# HKUDS/MiniRAG

[ACL2026] "MiniRAG: Making RAG Simpler with Small and Open-Sourced Language Models"

Repository: https://github.com/HKUDS/MiniRAG
Canonical: https://ross.abutalabs.com/products/minirag
Homepage: https://arxiv.org/abs/2501.06713
Language: Python
License: MIT
License Family: permissive
Topics: large-language-models, rag, retrieval-augmented-generation
Last push: 2025-10-16T07:43:16+00:00

## Health v2 (maintenance only)
Score: 39/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 47, release rhythm 28, longevity 42
- inputs: {"age_days": 599, "days_push": 321, "days_rel": 552, "gap_med": 42, "n_releases_24m": 2}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2008, forks 257 (observed 2026-08-28T04:06:04.906847+00:00)

## What it is
MiniRAG is an extremely simple retrieval-augmented generation framework designed to work with small, open-source language models. It uses semantic-aware heterogeneous graph indexing and lightweight topology-enhanced retrieval to achieve LLM-level RAG performance with small models while using only 25% of the storage.

## Use cases
- build a RAG pipeline with small open-source language models
- run retrieval-augmented generation on resource-constrained or on-device hardware
- index documents into a heterogeneous knowledge graph for retrieval
- reduce RAG storage requirements compared to LLM-based systems
- evaluate lightweight RAG systems with the LiHua-World benchmark
- query documents with graph-based topology-enhanced retrieval

## When to choose
- you want RAG performance without relying on large proprietary LLMs
- you are deploying on-device or in resource-constrained environments
- you need lower storage footprint for your RAG system
- you prefer graph-based retrieval over pure vector search
- you want to benchmark lightweight RAG approaches

## When to avoid
- you already have access to strong large language models and want maximum retrieval quality
- you need a turnkey managed RAG service rather than a Python framework
- your data does not benefit from entity/graph-based indexing
- you require extensive production tooling and enterprise support

## Facets
- artifact type: framework
- maturity: active
- function: rag, search-engine, nlp, llm-inference, machine-learning
- domain: large-language-models, artificial-intelligence
- platform: python, self-hosted, cross-platform
- tags: small-language-models, graph-indexing, knowledge-graph, lightweight-rag, on-device, heterogeneous-graph, benchmark-dataset, retrieval-augmented-generation, natural-language-processing, search, docker

## Member repositories
- HKUDS/MiniRAG (main) score 39

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:04.906847+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:01:16.793834+00:00, confidence not recorded.
  - readme: https://github.com/HKUDS/MiniRAG (fetched 2026-08-28T04:06:04.906847+00:00, sha 30edb9add276)
  - homepage: https://arxiv.org/abs/2501.06713 (fetched 2026-08-29T10:41:15.169935+00:00, sha c9fd41a5371d)
  - site_page: https://info.arxiv.org/about/donate.html (fetched 2026-08-29T10:41:15.179658+00:00, sha cca9c3a11c56)
  - site_page: https://info.arxiv.org/about/ourmembers.html (fetched 2026-08-29T10:41:15.183982+00:00, sha 47cbc55ff1de)
  - site_page: https://info.arxiv.org/about (fetched 2026-08-29T10:41:15.186169+00:00, sha a1f16f915a9a)
  - site_page: https://info.arxiv.org/labs/index.html (fetched 2026-08-29T10:41:15.181921+00:00, sha b14a8d05a0ec)
- Data as of 2026-08-30T08:39:29.467469+00:00.