NovaSearch-Team/RAG-Retrieval

Health v2 · maintenance only

Adoption not part of the score

1127 stars · 88 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

A Python library and toolkit for unified fine-tuning, inference, and distillation of RAG retrieval models, including embedding models, ColBERT-style late-interaction models, and rerankers. It also provides a lightweight pip-installable library for calling different RAG ranking models through a unified interface.

Use cases

• fine-tune an embedding model for RAG retrieval
• train a ColBERT late-interaction retriever
• fine-tune a reranker model with an LLM
• distill a large reranker into a smaller BERT model
• call different reranking models through one unified Python API
• train embedding models with MRL loss

When to choose

• you need to fine-tune or distill embedding, ColBERT, or reranker models for a RAG pipeline
• you want a unified inference interface for multiple RAG ranking model types
• you want to reproduce training of models like Stella and Jasper

When to avoid

• you only need to call a hosted embedding or reranking API without training
• you need a full end-to-end RAG framework with vector store and generation, not just retrieval model training
• you work outside Python

Facets

library · maturity active

machine-learning llm-training rag nlp search-engine machine-learning large-language-models python cross-platform embedding-models reranker colbert fine-tuning distillation information-retrieval retrieval-augmented-generation natural-language-processing search gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="NovaSearch-Team/RAG-Retrieval")