README.md

Jina 2.0 beginners guide

For this guide we are using the repositories from alexcg1 cloned on July 2, 2021:

• Search frontend with Streamlit
• Search backend using Jina

This steps were tested on Linux (Ubuntu 20.04), this may work the same for Mac, for Windows 10 it is supposed to work on WSL or Docker but my experience dealing with configurations alone was painful and out of the scope for developing with Jina.

Requirements

• Working installation of virtualenv
• Git

Steps to test or develop applications

Open Terminal at project folder

Clone repository and cd into it

git clone https://github.com/alexcg1/jina-meme-search-image-backend
cd jina-meme-search-image-backend

Create and activate virtual environment

virtualenv env --python=python3.8
source env/bin/activate

Install requirements.txt

pip install -r requirements.txt

Get models with shell scripting

sh get_models.sh

Get json with images metadata

sh get_data.sh

⚠️ Check on folder data that your memes.json file is more than 70 MB, if it isn't replace with file from Kaggle

Get images with python script, the parameter is the number of images you want to download minus 1

python get_images.py 101

Launch indexing, check flows/index.yml which has good comments about what each line does

python app.py -t index -n 100

Launch search endpoint, keep this Terminal alive

python app.py -t query_restful

For the front end, open a new Terminal at your base project folder, cd into it, create and activate virtual environment, install the requierements and launch the Streamlit application

git clone https://github.com/alexcg1/jina-meme-search-frontend
cd jina-meme-search-frontend
virtualenv env --python=python3.8
source env/bin/activate
pip install -r requirements.txt
streamlit run app.py

If it is your first time running streamlit it may prompt for an email, just press Enter.

Now you can see Jina in action:

• Open a web browser with your streamlit URL
• Select the Image radio button
• Expand Settings, set your endpoint to the one your Jina service is using, i.e. http://localhost:45680/search and hit Enter
• Click Browse files and select an image to search (one from the data folder or others could be used for testing)
• Click Search

For better understanding of how Jina works check the repository Readme and take a look at the cookbooks