Skip to main content

Abacus.AI Developer Platform Overview

The Abacus.AI Platform empowers you to build, train, and deploy sophisticated AI applications—from custom Chatbots and AI agents to production-grade machine learning models—all without extensive infrastructure setup.

Get started at: abacus.ai

What You Can Build

The platform enables you to create:

  • Custom ChatBots and AI Agents for conversational AI and workflow automation
  • Machine Learning Models including time series forecasting, classification, anomaly detection, and more

Deploy these solutions by:

Core Platform Concepts

Abacus.AI offers multiple project types, each designed to create specific AI solutions. While the available features vary by project type, the fundamental building blocks remain consistent across the platform.

Below is an overview of these core concepts.

ComponentDescription
DatasetsRaw data imported directly from connectors or uploaded by users
Feature GroupsTransformed datasets created using SQL or Python, including combinations of multiple datasets
NotebooksWeb-based Jupyter notebooks providing an IDE for rapid development and experimentation
Document RetrieversAbacus.AI's native vector store, used exclusively for Generative AI applications
ModelsTrained models within a project; multiple models with different configurations can coexist
DeploymentsProduction-ready models accessible via API endpoints
Batch PredictionsBulk inference on datasets using trained models (primarily for ML use cases)
MonitorsData drift and quality monitoring tools (primarily for ML use cases)
PipelinesFlexible orchestration workflows for automating data refreshes, model training, and custom tasks

Project-Specific Requirements

While these core concepts are universal, each project type has specific Feature Group Type requirements. For instance:

  • A Demand Forecasting project requires Forecasting Timeseries Data
  • A Custom ChatBot project requires a List of Documents

You cannot train a time series forecasting model using PDF files, just as you cannot build a document-based ChatBot with Timeseries data.

The following sections explore different project types and their specific requirements in detail.