Skip to main content
Macroa Pulse is a Python library for proactive AI agent cognition. It runs entirely on local compute with zero LLM API calls in normal operation.

​
Requirements

Macroa Pulse requires Python 3.11 or higher. The library uses PyTorch for neural network inference and runs efficiently on CPU.

​
Installation Methods

1

Install with pip

Install the latest version from PyPI:
pip install macroa-pulse
This will install Macroa Pulse along with its core dependencies:
  • torch>=2.0 - PyTorch for LSTM/TCN models
  • watchdog>=3.0 - Filesystem event detection
  • numpy>=1.24 - Feature vector construction
2

Verify installation

Verify that Macroa Pulse is installed correctly:
from pulse import PulseRegistry, EscalationDecision, ModuleFingerprint

print("Macroa Pulse installed successfully!")

​
Alternative: Poetry

If you use Poetry for dependency management: 
poetry add macroa-pulse

​
Alternative: Pipenv

If you use Pipenv: 
pipenv install macroa-pulse

​
Development Installation

If you want to contribute to Macroa Pulse or run the test suite, install the development dependencies: 
pip install macroa-pulse[dev]
This installs additional packages:
  • pytest>=7.0 - Testing framework
  • pytest-asyncio>=0.21 - Async testing support

​
System Requirements

Macroa Pulse runs entirely on local compute. No external services, network requests, or API keys are required.

​
Operating Systems

  • Linux - Full support with inotify for filesystem events
  • macOS - Full support with FSEvents
  • Windows - Full support with ReadDirectoryChangesW

​
Hardware

  • CPU: Any modern CPU (no GPU required)
  • RAM: Minimal footprint - cluster models are ~50K-200K parameters each
  • Disk: ~10MB for the library, additional space for training data storage

​
Filesystem Watching

Macroa Pulse uses the watchdog library for filesystem event detection. This provides cross-platform support with native OS event APIs:
  • Linux: inotify
  • macOS: FSEvents
  • Windows: ReadDirectoryChangesW
Watching very large directory trees (>10,000 files) may impact performance on some systems. Configure watch_directories in your module fingerprints to monitor only relevant paths.

​
Data Storage

Macroa Pulse stores training data and model weights locally: 
~/.macroa/pulse/
├── training_data.db    # SQLite database for training examples
└── models/             # PyTorch .pt files for cluster models
All data stays on your machine. Macroa Pulse never sends data to external services.

​
Next Steps

Now that you have Macroa Pulse installed, proceed to the Quickstart guide to create your first proactive agent.

Build docs developers (and LLMs) love

Get started for freeTalk to us