Skip to content

Full setup

Jump to bottom
Sivasurya Santhanam edited this page Oct 8, 2019 · 1 revision

The architecture shows the component diagram of integrating IslandViz with a conversational interface. These modules are developed independently so as to facilitate modularity and development. Full setup ensures setting up Island-viz with the database and conversational services.

Component diagram of the architecture

Device & Context Monitor : Hololens is the device at use and the code is maintained in the current repository

Controller : Controller is the Central unit which communicates with all the services. Controller is maintained at iviz-controller

Repository : Repository is where the database is maintained. We have used a Java project based on OSGi framework from RCE. The architecture is translated into a property graph (Neo4j Graph DB) using jQAssistant.

Conversational services : This includes detecting intents and entities from the user utterance and Converting those intents into Graph queries. The NLP pipeline is maintained at rique

Installation and Setup

Clone the repositories of Device, rique and the controller in the same directory adjacent to each other.

Follow the readme.md instructions in each repository for the setup.

⚠️ It is highly encouraged to create a virtual environment while installing python packages, which otherwise might have dependency issues

Deployment

  1. Build the island-viz application and deploy it to the Hololens device (Device)
  2. Start the Neo4j server and ensure the graph is running (Repository)
  3. Run the conversational interface services such as Rasa x and Rasa actions (rique)
  4. Run the flask application (controller)

The island-viz application in Hololens now should be able to successfully communicate with all the modules. Check out rique for the list of user utterances one can ask.

🔫 Troubleshooting

Clone this wiki locally