mindtrace 0.7.0

Mindtrace monorepo with modular packages

Mindtrace Module Dependency Structure

Mindtrace is organized into a layered workspace to support ML components as Python modules with clearly defined boundaries and dependencies.

---

📐 Layered Architecture

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.
• 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, services
registry	core
database	core
services	core
ui	core
cluster	jobs, registry, database, services
datalake	registry, database, services
models	registry, services
hardware	core
automation	jobs, registry, database, services, datalake, models, cluster
apps	Everything

---

🛠️ Build

Building wheels and source distributions, from the root of the repo:

uv build --all-packages
ls dist/

For building only wheels:

uv build --all-packages --wheel
ls dist/

They may then be installed in a new venv (the entire mindtrace package or any submodule mindtrace-core) via:

uv pip install mindtrace --find-links /path/to/dist
# or
uv pip install /path/to/dist/mindtrace.whl

Note: You may need to use uv pip install --force-reinstall in case you encounter ModuleNotFoundError.
Checking the installation:

uv run python -c "from mindtrace.core import Mindtrace; print('OK')"

🛠️ Usage Examples

Installing the full Mindtrace package:

uv add mindtrace

Installing a minimal dependency chain (e.g., for Datalake development):

uv add mindtrace-datalake

Python Imports

from mindtrace import core, registry, database, services