From URL to answer, in five stages.
No mystery here. The pipeline is straightforward. Below: each stage in order, what happens, and what you control.
Crawl.
You give us a domain. We discover URLs from sitemaps, robots.txt, and link traversal. Headless rendering handles JavaScript-heavy pages and SPAs.
Embed.
Pages are chunked along semantic boundaries (headings, lists, code). Each chunk is embedded with a current-generation model and stored against its source URL and metadata.
Index.
Vectors live in a tuned index alongside lexical signals. Queries take both routes — semantic similarity and keyword matching — then merge results with a reranker.
Serve.
A query hits AWS Lambda in Sydney. Both semantic and lexical lookups run in parallel, then merge. Network round-trip dominates, not search.
Tune.
You see what visitors search, what they click, and what returns nothing. Boost rules and metadata schema are the knobs you turn — ranking changes when you change them, not silently in the background.
Your site has the answer.
Help them find it.
Free tier, no credit card. Cancel by removing the script tag — that's the whole thing.
Setup
One script tag
Uptime
100% since launch
Hosting
AU-sovereign
Cancel
Delete the tag