Set up sample Docker containers to run your Python topic detection models, and
then serve the models to Pega Platform through an API endpoint.
Deploy the sample containers in a cloud or on-premises environment.
Important: Before using the sample containers in production, you need to manage
any additional production-level requirements, such as security, load balancing, and
monitoring.
Before you begin: Set up your Docker environment. For more information, see
the Docker documentation.
-
Train your topic detection model.
-
Save the model in one of the supported formats:
- For machine learning models:
.bst,.joblib,.pkl
- For deep learning models: .h5
Note: Ensure that the model file contains feature vectorization and model
hyperparameter information for deployment purposes.
-
Go to the Pega
GitHub repository, and then clone or download the
sample containers.
The repository provides two sample containers:
- machine-learning-nlp-container for deploying
machine learning models.
- deep-learning-nlp-container for deploying deep
learning models.
-
Deploy your model in the sample container:
-
Copy the model to the specified location.
-
Build a Docker image.
-
Run the container.
For instructions, see the README.md file that is provided
with the sample container.
Result: When the container is running, you can access the API at one of the
following endpoints:
- If you deployed the container using OAuth 2.0:
https://IP-address:port/auth/predict
- If you deployed the container without authentication:
http://IP-address:port/noauth/predict
where:
- IP-address is the IP address of the machine that
hosts the container.
- port is a port on the machine that hosts the
container.
-
Test your model endpoint API using an API testing tool, such as Postman, to
ensure that the model works properly.
What to do next: Configure a machine learning service to connect to the
model through an API. For more information, see Configuring a machine learning service connection for topic detection models using REST API.