md-experiments/elastic_transformers

Combines Elasticsearch's dense vector search with sentence-transformers to enable hybrid lexical and semantic search on large document collections. The `ElasticTransformers` class abstracts index creation, bulk embedding, and querying workflows—handling chunked CSV ingestion with configurable dense field dimensions and supporting both keyword ('match') and embedding-based dense search. Integrates BERT-based contextual embeddings directly into Elasticsearch indices for near state-of-the-art semantic retrieval at scale.

161 stars. No commits in the last 6 months.