simlar on BEIR: Hybrid Retrieval Across 14 Datasets
A BEIR benchmark across 14 datasets: how simlar's hybrid retrieval stacks up against brute-force, FAISS, and turbovec dense search — on ranking quality and, especially, on latency at scale.
Three Search Problems, One Library: sim_LAR by Example
A support knowledge base, a product catalog, and a 120,000-document streaming archive — three different shapes of search problem, one pip install. Here's what each one actually took.
Inside the sim_LAR RAG Retrieval Benchmark Notebook
The full notebook, code and all: how sim_LAR compares to Chroma, Qdrant, FAISS, and turbovec on Hit@1, Hit@10, and MRR against a 500,000-passage MS MARCO slice.
Installing and Using sim_LAR: A Practical Guide
sim_LAR is TekDatum's search library for Python — keyword search, semantic search, and hybrid search behind one simple API. Here's how to install it and start searching in minutes.
Building sim_LAR: Why Similarity Search Needs to Be Exact, Not Approximate, in the RAG Era
Most vector databases only ask one question: what's close enough? We built sim_LAR to ask a better one: did every relevant piece of information actually show up?