kgtk-browser

Download the codebase and setup the python environment. In a terminal run the following commands ,

git clone https://github.com/usc-isi-i2/kgtk-browser
cd kgtk-browser
git checkout dev
conda create -n kgtk-env python=3.9.7
conda activate kgtk-env
pip install -r requirements.txt
pip install -e git+https://github.com/usc-isi-i2/kgtk.git@ee053b021d83c4d74797a24e98c25b71c6b852c3#egg=kgtk

Install graph-tools and jupyterlab

conda install -c conda-forge graph-tool
pip install jupyterlab

Build the Graph Cache required for KGTK Browser Backend

If you are bringing your own data (in contrast to Wikidata), please refer to BYOD. Successful run of steps in the document will produce the files required to proceed from here on. Only exception being metadata.pagerank.undirected.tsv.gz. Please read on.

The following files are required ,

• labels.en.tsv.gz
• aliases.en.tsv.gz
• descriptions.en.tsv.gz
• claims.tsv.gz
• metadata.property.datatypes.tsv.gz
• qualifiers.tsv.gz
• metadata.pagerank.undirected.tsv.gz
• class-visualization.edge.tsv.gz optional
• class-visualization.node.tsv.gz optional

The file metadata.pagerank.undirected.tsv.gz can be created by running this command ,

 kgtk --debug graph-statistics \
 -i claims.tsv.gz \
 -o metadata.pagerank.undirected.tsv.gz  \
 --compute-pagerank True  \
 --compute-hits False  \
 --page-rank-property Pundirected_pagerank \
 --output-degrees False  \
 --output-pagerank True  \
 --output-hits False  \
 --output-statistics-only \
 --undirected True \
 --log-file metadata.pagerank.undirected.summary.txt

Move the file metadata.pagerank.undirected.tsv.gz to the folder containing the other files required for the browser cache.

SQLITE or ElasticSearch

KGTK Browser can be setup with either a SQLITE DB Cache file or a KGTK Search api. We'll describe both options.

Building a SQLITE Cache DB file

• Execute this notebook.
• Set parameters: create_db = 'yes' and create_es = 'no' to create only the SQLITE DB Cache file.
• Setup other parameters as described in the notebook.

Setting up ElasticSearch Index and KGTK Search API

• Execute this notebook.
• Set parameters: create_db = 'no' and create_es = 'yes' to create and load the ElasticSearch index.
• Setup other parameters as described in the notebook.
• This will result in a ElasticSearch index which can now be used to setup the KGTK Search api.
• Setup the KGTK Search API

Running the web app using SQLITE Cache File

Update the parameter GRAPH_CACHE in the file kgtk_browser_config.py and set it to the cache file location as created in the step Building a SQLITE Cache DB file.

Ensure that recent versions of "npm" and "node" are installed:

npm --version
7.20.3

node --version
v16.6.2

NOTE: node version 18 (and above) gives the following error after npm start. Please use node version v16.6.2

Proxy error: Could not proxy request /kb/info from localhost:3000 to http://localhost:3233.
See https://nodejs.org/api/errors.html#errors_common_system_errors for more information (ECONNREFUSED).

Terminal-1: Run kgtk_browser_app.py

Set the following ENV variables in the terminal.

• KGTK_BROWSER_GRAPH_ID: Sets the title of the KGTK Browser. For example: DWD Knowledge Graph
• KGTK_BROWSER_GRAPH_CACHE: absolute path to the sqlite db graph cache file
• KGTK_BROWSER_CLASS_VIZ_DIR: path to folder where graph visualizations will be stored. This is optional and requried only in the case where you have class visualization files.

export KGTK_BROWSER_GRAPH_ID=My Dataset Browser
export KGTK_BROWSER_GRAPH_CACHE=<path to the SQLITE DB Cache created>
export KGTK_BROWSER_CLASS_VIZ_DIR=/tmp # or to some folder

The following commands will start the backend flask server,

cd kgtk-browser
conda activate kgtk-env
python kgtk_browser_app.py

Terminal-2: Run front end

The following steps will start a server on your local host at the default port (3000):

Build the frontend files ,

cd app
export REACT_APP_FRONTEND_URL='/browser'
export REACT_APP_BACKEND_URL=''

To use the SQLite text instead of Elasticsearch API, set the following environment variable.

export REACT_APP_USE_KGTK_KYPHER_BACKEND='1'

Continue ,

npm run build
npm start

A new tab should open in the default browser at http://localhost:3000

Start the server using kgtk browse command

Set the ENV variables described in the previous section and run the following commands from a terminal.

cd kgtk-browser
cd app
npm run build

export PYTHONPATH=$PYTHONPATH:$PWD
kgtk browse --host localhost --port 5000

A new tab should open in the default browser at http://localhost:5000

NOTE: using development mode turns on JSON pretty-printing which about doubles the size of response objects. For faster server response, set FLASK_ENV to production.

Build a docker image for KGTK Browser deployment (internal use only)

Running the following commands will create and push a new docker image with the tag kgtk-browser:06022022

Open a terminal and type in the following commands

cd kgtk-browser
export REACT_APP_FRONTEND_URL='/browser'
export REACT_APP_BACKEND_URL='/browser'

To use the SQLite text instead of Elasticsearch API, set the following environment variable.

export REACT_APP_USE_KGTK_KYPHER_BACKEND='1'

Set the following environment variable to 0 or 1 if you have have the file derived.P31279star.tsv.gz loaded into the cache.

export REACT_APP_P31279STAR_NA='0' # if the file is available, otherwise

export REACT_APP_P31279STAR_NA='1'

continue ,

cd app
npm run build

cd ..
docker build -t docker-reg.ads.isi.edu:443/kgtk-browser:06022022 . && docker push docker-reg.ads.isi.edu:443/kgtk-browser:06022022