Documentation

Everything you need to install, configure, and operate MemoryMap on NVIDIA Jetson.

Install Guide

MemoryMap runs on any NVIDIA Jetson device with JetPack 5.x or 6.x. Zero configuration required.

Prerequisites

• NVIDIA Jetson (Orin, Xavier, or Nano series)
• JetPack 5.1+ or 6.0+
• Python 3.8+
• Root or sudo access (for /proc and /sys reads)

Quick Install

# Clone the repository
git clone https://github.com/ResilientMindAI/MemoryMap.git
cd MemoryMap

# Install dependencies
pip install -r requirements.txt

# Launch the dashboard
python -m memorymap serve

Verify Installation

# Check version
memorymap --version
# Expected: memorymap v0.1.0

# Run health check
memorymap health
# Expected: All checks passed

CLI Commands

MemoryMap provides a simple CLI for launching the dashboard, exporting snapshots, and querying memory state.

Command	Description
memorymap serve	Start the web dashboard on :8050
memorymap serve --port 9090	Start on a custom port
memorymap snapshot	Export current memory state to JSON
memorymap snapshot --format csv	Export as CSV
memorymap status	Print summary to terminal
memorymap health	Run system health checks
memorymap --version	Print version

Web Dashboard

The web dashboard renders six panels in real time, updating every second by default.

Panels

1. Memory Heatmap — 256MB cell grid showing CPU (blue) vs GPU (amber) allocation
2. Process Breakdown — Top 10 processes by memory consumption
3. CPU Grid — Per-core utilization bars for all 12 cores
4. Thermal Zones — CPU / GPU / SoC temperatures with color coding
5. Power Rails — Total + CPU + GPU power in mW/W
6. Timeline — 5-minute rolling chart of CPU Used, GPU Used, Total Used

Configuration

# Environment variables (all optional)
MEMORYMAP_PORT=8050         # Dashboard port
MEMORYMAP_INTERVAL=1000     # Update interval in ms
MEMORYMAP_CELL_SIZE=256     # Cell size in MB for heatmap
MEMORYMAP_HISTORY=300       # Timeline history in seconds

Export / Snapshot Format

Snapshots capture the complete system state at a point in time.

{
  "timestamp": "2026-04-07T12:00:00Z",
  "version": "0.1.0",
  "device": "Jetson AGX Orin 64GB",
  "memory": {
    "total_mb": 65536,
    "cpu_used_mb": 8192,
    "gpu_used_mb": 2048,
    "free_mb": 55296,
    "cells": [ ... ]
  },
  "processes": [ ... ],
  "cpu_cores": [ ... ],
  "thermals": { "cpu": 42.5, "gpu": 39.0, "soc": 41.0 },
  "power": { "total_mw": 15200, "cpu_mw": 5800, "gpu_mw": 4200 }
}

jtop Integration

MemoryMap uses jtop (from the jetson-stats package) as the primary data source when available.

# Install jetson-stats (if not already present)
sudo pip3 install jetson-stats

# MemoryMap auto-detects jtop
memorymap health
# Expected: jtop: available (v4.x.x)

When jtop is available, MemoryMap uses its Python API for low-overhead data collection. All six panels benefit from jtop's unified telemetry stream.

tegrastats Fallback

If jtop is not installed, MemoryMap falls back to parsing tegrastats output directly.

# tegrastats is included with JetPack
# MemoryMap spawns it automatically
sudo tegrastats --interval 1000

The tegrastats fallback parses memory, CPU, GPU, thermal, and power data from the tegrastats text stream. While functional, the jtop integration is recommended for lower overhead and richer data.

Data Source	Overhead	Recommended
jtop (jetson-stats)	Low (~0.5% CPU)	Yes
tegrastats (JetPack)	Moderate (~1.5% CPU)	Fallback