About WOXA

WOXA is a web search engine that shows you real pages, ranked by rules you can inspect. It does not generate answers, does not keep your searches, and works in Thai and English.

What WOXA is

WOXA is pure search. You type a query and get a list of real web pages, posts, papers and repositories. There are no AI answers and no generated summaries above the results, so what you read is what the original authors wrote.

Ranking is deterministic: the same query, settings and index give the same order. Every result has a Why this result panel that shows the scores behind its position.

Where results come from

WOXA has its own index of Thai and English pages, collected by a polite crawler that follows robots.txt. Thai text is split into words properly, so Thai searches match the way people actually write.

Each search can also draw on open sources at the same time, asked by our server rather than your browser:

  • General web and reference: DuckDuckGo, Wikipedia (English and Thai) and the Internet Archive.
  • Technology and research: Hacker News, GitHub, arXiv and Stack Exchange.
  • Markets: CoinGecko and DexScreener.
  • Communities and the decentralized web: Reddit, Lemmy, Mastodon, Nostr and IPFS.

How results are ranked

Every search goes through the same steps, in this order:

  1. Read the query. Operators such as site: or intitle: and any !bang are recognised first.
  2. Search our own index with BM25F, a well-known relevance formula that gives words in the title and address more weight than words deep in the page.
  3. Ask the open sources in parallel and collect what they return.
  4. Merge all the lists with Reciprocal Rank Fusion, which rewards pages that several sources agree on rather than trusting any single one.
  5. Apply the anti-SEO-spam score and the active lens.
  6. Add Local Citation Rank: a small lift, at most 20%, for sites that other pages in our index link to.
  7. Apply your own site rules last, so pinned, boosted, lowered or blocked sites behave exactly as you asked.

Keeping SEO spam down

Pages written to game search engines tend to repeat the same phrases and carry heavy tracking. WOXA measures how varied a page's text is (its entropy), so keyword-stuffed pages score lower, and it strips tracking parameters from links and penalises pages loaded with trackers. Some lenses apply this more strongly than others.

Lenses

A lens changes which sources and sites count most, without hiding the rest of the web. There are six built in, and you can create your own.

  • All web: everything, ranked evenly.
  • Deep tech: repositories, papers, standards and engineering discussion.
  • Finance & markets: market data, filings and on-chain metrics.
  • Indie web: personal blogs, small forums and independent feeds.
  • Decentralized: Nostr, the Fediverse and IPFS.
  • OSINT & security: DNS, vulnerabilities, certificates and archives.

Local Citation Rank

Links are a useful signal of trust, but only when you can see where they come from. WOXA counts links between sites inside its own index and works out a domain-level authority score from them, ignoring links a site makes to itself. That score can lift a result by at most 20%, so relevance always comes first.

No AI, on purpose

WOXA does not use language models to write, summarise or filter results. Instant answers such as the calculator, unit conversion, hashes, DNS records and price tickers are computed by ordinary, predictable code.

Who WOXA is for

  • Anyone who wants to search without being tracked or asked to sign in.
  • Thai readers who want a search engine that handles Thai text well.
  • Engineers, researchers, journalists and security analysts who need to see why a result ranks where it does.
  • People who prefer reading original sources over generated summaries.
  • Organisations that want to run their own private search with Docker.

Learn more

See exactly what we keep in the privacy statement, or learn the operators and shortcuts in Help.

WOXA privacy statementWOXA help