About 400bc

This is a search engine for Nepali-language web pages. It crawls the web itself, builds its own index, and analyses Devanagari text the way the Nepali language actually works.

522,144 pages from 6,314 hosts have been indexed so far, and 45,085 addresses are still waiting to be crawled.

This is a prototype

The index is small, so many searches find nothing. An empty result does not mean the page is absent from the internet — it means it has not been crawled here yet. That is why the foot of every page shows how much has been crawled.

Devanagari analysis

Queries and pages are analysed in exactly the same way — NFC normalisation, ZWNJ/ZWJ stripping, tokenising that keeps a word whole through its matras, and suffix stemming. That is why searching for नेपाल also finds a page that writes नेपालको.

How this works

• Ranking is BM25 over a Devanagari analyzer, then blended with host authority and recency: a fresh page on a well-linked site rises, an old reference page is never pushed down.
• Coverage is a curated list of hosts — an operator adds a site with npadmin, and only those sites are crawled. There are no content farms and no open submission.
• Pages are refreshed on their host's own schedule, from every fifteen minutes for a news site to fortnightly for an archive, and the index is rebuilt from the database.

Why not just use Google

A curated index with nothing AI-written in it; an analyzer built for Nepali that knows नेपालको is नेपाल; place answers assembled from cited pages with nothing composed; and no tracking of what you search.