Setup

Prerequisites

• Bun 1.2+
• Python 3.x
• nfdump available on your PATH (or via Nix — see shell.nix)
• SSH access to the research host if using live ONRG data

For Multifractal Analysis

Only needed if you want the vendor analyzer binaries:

• ghc plus the Haskell packages listed in vendor/maad/shell.nix for vendor/maad
• zig for vendor/burstify
• pkg-config and libpcap if you also want the full burstify toolchain, not just StructureFunction
• or nix-shell, then use the compile scripts under vendor/scripts/

Install

Configure Datasets

Copy the example config and edit paths for your machine:

root_path should point at the directory containing router/source subdirectories.

See datasets-json.md for more.

Environment

Start from the template:

At minimum, make sure DEFAULT_DATASET matches one of the dataset ids in datasets.json.

Optional overrides: DATASETS_CONFIG_PATH, NETFLOW_DATA_PATH, DATABASE_PATH.

Populate the Database

This discovers and processes NetFlow files into SQLite. See pipeline-usage.md for flags and scheduling patterns.

Optional: Compile Vendor Analyzers

If you need spectrum_stats / structure_stats for real captures, initialize the analyzer submodules and compile the vendor binaries:

compile-burstify.sh builds vendor/burstify/zig-out/bin/StructureFunction, which structure_stats uses. If libpcap is also available, it additionally builds the rest of burstify.

compile-maad.sh builds vendor/maad/Spectrum and related Haskell executables used by spectrum_stats.

Those binaries are not required for the no-data setup verification flow above.

Run the App

The web app starts at http://localhost:5173.

SSH Tunnel (remote host)

Then open http://localhost:5173 locally.