ExtractThinker - extract and classify data from documents - ORM for documents

ParadeDB - Elasticsearch-alternative that runs in Postgres

vLLM Semantic Router - Intent aware semantic router, route requests to the best model based on your defined pool (intent/complexity/cost-aware)

Docling - Deep featured document parsing with info hierarcy, content bounding boxes and header/footer disambiguation

LLM Guard - Input and output filters capabale of detecting or removing text like code, competitors, gibberish, secrets, etc

ck - CLI for semantic code and text search that blends BM25 and embeddings to quickly surface chunks you can feed into RAG pipelines

Steel - hosted browsers for AI agents - use in combination with OpenAI/Claude computer use