Search engine
Beacon
A web search engine, from the crawler up.
A polite async crawler, a boilerplate stripper, a positional inverted index, a postings codec (delta, varint, bitpacking), BM25, PageRank over sparse matrices, a gradient-boosted learning-to-rank reranker, a query parser with spellcheck and autocomplete, and a sharded fan-out/merge layer.
beacon-search-console wires all nine together through their documented on-disk contracts and serves 181 pages crawled live from docs.python.org. beacon-search-engine-at-scale then rebuilds the substrate — object storage, a message queue, a service registry, map-reduce indexing — for a few million pages. No Elasticsearch, no Lucene.
- Repositories
- 11
- Lines
- 26,158
- Commits
- 223
- Test files
- 132
- Built
- Jul 2026
- Mostly
- Python
L1 Ingestion
L2 Indexing
L3 Ranking & relevance
L4 Query serving
↳ Flagship
↳ Flagship
Where it all comes together
L1
Ingestion
L2
Indexing
L3