GitHub - redis-developer/redis-product-search: Visual and semantic vector similarity with Redis Stack, FastAPI, PyTorch and Huggingface. (original) (raw)

Redis Vector Search Demo Application

This demo showcases the vector search similarity (VSS) capability within Redis Stack and Redis Enterprise. Through the RediSearch module, vector types and indexes can be added to Redis. This turns Redis into a highly performant vector database which can be used for all types of applications.

The following Redis Stack capabilities are available in this demo:

• Vector Similarity Search
  • by image
  • by text
• Multiple vector indexing types
  • HNSW
  • Flat (brute-force)
• Hybrid Queries
  • Apply tags as pre-filter for vector search

Application

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

• Redis Stack: Vector database + JSON storage
• RedisVL for Python vector db client
• FastAPI for backend API
• Pydantic for schema and validation
• React (with Typescript)
• Docker Compose for development
• MaterialUI for some UI elements
• React-Bootstrap for some UI elements
• Pytorch/Img2Vec and Huggingface Sentence Transformers for vector embedding creation

Some inspiration was taken from this Cookiecutter projectand turned into a SPA application instead of a separate front-end server approach.

General Project Structure

Much inspiration taken from tiangelo/full-stack-fastapi-template

/backend
    /productsearch
        /api
            /routes
                product.py # primary API logic lives here
        /db
            load.py # seeds Redis DB
            utils.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

Datasets

The dataset was taken from the the following Kaggle links.

• Large Dataset
• Smaller Dataset

A formatted version is available for use with this demo at:

Running the App with docker-compose

Before running the app, install Docker Desktop.

1. Copy .env file
2. Run docker containers with make

Note: you can add --build and --force-recreate if caching old images.

Running local without docker-compose

Run frontend

1. Install NPM packages
   $ cd frontend/
   $ npm install
2. Use npm to serve the application from your machine
3. Navigate to http://localhost:3000 in a browser.

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

Run backend

Pre-step: install poetry.

1. cd backend
2. poetry install to get necessary python deps
3. poetry run start to launch uvicorn server with FastAPI app

vscode debugger

Included in the project is a ./vscode/launch.json for local debugging purposes.

Troubleshooting

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

Open an issue here on GitHub and we will try to be responsive to these. Additionally, please consider contributing.