The Universal NFT Vector Database is an initiative to store a vectorized version of every image-based Ethereum NFT (ERC-721). This project leverages a variety of technology to make this possible including Alchemy API, the Graph Protocol, and Pinecone. Our complete software infrastructure can be found below:
Below we have outlined the major parts/systems as part of this project and their use case in our overall system.
| System | Use Case |
|---|---|
| Analytics Dashboard | This is the frontpage of our client side application, primarily used to display statistics and other information regarding the data in our database and task queue. |
| Search Page | The search page is an interactive page for users to query the database and find similar NFTs while also visualizing their NFTs relative to similar NFTs on a two-dimensional graph. |
| Graph Protocol Integration | The Graph Protocol is used as a querying layer to obtain all of the NFT collection information on the blockchain. |
| Alchemy API Integration | Alchemy API is used to obtain all the individual NFT information for each NFT collection which is then stored in our system. |
| Backend Server | Our backend server allows us to interact with the data and obtain information from our databases to display on our analytics dashboard. |
| RegNetY-080 Image Embedding Integration | The image embeddings serve as a way to standardize the representation of the images and store them in a vector database. |
| Search API | The search API allows users to query the vector database for similar NFTs based on a source image. It uses cosine distance as the metric for finding the closest NFTs. |
| Visualization API | The visualization API allows users to visualize their source image relative to similar NFTs -- it uses truncated singular value decomposition to do this. |
| Task Queue Worker Server | The task queue workers server enables horizontal scalability for the system because it allows hundreds of thousands of NFTs to be sitting in a queue as they eventually get processed. |
We always welcome pull requests to this repository for fixing and further improving our database.
We have provided a setup script available (setup.sh) to setup the system locally -- this has only been tested with WSL and Ubuntu so use at your own risk. Before running the setup script, please ensure you have the following installed:
- MongoDB Community Edition: https://www.mongodb.com/docs/manual/administration/install-community/
- Python 3.8+: https://www.python.org/downloads/release/python-380/
- Nodejs v16+: https://nodejs.org/en/download/
You will also need to add an enviornment file for this project. Directories with an .env file:
python_server/src/client/celery_queue/benchmarks/
We have provided an .env.example for each .env file. We also provide a doppler.example to see what other environment variables we have because we use Doppler as our unified secret management software.
Then, to Setup this project:
sudo chmod +x ./setup.sh
sudo ./setup.sh
Alternatively, to setup this project manually:
- Go to the
clientfolder and npm install all of the dependencies:
cd client
npm install
- Go to the
python_server/srcfolder and create a python virtual environment folder. Then install all of the dependencies:
cd python_server/src
python3 -m venv venv
source venv/bin/activate
pip3 install -r requirements.txt
- Go to the
celery_queue/folder and create a python virtual environment folder. Then install all of the dependencies:
cd celery_queue
python3 -m venv venv
source venv/bin/activate
pip3 install -r requirements.txt
- To Run this Project:
sudo chmod +x ./run.sh
sudo ./run.sh
- Run the Client:
npm start --prefix client/
- Activate the python server virtual enviornment and run the search API:
cd python_server/src
source venv/bin/activate
python app.py
- Activate the Celery queue virtual enviornment and run the task queue
cd celery_queue
source venv/bin/activate
celery -A tasks worker