PacketLoss

The pipeline is four distinct steps, run in order:

make atlas    # sync config -> RIPE Atlas measurements (ensure + reconcile probes)
make data     # download the rolling window of results -> local cache (data/json/)
make web      # build the static site from the cache -> web/dist
make publish  # upload web/dist to bunny.net (bunny CLI)

Data

All data is real RIPE Atlas and the builder keeps no local state (RIPE retains the measurement history). Each hourly make run:

1. resolves live probe coverage per provider (gates out networks with < min_probes);
2. fetches the rolling window from RIPE, attributes each result to its (country, provider) by the probe's country + ASN, scores, and exports data/json/.

Measurement creation needs a key so its run manually

Config & env

• config/thresholds.yaml — scoring window, loss weight K, min probes, status cutoffs.
• config/<country>.yaml — providers (curated from bgp.tools cone rank) + targets. Each target is a pinned address: (IP) or a host: (resolved per-probe at runtime). DE is wired so far.
• .env (loaded via dotenv, gitignored) — RIPE_ATLAS_API_KEY. See .env.example.