domains: 3,212 crawlable domains, 39,442,105 URLs discovered

README.md

@@ -0,0 +1,86 @@
+---
+license: other
+license_name: derived-from-public-lists
+language:
+  - en
+pretty_name: Desearch crawlable domains
+tags:
+  - web-crawl
+  - domains
+  - sitemaps
+  - bittensor
+configs:
+  - config_name: domains
+    data_files: domains/domains.parquet
+---
+
+# Desearch crawlable domains
+
+The English-language domains Desearch miners crawl. Every host listed here was visited by the
+Desearch Bot: its robots.txt was read and obeyed, a sitemap was fetched and parsed, and its
+homepage was checked for readable English text. `sitemap_url` is where a crawl of that domain
+starts.
+
+`domains/domains.parquet` is rewritten as the bot works through its candidate list, so it always
+reflects the current qualified set. `domains/stats.json` carries the counts and the rejection
+breakdown for the same moment.
+
+## Columns
+
+| column | meaning |
+|---|---|
+| `host` | registrable domain, lowercase, no `www.` |
+| `rank` | best rank across Tranco, Majestic and Open PageRank; null when only the unranked sources list it |
+| `tld_group` | `big_generic` (com/net/org), `en_cctld`, `new_generic` |
+| `type_hint` | `news`, `blog` or `company` when a source category suggested it |
+| `sitemap_url` | the sitemap that parsed; where a crawl starts |
+| `sitemap_kind` | `index` when it points at more sitemaps, `urlset` when it points at pages |
+| `url_count` | URLs discovered across this domain's sitemap tree |
+| `crawl_delay` | seconds requested by robots.txt for our token, when set |
+| `language` | detected from the homepage text |
+| `checked_at` | when the bot visited |
+
+## How a domain qualifies
+
+1. **robots.txt** is fetched and parsed. A `Disallow: /` for `DesearchBot`, or for `*` when we are
+   not named, rejects the domain. A declared `Crawl-delay` is recorded and respected.
+2. **A sitemap** is taken from robots.txt, or tried at `/sitemap.xml` and `/sitemap_index.xml`. It
+   must parse, and when it lists pages directly it must hold at least ten.
+3. **The homepage** must return 200, show at least 200 characters of visible text, not be a bot
+   wall, and be detected as English.
+
+## What is excluded before any visit
+
+Adult, malware, phishing, cryptomining, stalkerware, hacking, warez, gambling, link shorteners, ad
+networks, dynamic DNS and proxy hosts, from the UT1 category lists, matched on the exact host.
+Search engines, social networks, forums, chat, webmail, file hosting, media streaming and
+marketplaces, whose pages are generated per user rather than published as documents. CDNs,
+resolvers, certificate authorities, registrars and analytics endpoints. Country-code domains
+outside English-speaking markets.
+
+## Crawling policy
+
+The bot identifies itself as
+`Mozilla/5.0 (compatible; DesearchBot/1.0; +https://www.desearch.ai/crawler)`, obeys robots.txt
+for that token, and requests at most one page per second per host unless a longer `Crawl-delay` is
+declared. It reads sitemaps and homepages only; page content is fetched by miners under the same
+rules.
+
+## Sources
+
+Candidate hosts come from Tranco, Majestic Million, Open PageRank, BuiltWith Top 1M and Cisco
+Umbrella. Categories come from the UT1 blacklists. Registrable domains follow the Public Suffix
+List. Each source carries its own terms; this dataset is derived data with attribution, and the
+ranks are each source's own and are not comparable between sources.
+
+## Reproducing
+
+The bot is open source in [subnet-22](https://github.com/Desearch-ai/subnet-22) under
+`desearch_bot/`:
+
+```bash
+python -m desearch_bot.cli candidates --out build
+python -m desearch_bot.cli load --candidates build/candidates.parquet
+python -m desearch_bot.cli discover --limit 100000
+python -m desearch_bot.cli publish --push
+```

domains/domains.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7afe977e80c769106adcfb414a600b971e9366e3ab2b302c0484866451ebbd8b
+size 96727

domains/stats.json

@@ -0,0 +1,25 @@
+{
+  "built_at": "2026-09-08T14:22:22+00:00",
+  "qualified": 3212,
+  "rejected": 5788,
+  "still_to_check": 5966916,
+  "urls_discovered": 39442105,
+  "sitemaps_tracked": 42517,
+  "qualified_by_tld_group": {
+    "big_generic": 2571,
+    "new_generic": 440,
+    "en_cctld": 201
+  },
+  "rejected_by_reason": {
+    "no_sitemap": 3019,
+    "unreachable": 1507,
+    "not_english": 416,
+    "no_homepage_text": 301,
+    "homepage_error": 185,
+    "robots_disallowed": 175,
+    "sitemap_too_small": 113,
+    "robots_unavailable": 36,
+    "homepage_unreachable": 33,
+    "bot_wall": 3
+  }
+}