, choose one of the methods below
   • Redis Stack runs Redis as a local docker container.
   • Redis Cloud will manage a Redis database on your behalf in the cloud.

Redis Stack Docker (Local)

Using Redis Stack locally doesn't require any additional steps. However, it will consume more resources on your machine and have performance limitations.

Use the provided docker-compose file for running the application locally:

$ docker compose -f docker-local-redis.yml up

Redis Cloud

1. Get a FREE Redis Cloud Database. Make sure to include the Search module.

2. Add the REDIS_HOST, REDIS_PASSWORD, and REDIS_PORT environment variables to your .env file.

3. Run the App:

   $ docker compose -f docker-cloud-redis.yml up

Customizing (optional)

Run local redis with Docker

docker run -d --name redis -p 6379:6379 -p 8001:8001 redis/redis-stack:latest

FastApi with poetry

To run the backend locally

1. cd backend
2. poetry install
3. poetry run start-app

poetry run start-app runs the initial db load script and launch the API

React Dev Environment

It's typically easier to build front end in an interactive environment, testing changes in realtime.

1. Deploy the app using steps above.
2. Install packages

   $ cd frontend/
   $ npm install

3. Use npm to serve the application from your machine

   $ npm run start

4. Navigate to http://localhost:3000 in a browser.

All changes to your frontend code will be reflected in your display in semi realtime.

Troubleshooting

Every once and a while you need to clear out some Docker cached artifacts. Run docker system prune, restart Docker Desktop, and try again.

This project is maintained by Redis on a good faith basis. Please, open an issue here on GitHub and we will try to be responsive to these.