OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
RAG & Knowledge Systems

Memorag

MemoRAG is an Apache-2.0 Python framework for retrieval-augmented generation. It builds a global memory representation of a long input, recalls clues relevant to a query, and uses those clues to guide evidence retrieval and response generation.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Memorag — OpenViST project profile in RAG & Knowledge Systems.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use MemoRAG when questions require clues from across a long context or document collection, especially when the information need is not fully explicit in the query.

Who it is for

  • Developers building retrieval-augmented generation applications over long documents or large collections.
  • Research teams creating corpus question-answering or document exploration tools.

Documented uses

  • Build a memory representation of a long context.
  • Recall query-specific clues from that context to guide evidence retrieval.
  • Generate responses using retrieved evidence.
Before you choose

Requirements & limits

The repository offers installation from source or through pip, and documents a Google Colab quick-start.

  • The pipeline separates memory, retrieval and generation models; the generation model is optional in the documented example.
  • The README provides model-specific context capacities, which vary by model and beacon-ratio settings.
  • The documented local setup includes Python, PyTorch 2.3.1 and GPU-enabled FAISS 1.8.0.
  • MemoRAG Lite's documented language support is English and Chinese; other languages may have lower performance.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Set up MemoRAG Lite for a context and query

Install the documented dependencies and use the Lite pipeline to memorize a text context and submit a question.

Before you start

  • Python
  • PyTorch 2.3.1
  • GPU-enabled FAISS 1.8.0
  • A GPU; the README recommends 24 GiB, and says 16 GiB can work with default settings in most cases

Steps

  1. Install PyTorch with the command below.

    Documented command
    pip install torch==2.3.1

  2. Install GPU FAISS with conda install -c pytorch -c nvidia faiss-gpu=1.8.0.

  3. Install MemoRAG from the repository with the command below, or install the package with the command below.

    Documented command
    pip install -e .
    Documented command
    pip install memorag

  4. Follow the README's Lite example: create MemoRAGLite(), pass a text context to memorize(), then submit a query to the pipeline.

Expected outcome

A Lite pipeline that has memorized a supplied context and can return a response to a query, following the README example.

Guide limitations

  • The README recommends a 24 GiB GPU; it says 16 GiB can work under default settings in most cases.
  • Lite mode is described for English and Chinese; other languages may perform worse because its default prompts are in English.
Documentation 1 ↗
COMPARE OPTIONS

Compare options for Retrieval-augmented applications

These projects document a common task and share a project role. This is a source-based comparison, not an execution benchmark or a guarantee of interchangeability.

ProjectDocumented purposePlatformAccess and requirements
MemoragBuild a memory representation of a long context.CloudThe repository offers installation from source or through pip, and documents a Google Colab quick-start.
LangsearchCrawl website or local data for a knowledge base.CLI, Self-hosted, Server
Hybrid Search RagCombine document retrieval and language-model generation in a configurable workflow.
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
14 October 2026
Original website title
qhjqhj00/MemoRAG · Official homepage title · retained in its source language
How we review sourcesSuggest a correction