Skip to content

Welcome to Mindtrace

Unifying asset inspection under a single package, Mindtrace brings together industrial hardware, machine learning, and automation, bridging the gap between edge devices and scalable intelligence.

Monitor, automate, and scale your on-prem AI solutions effortlessly with Mindtrace package.
From training robust AI models, hardware integration to deploying and scaling edge intelligence for Asset inspection, Mindtrace enables full-cycle orchestration.

It seamlessly connects hardware, data, and machine learning, empowering teams to deploy context-aware decision systems, derive real-time insights, and visualize results through interactive dashboards

Features

  • Unified Architecture
    Integrates data ingestion, model training, deployment, and system coordination under one modular ecosystem that scales seamlessly across services and clusters.

  • Hardware-Aware Intelligence
    Connect directly to PLCs, cameras, and sensors for real-time inference, control, and closed-loop feedback across industrial environments.

  • ML-Native Design
    Provides end-to-end pipelines for dataset management, model registry, and distributed training workflows.

  • Datalake Integration
    Built-in connectors and APIs for storing, indexing, and retrieving structured or unstructured data — powering analytics, retraining, and traceability.

  • Cluster-Aware Orchestration
    Enables coordinated operation across multiple nodes or services, supporting distributed execution and horizontal scaling.

  • Service Collaboration Layer
    Seamlessly launch, register, and interconnect FastAPI or MCP-based microservices through a unified control plane and shared state system.

Layered Architecture

Mindtrace is organized into a layered workspace to support ML components as Python modules with clearly defined boundaries and dependencies. We use a level-based system for organizing modules based on dependency direction and build order.

Level 1: Core

  • core: Foundational utilities and base classes used across all other modules.

Level 2: Core Consumers

  • jobs: Job execution and backend interfaces.
  • registry: Artifact and metadata management.
  • database: Redis, Mongo, and DB access layers.
  • services: Service base classes, authentication, and gateways.
  • storage: Storage functionality for cloud storage integration.
  • ui: Optional UI libraries and components.

Level 3: Infrastructure Modules

  • hardware: Interfaces for cameras, PLCs, scanners, etc.
  • cluster: Runtime cluster management, nodes, and workers.
  • datalake: Dataset interfaces for HuggingFace and Mindtrace datasets.
  • models: Core model definitions and leaderboard utilities.

Level 4: Automation

  • automation: Integration of pipelines and orchestration using level 2–3 modules.

Level 5: Applications

  • apps: End-user applications composed of all previous levels.
  • E.g., Demo pipelines

Dependency Flow

Each layer only depends on modules in lower levels.

Module Depends On
core –
jobs core
registry core
database core, registry
services core
storage –
ui core
cluster core, jobs, registry, database, services
datalake core, registry, database, services
models core, registry, services
hardware core, services, storage
automation core, registry, database, services, datalake, models, cluster
apps core, registry, database, services, datalake, models, cluster, jobs, hardware, ui, automation

Quick Start

Installation

# Install the full Mindtrace package
uv add mindtrace

# Or install a minimal dependency chain
uv add mindtrace-datalake

Basic Usage

from mindtrace import core, registry, database, services

Contribute

We welcome contributions! Whether you're fixing bugs, adding features, or improving documentation, your help makes Mindtrace better.