Skip to content

Latest commit

 

History

History

polyglot-chat-app

README.md

Polyglot Chat Application

This example demonstrates how to integrate Python on GraalVM with a Micronaut application. The application uses the Gradle build tool.

Prerequisites

Preparation

  1. Download the latest GraalPy as described on https://www.graalvm.org/python/. For example on Linux:

    wget https://github.com/oracle/graalpython/releases/download/graal-23.1.1/graalpy-23.1.1-linux-amd64.tar.gz
tar xzf graalpy-23.1.1-linux-amd64.tar.gz

  2. Install the required packages for this demo into the resources directory:

    graalpy-23.1.1-linux-amd64/bin/graalpy -m venv src/main/resources/venv
src/main/resources/venv/bin/graalpy -m pip install nltk

  3. Optional: Download and install GraalVM JDK for Java 21 or later to run Python with runtime compilation and to build a native image. The demo will work on any OpenJDK distribution, but will be much faster on GraalVM JDK.

Building and Running the application:

  1. Build application with Gradle:

    ./gradlew run

  2. Navigate to http://localhost:12345/#/chat/bob

    You can connect from multiple browsers and chat via websockets. The Python code will load a language model in the background. Once it is ready, it will analyse the sentiments of all messages and add an emoji to the message to indicate the feelings conveyed. A chat may look like this (newest message at the top):

    [bob 😀] awesome, GraalVM and GraalPy rock!
[bob 🫥] are we done yet?
[bob 💬] still loading the sentiment model I believe

Building a Native Image

The application can be AOT compiled using GraalVM Native Image. The Python code has to be shipped in a resources directory that is kept next to the native executable.

  1. Build a native executable with the Micronaut AOT support:

    ./gradlew nativeCompile

  2. Copy the venv into the output resources directory:

    cp -R src/main/resources/venv/ build/native/nativeCompile/resources/python/

  3. Run the native executable:

    build/native/nativeCompile/websocket.chat

Learn More

Learn more about GraalVM polyglot capabilities here.