sysgraph

An interactive force-directed network graph visualizer for the browser — with two modes:

• Import any graph — load any JSON graph (nodes + edges) to explore and visualize it interactively
• Live process graph — discover running OS processes and their inter-process communication channels in real time (cross-platform; richest on Linux)

Features

• Import any graph — load a JSON file with nodes and edges to visualize any network, social graph, dependency tree, or dataset
• Export/Import — save and reload graph snapshots as JSON; use the sample at data/simplest-graph.json as a format reference
• Interactive graph — force-directed graph rendered in the browser with zoom, pan, drag, and search
• Fuzzy search — find nodes by any property
• Adjacency filtering — right-click a node to show only its neighbors
• Configurable — tune d3 force parameters, colors, and type filters via the settings panel
• Process discovery — enumerates running OS processes and their parent-child relationships (cross-platform via psutil)
• IPC visualization — discovers TCP/UDP connections (all platforms), Unix domain sockets and pipes (Linux only)
• Real-time — fetch the latest process graph on demand via the web UI

Demo

demo.mp4

Requirements

• Python ≥ 3.12
• Linux, macOS, or Windows — process and network discovery works on all platforms via psutil; Unix domain sockets and pipe discovery require Linux
• Root/sudo recommended for full process visibility on Linux/macOS

Installation

pip install sysgraph

Usage

# Start the web server (default: http://localhost:8000)
sysgraph

# Specify a custom port
sysgraph --port 9000

# Or run as a module
python -m sysgraph

Open your browser to the displayed URL.

Visualize your own graph

Use the Import button in the UI to load any JSON file in the following format:

{
  "nodes": [
    {"id": "1", "type": "person", "properties": {"name": "Alice"}},
    {"id": "2", "type": "person", "properties": {"name": "Bob"}}
  ],
  "edges": [
    {"source_id": "1", "target_id": "2", "type": "knows", "properties": {}}
  ]
}

See data/simplest-graph.json for a minimal example.

Live process graph

For full visibility into all processes and their connections, run with elevated privileges (Linux/macOS):

sudo sysgraph

Docker

docker run --rm -it --pid=host --net=host gubenkoved/sysgraph

The --pid=host and --net=host flags allow the container to see host processes and network connections.

How It Works

1. The browser frontend renders interactive force-directed graphs using force-graph with d3 physics simulation.
2. Graphs can be imported from JSON directly in the browser, or fetched live from the backend.
3. The FastAPI backend uses psutil to discover processes and network connections (cross-platform), plus Linux-specific APIs (/proc, ss) for pipe and Unix domain socket discovery, building a graph served via GET /api/graph.

Development

Prerequisites

• Python ≥ 3.12, Docker (for frontend builds)
• Node.js 22 runs inside Docker; no host installation required

Backend

python3 -m venv .venv && source .venv/bin/activate
pip install -e . && pip install -r requirements-dev.in
python src/sysgraph/app.py   # → http://localhost:8000

Frontend (TypeScript + Vite)

./scripts/build-ui.sh         # production build → src/sysgraph/dist/
./scripts/dev-ui.sh           # Vite dev server with HMR on :5173
./scripts/typecheck-ui.sh     # TypeScript type checking
./scripts/lint-ui.sh          # Biome linter (pass --fix to auto-fix)

Tests

pytest src/sysgraph/tests/

Python linting

./scripts/lint.sh             # ruff check + ruff format + isort

License

MIT