Skip to content

Latest commit

 

History

History
114 lines (72 loc) · 3.87 KB

local_execution.md

File metadata and controls

114 lines (72 loc) · 3.87 KB

CosmosAIGraph : Local Execution

These are the instructions for those who want to explore and execute CosmosAIGraph on their workstation.

Other users may simply wish to deploy the pre-built Docker containers to an Azure Container Apps as described here.

Compile Graph microservice

Before launching the solution, be sure that the Graph microservice, which is implemented in Java, is compiled. To do so, execute the provided Gradle script in your environment of choice.

For example, in IntelliJ IDEA, you'd need to open a folder impl/graph_app, navigate to Gradle sidebar and double-click build task:

[Optional] Unzip file libraries-graph.zip

If you'd like to load the graph data from local file, follow these steps. Otherwise, the data is provided as part of Cosmos DB account which is populated with this data as described here.

The full-size libraries-graph.nt (N-triples) file is too large for GitHub. Therefore it has been zipped and added to this repo as a file impl/app/rdf/libraries-graph.zip. Navigate to this directory and unzip this file to impl/app/rdf/libraries-graph.nt .

Modes of Execution

Three different modes of execution are recommended, please use the mode most natural to your development style.

The three modes are:

  • App-per-Terminal (Recommended)
  • Launcher Script
  • Docker Compose

App-per-Terminal

In this mode, two Terminal windows are created each hosting its application - web and graph.

Before you run them, be sure to set your local environment variables, either by preparing an .env file, or by preparing and running impl/set-caig-env-vars-sample.ps1 script, or just adding them manually.

Then be sure you have created Python virtual environment in the web app, by running impl/web_app/venv.ps1 script.

To launch the terminals, simply run impl/run.ps1 script.

Docker Compose

This is the closest method to running your solution to Azure. This approach executes the application packaged as Docker Containers rather than as local files.

Start your Docker Desktop application if it's not already running.

Be sure to modify your environment variables in the appropriate docker-compose-xxx.yml ile before starting the microservices.

The following docker-compose YAML file is provided:

  • docker-compose-with-jena.yml
    • This also uses the Python-based web application
    • This uses the Java-based graph microservice using Apache Jena

Create two PowerShell Terminal windows, and navigate to the impl/app/ directory in each.

In the first terminal window, execute the following command to start the application (both microservices).

docker compose -f docker/docker-compose-with-jena.yml up

You should see similar verbose output that includes the following:

In the second terminal window, execute the following command to terminate the application.

docker compose -f docker/docker-compose-with-jena.yml down

You should see similar verbose output that includes the following:

The Docker Containers

These three pre-built Docker containers exist on DockerHub:

  • cjoakim/caig_web_v3:latest
  • cjoakim/caig_graph_v3:latest
  • cjoakim/caig_graph_java_jena_v1:latest

These are used by default by the above docker-compose scripts and also by the Azure Container Apps deployment process.

If you wish to rebuild these containers and deploy them to your own Container Registry, please see the following Dockerfiles in this repo. You're free to modify these as necessary. Please change the cjoakim prefix to your own identifier.

  • impl/app/docker/Dockerfile_graph
  • impl/app/docker/Dockerfile_web
  • impl/java_jena_graph_websvc/Dockerfile