A Heterogeneous Benchmark for Information Retrieval. Easy to use, evaluate your models across 15+ diverse IR datasets.

Open-source inference server and production cluster for all the models your agent needs.

Efficient Retrieval Augmentation and Generation Framework

🥤 RAGLite is a Python toolkit for Retrieval-Augmented Generation (RAG) with DuckDB or PostgreSQL

Open Source Semantic Search for your AI Agent

Use late-interaction multi-modal models such as ColPali in just a few lines of code.

Late Interaction Models Training & Retrieval

Neural Search

High-Performance Engine for Multi-Vector Search

PyLate efficient inference engine

ColBERT humor dataset for the task of humor detection, containing 200,000 jokes/news

Production inference for encoder models - ColBERT, GLiNER, ColPali, embeddings etc. - as vLLM plugins for online and in-process deployment

An easy-to-use python toolkit for flexibly adapting various neural ranking models to target domain.

Vector Database with support for late interaction and token level embeddings.

A demonstration of hybrid search with reranking using Qdrant and BGE-M3 model. A showcase of dense and sparse retrieval combined with ColBERT reranking for optimal search results

This repository helps you evaluate your models on the FreshStack benchmark!

Tree-based indexes for neural-search

LEMUR reduces multi-vector retrieval for late interaction models such as ColBERT into regular single-vector retrieval.

Fused Triton kernels for late-interaction (MaxSim) scoring

An overview of popular reranking models and architectures for 2 stage RAG pipelines