Show HN: A 34 MB search engine that can't make a network call
Korvo is a 34 MB fully offline search engine that runs entirely in the browser with no network calls, posted to Hacker News by developer akshay_bhardwaj.
Korvo, a self-contained search engine weighing 34 MB, launched on Hacker News on August 25, 2026, posted under the "Show HN" format by developer akshay_bhardwaj. The project's defining constraint is architectural: once loaded, it cannot make any network calls, meaning all indexing, querying, and result rendering happens entirely within the browser.
The product is available at korvo.xyz/embed, per the Hacker News post by akshay_bhardwaj. The "embed" path in the URL suggests the engine is designed to be dropped into existing sites or applications as a standalone component, though the source does not elaborate on specific integration methods or supported data formats.
The 34 MB footprint packages everything a search engine requires — index structures, ranking logic, and query processing — into a single bundle that runs client-side. Fully offline search tools of this kind typically rely on WebAssembly or compiled JavaScript to achieve acceptable query performance without a server round-trip; the source does not specify the underlying technology stack.
The offline-only constraint has practical implications for privacy-sensitive deployments, air-gapped environments, and embedded documentation search where a backend is unavailable or undesirable. Comparable open-source projects in the client-side search space include Pagefind and Lunr.js, though neither is cited by the source and the relative size or performance benchmarks between Korvo and those tools are not addressed in the available material.
The Hacker News listing had two points and no comments at the time of publication. No company name, funding, team size, or prior product history is mentioned in the source material. Forward-looking details — pricing, a roadmap, or a broader product suite — are absent from the available information.
No comments yet — start the thread.