BrandFuzzysearch-LLM

Project information

BrandFuzzySearch is a multi-source search tool that tolerates spelling variants to look up registered trademarks, patents, domain names and companies (legal names, acronyms) across official registries, with geographic priority on the EU and Switzerland. It is built as a Rust cargo workspace: a shared core (bfs-core) provides an async Source trait, text normalization (including cross-script transliteration) and a hybrid fuzzy re-ranking engine, while dedicated crates wrap real data sources — EUIPO, EPO, INPI, INSEE, Swissreg, Zefix, ARES, PRH, CVR/DataFordeler and RDAP. An optional multi-agent LLM pipeline (bfs-llm, llama.cpp) enriches results with analysis grounded in cached open data. It ships as a CLI (bfs), an Axum HTTP API (bfs-api) and an Astro front end. External access is read-only by design (no writes to upstream APIs, enforced by a build-time guardrail), with encryption at rest, GDPR purge, an SBOM and supply-chain checks in CI. ⚠️ The tool produces a preliminary study only. It does not replace the explanations, research or advice of a professional or of a lawyer specialising in trademark or company registration. Licensed under MPL-2.0.