Guides
WebSiteHomeAcademy

AI & Machine Learning

Suggest Edits

AI & Machine Learning Overview

The Reasoning Flows layer dedicated to intelligent systems, model training, and generative AI.

🧭 Purpose

The AI & Machine Learning object types empower users to build, deploy, and leverage intelligent systems directly within Reasoning Flows.
These tools support a wide range of use cases — from generating embeddings and training models to real-time inference and custom ML workflows using high-performance compute environments.

With both GUI-based and code-based development options, teams can move seamlessly from data preparation to advanced analytics and generative reasoning, even across GPU-backed pipelines.

🔹 Where It Fits in Reasoning Flows

In the Reasoning Flows architecture:

  1. Extract & Load brings data into the platform.
  2. Repository Tables register datasets for reuse.
  3. Transform & Prepare cleans and structures them for modeling.
  4. AI & Machine Learning builds, trains, and deploys intelligent models.
  5. Knowledge Atlas integrates model outputs into Arpia’s Generative AI and Semantic reasoning systems.

Goal: This stage turns structured and prepared data into intelligent models, predictions, and AI-driven applications.

🧩 Development Environments

AutoML

This object type provides access to the Machine Learning and AI tools available in Arpia’s Reasoning Flows.
They support a wide variety of ML and NLP functions — from generating embeddings to training predictive models.

AutoML

  • AP AutoML Engine
    GUI-based tool that lets you train ML models on structured datasets with zero-code setup.
    Ideal for regression, classification, and basic time-series forecasting tasks.

  • SingularAI Text Embeddings
    Generates semantic vector embeddings from raw text, enabling semantic search, classification, and clustering.
    Commonly used in NLP and Retrieval-Augmented Generation (RAG) workflows.

  • AP Generative AI Workflow
    Framework for integrating LLM-based components within Reasoning Flows.
    Useful for summarization, chat assistants, or content generation pipelines.

AutoML GPU

This object type provides access to GPU-backed compute environments, ideal for large-scale or deep learning tasks.

  • AP AutoML GPU Engine
    GPU-accelerated version of the standard AutoML Engine.
    Designed for big data, image processing, and deep learning workloads where performance and parallelization are critical.

Extract

These objects automate extraction from MySQL-compatible databases registered as Data Sources in Reasoning Flows.
They can move data directly between tables or execute custom SQL-based retrieval logic.

Extract

  • AP DataPipe Engine - MySQL
    GUI-based builder for extracting data from MySQL sources and loading it into target tables.

  • Python 3.12 DataPipe Engine
    Script-based DataPipe for developers who need advanced logic or conditional data handling in Python.

High Performance Computing

These environments offer full programming flexibility for compute-intensive or custom ML pipelines.
They support Python, R, Node.js, and other languages for advanced analytics, model deployment, and integrations.

High Performance Computing

  • PHP 7.4 Application (Legacy)
    Legacy support for procedural logic and lightweight backend integrations. (EOL — maintained for compatibility only.)

  • Python 3.12 Advanced ML Application
    Flexible Python environment for training, testing, and deploying custom ML models.

  • Python 3.12 Advanced ML & Plotly
    Same as above but includes Plotly for interactive data visualization and exploratory analysis.

  • R Environment (v4.x)
    Optimized for statistical modeling and analytical reporting.

  • Python 3.12 Google Cloud Speech
    Preconfigured for speech-to-text workflows using Google’s Cloud Speech API.

  • Node.js 16 Application
    Full Node.js runtime for real-time APIs, integrations, or AI-driven web services.

Image Analyzer

This object enables image categorization and analysis using pretrained AutoML vision models.
It references image URLs stored in Repository Tables within Reasoning Flows.

  • AP AutoML Vision Predictor
    GUI-driven object that classifies images based on content.
    Ideal for tagging, validation, or computer-vision automation.

Notification Engine

The Notification Engine allows users to configure email notifications and alerts directly within Reasoning Flows.
Requires a Mailgun API key.

  • AP Notification Engine
    GUI-based interface for defining email triggers, templates, and delivery conditions.
    API keys should always be stored securely using the platform’s secrets manager.

Prepare, Transform & Inference Tools

These objects refine and transform data before and after model training, bridging the gap between data preparation, modeling, and inference.

  • AP Prepared Table
    GUI tool that converts a table into a modifiable, analysis-ready dataset.

  • AP Transform String to Binary
    Converts categorical or binary text values (e.g., Gender M/F → 0/1).

  • AP Transform String to Numeric
    Maps string categories to numeric codes (e.g., Category A/B/C → 1/2/3).

  • AP Transform Dates to Numeric
    Converts date fields into numeric features like day-of-week, Unix time, etc.

  • AP SQL Code Execution
    Enables execution of advanced SQL logic for preprocessing or feature engineering.

  • AP Model Render
    Executes trained models for real-time or batch inference, writing results back into tables.

  • SingularAI Text Splitter
    Splits long-form text into manageable chunks for LLM, semantic search, or embedding generation workflows.

Web-Hook Sender

Enables real-time communication between Reasoning Flows and external systems using webhooks.

  • AP Web-Hook Sender
    GUI-based configuration for defining outgoing payloads, target endpoints, and trigger events.
    Supports HTTPS and authentication headers for secure data exchange.

🧠 Best Practices

  • Use AutoML Engines for rapid model development and deployment.
  • Tag datasets and outputs according to their Arpia Data Layer (CLEAN → GOLD → OPTIMIZED).
  • Use GPU engines for compute-intensive training or real-time inference.
  • Document model outputs and performance metrics in the Knowledge Atlas for governance and traceability.
  • Always secure API keys, credentials, and webhooks through the platform’s Secrets Manager.

🔗 Related Documentation

Updated 18 days ago

What’s Next