mmore Documentation

mmore is an open-source multimodal ingestion and retrieval framework designed for heterogeneous document collections.

It provides tools to process documents, build indexes, run retrieval pipelines, support multimodal workflows, and run distributed processing for larger collections and production-oriented settings.

What is mmore?

mmore helps you build retrieval systems over complex document collections by combining:

• document ingestion and processing

• indexing pipelines

• retrieval and RAG workflows

• multimodal retrieval support

• distributed processing support for large-scale document ingestion

• evaluation and profiling tools

Where to start

Depending on what you want to do, start in different places:

• to install mmore, read Installation

• to run a first workflow, read Quickstart

• to understand the overall system, read Architecture

• to understand ingestion and indexing, read Process and Indexing

• to work on retrieval workflows, read RAG

• to work on multimodal retrieval, read ColVision

• to run distributed processing, read Distributed processing

• to contribute to the codebase, read For developers

Documentation map

Getting started

• 📦 Installation
• 🚀 Quickstart
• 🏗️ Architecture
• ⚙️ Process
• 🗂️ Indexing
• 🤖 RAG
• 🪟 Running mmore on Windows

Core features

• 🖼️ ColVision Integration
• 🌐 WebSearch Integration
• 📊 RAG Evaluation Pipeline
• LLM as a judge

Advanced usage

• ⚡ uv
• 🌐 Distributed Document Processing
• ⏱️ Profiler
• 🚀 Cluster and Production Deployment

Developer documentation

• 💻 Developer Documentation
• 🔌 Indexer API Documentation

Page guide

Here is a quick overview of the main pages:

• Installation: set up mmore and prepare your environment

• Running on Windows: what differs on Windows and how to fix it

• Quickstart: run a first minimal workflow end to end

• Architecture: understand the main system components and how they interact

• Processing pipeline: understand how documents are ingested and transformed

• Indexing: build and manage indexes

• RAG: structure retrieval-augmented generation workflows

• ColVision: multimodal retrieval-related documentation

• Websearch: web search integration and related workflows

• Evaluation: assess system performance

• LLM as a judge: corrective retrieval with an LLM judge

• Distributed processing: scale processing across larger workloads

• Profiler: profile and analyze performance

• uv: environment and dependency workflow

• Cluster and production: deployment and production-oriented guidance

• For developers: contributor and internal development documentation

• Index API: API-oriented reference for indexing-related functionality