A small toolkit to discover Nostr relays and estimate their locations. It can also optionally filter for BitChat-capable relays (kind 20000) before geolocating.

The repository contains three scripts that form a simple pipeline:

1. Discover functioning relays (Python)

• Script: nostr_relay_discovery.py
• What it does:
  • Starts from a seed relay URL and connects to it via WebSocket (ws/wss).
  • Breadth‑first searches follow lists (kind 3) and relay lists (kind 10002), extracting relay URLs from event tags (r/p tags).
  • Tests each candidate relay for basic Nostr protocol responsiveness by issuing a REQ and evaluating responses (EVENT/EOSE/NOTICE).
  • Processes relays in concurrent batches and periodically saves progress to relay_discovery_results.json.
• Output: relay_discovery_results.json with:
  • functioning_relays (array of ws/wss relay URLs)
  • discovery settings, progress, and statistics

2. Filter for BitChat relays (kind 20000)

• Script: filter_bitchat_relays.sh
• What it does:
  • Reads relay URLs from stdin and, in parallel, checks whether each relay:
    • can return kind 20000 events (read), and
    • accepts posting a kind 20000 event (write).
  • Uses the nak CLI to query (nak req -k 20000) and post (nak event -k 20000 ...).
• Output: writes to stdout the subset of input relay URLs that pass both checks. You can redirect this to a file, e.g., bitchat_relays.txt.

3. Geolocate relays (Bash)

• Script: relays_geo_lookup.sh <output.csv>
• What it does:
  • Downloads the DB‑IP city IPv4 ranges (dbip-city-ipv4-num.csv.gz), loads it into memory, and uses a binary search to map IPv4s to latitude/longitude.
  • Reads hostnames from stdin, resolves them via dig to A records, and looks up the first IPv4 with available coordinates.
• Input: hostnames only (e.g., relay.example.com), not full ws/wss URLs.
• Output: writes a CSV with header Relay URL,Latitude,Longitude to the file you specify.

Note: The geolocation step is IPv4‑only (the DB‑IP file used here is IPv4). IPv6‑only relays will be skipped.

• The discovery step uses timeouts and concurrent batches; tune --batch-size, --timeout, and --max-depth for your environment.
• The BitChat filter posts a test kind 20000 event. Ensure nak is configured to publish (e.g., via its config or environment variables as per nak docs).
• Geolocation is an estimate based on DB‑IP and only for IPv4. Accuracy varies and may reflect the ISP/hosting POP rather than precise server location.
• relays_geo_lookup.sh stops at the first IPv4 for which DB‑IP provides coordinates. Some hostnames resolve to multiple IPs.

This repository includes a GitHub Actions workflow under .github/workflows/ that automates the end‑to‑end process:

• Checks out the repository
• Sets up Python and system dependencies
• Runs nostr_relay_discovery.py to produce relay_discovery_results.json
• Extracts the functioning relays with jq
• Runs filter_bitchat_relays.sh
• Runs relays_geo_lookup.sh to generate nostr_relays.csv
• Commits the updated artifacts back to the repository

To change the schedule, seed relay, or enable BitChat filtering in CI, edit the workflow file in .github/workflows/.

These charts show the number of Nostr relay entries in our dataset over time:

The chart below shows relays that support BitChat events (kind 20000):

The chart below shows all functioning relays discovered during the relay discovery process:

The charts are automatically updated daily to reflect changes in the number of relays and show approximately 70 days of history.

The map below shows the geographical distribution of BitChat-compatible Nostr relays around the world:

Additional visualizations available in this repository:

• Heatmap: A density visualization showing relay concentration areas (assets/relay_locations_heatmap.png)
• Interactive Map: An HTML-based interactive map that allows zooming and clicking on individual relays (assets/relay_locations_interactive.html) - download and open in a browser to explore

All maps are automatically updated alongside the relay data.

nostr_relays.csv and relay_discovery_results.json use a database curated by DB‑IP, available at https://www.db-ip.com.