Skip to content

Introduction

Getting started with Practicus AI is straightforward.

Typical Practicus AI usage

The following steps outline a typical scenario for users who write code:

  1. Log in to your chosen region (e.g., https://practicus.your-company.com).
  2. Create one or more workers with the desired features and resource capacities.
  3. Start an IDE, such as JupyterLab or VS Code, within a worker.
  4. Develop models, applications, and process data as usual.
  5. Deploy models, applications, or use add-ons (e.g., create Airflow workflows).
  6. Observe metrics, logs, events, errors. Create alerts.

Common usage

Leveraging Documentation and Developer Tooling

1. Access the SDK Documentation:
Experienced coders understand that having immediate access to detailed SDK references accelerates the development lifecycle. You can refer to the Practicus AI SDK Documentation to understand package structures, classes, methods, and parameters. This robust, searchable reference ensures you can quickly find the API calls needed to interact with Practicus AI resources programmatically.

2. Utilize IntelliSense and Contextual Help in JupyterLab or VS Code:
When working within JupyterLab or VS Code, take advantage of built-in IntelliSense (auto-completion) capabilities. As you type, your IDE will surface method signatures, docstrings, and parameter hints—especially helpful for complex ML pipelines or when invoking intricate model-serving APIs.

Contextual Help with Jupyter Lab

  • Contextual Tooltips: Hover over classes and methods to see in-line docstrings and parameter descriptions. This “just-in-time” help enables you to craft pipelines, preprocess data, or orchestrate model inference steps without constantly switching between your IDE and external docs.

  • Shift-Tab: Inside a Jupyter notebook, pressing Shift+Tab while your cursor is within a function call will reveal type hints, default values, and docstrings. This immediate feedback reduces trial-and-error and makes coding more efficient and error-free.

Jupyter help

Combining direct SDK reference materials with IntelliSense-driven guidance ensures data scientists spend more time crafting robust models and less time hunting down syntax or function definitions.

  • Contextual Help Tab: You can also right-click on a cell, select "Show Contextual Help" to leave the help tab always open.

Jupyter help

Contextual Help with VS Code

  • Ctrl+Space for Inline Help: In VS Code, pressing Ctrl+Space triggers IntelliSense to display inline suggestions, completion items, and parameter hints. This built-in guidance makes it easy to discover available functions, understand their expected parameters, and review docstrings—all without leaving your editor window.

VS Code help

  • Jupyter Panel: You can also keep the Jupyter panel open in VS Code for continuous, context-sensitive help as you work. This panel remains visible as you code, providing an always-on reference for classes, methods, and type hints.

VS Code help

Practicus AI Platform Components

Below are the primary components you will interact with when using Practicus AI.

Practicus AI Workers

Practicus AI Workers are dedicated compute environments that run ML, data processing, and other tasks.

Key characteristics include:

  • On-demand: Request as many workers as you need, available within seconds.
  • Interactive: Launch JupyterLab or VS Code for hands-on experimentation.
  • Batch-capable: Run tasks or jobs in non-interactive mode as well.
  • Isolated: Issues in other workers or systems do not affect your worker.
  • Configurable: Each worker is defined by a container image, which can be chosen from the provided options or customized.
  • Flexible Resources: Assign a specific amount of CPU, memory, and GPU resources.
  • Ephemeral: Workers can be replaced easily. Since each restart resets the file system, save important files in ~/my or ~/shared to preserve them, or push to a source control system such as git.

Practicus AI ModelHost

Practicus AI ModelHost deployments run classic ML and LLM models, optimized for CPUs and GPUs.

  • Shared deployments can host thousands of models, each with up to 100 versions.
  • Isolated deployments allow you to create a dedicated environment for a set of models.

Practicus AI AppHost

Practicus AI AppHost deployments are used to build visual Gen AI applications or microservices focused on ML workflows.

Practicus AI Add-ons

Practicus AI Add-ons, such as Airflow or MLflow, extend the platform’s core functionality. They integrate seamlessly, allowing you to manage and orchestrate complex workflows and track experiments.

Practicus AI Regions

Practicus AI is a multi-region environment, where each region is a separate deployment and isolated Kubernetes namespace. Regions can differ by geography, cloud vendor, lifecycle stage, department, or security requirements.

For example, you might have:

  • One region in a certain geographic location and another in a different one.
  • Regions across different cloud vendors (e.g., AWS, Azure, on-premises).
  • Separate regions for production, development, or testing.
  • Regions dedicated to different departments or security contexts.

Practicus AI Clients

Practicus AI clients enable you to connect to multiple regions seamlessly, allowing you to develop in one region and deploy in another. Common client options include:

  • Browser: Access the platform via a standard web interface to launch JupyterLab, VS Code, and manage workloads.
  • AI Studio: A desktop application for Windows, macOS, and Linux that connects to multiple regions for unified management.
  • SDK: Install the SDK (pip install practicuscore) to interact programmatically with any Practicus AI region.
  • CLI: With the SDK installed, use the prtcli command-line tool to manage tasks and resources.

Example: A Multi-Region Setup

Below is an example of a deployment where a customer utilizes three regions in two geographies, accessible through various clients.

Sample Regions

Next: Workers