# jerry-ai-dev/MODULAR-RAG-MCP-SERVER

A modular RAG (Retrieval-Augmented Generation) system with MCP Server architecture. Using Skill to make AI follow each step of the spec and complete the code 100% by AI.

Repository: https://github.com/jerry-ai-dev/MODULAR-RAG-MCP-SERVER
Canonical: https://ross.abutalabs.com/products/modular-rag-mcp-server
Language: Python
License Family: other
Last push: 2026-03-10T07:00:39+00:00

## Health v2 (maintenance only)
Score: 47/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 71, release rhythm 35, longevity 16
- inputs: {"age_days": 225, "days_push": 176, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1109, forks 253 (observed 2026-08-28T04:03:37.194729+00:00)

## What it is
A modular, pluggable RAG (Retrieval-Augmented Generation) framework exposed as an MCP (Model Context Protocol) server, so AI assistants like Copilot and Claude Desktop can query a knowledge base directly. It includes a full ingestion pipeline (PDF to embeddings), hybrid search with reranking, multimodal image captioning, Ragas-based evaluation, a Streamlit dashboard, and doubles as a hands-on learning project for LLM engineering job seekers.

## Use cases
- expose a document knowledge base to Claude Desktop or Copilot via MCP
- build a RAG pipeline over PDFs with hybrid BM25 and dense vector search
- add reranking with cross-encoders or LLM rerankers to retrieval
- make images searchable by captioning them with a vision LLM
- evaluate RAG quality with Ragas and golden test sets
- trace ingestion and query pipelines end to end
- learn RAG engineering for LLM job interviews
- swap LLM, embedding, reranker, or vector store backends via config

## When to choose
- you want a working MCP-accessible RAG server with pluggable components
- you need hybrid search plus reranking out of the box
- you want a complete, observable RAG reference implementation to study or extend
- you are preparing for LLM/RAG engineering interviews and want a portfolio project

## When to avoid
- you need a production-hardened, licensed product (no license is provided)
- you want a simple drop-in library rather than a full server with dashboard and pipelines
- you need non-Python runtimes or managed cloud deployment
- you require guaranteed long-term maintenance or enterprise support

## Facets
- artifact type: service
- maturity: active
- function: rag, search-engine, mcp, llm-inference, pdf, data-visualization, monitoring, machine-learning
- domain: large-language-models, artificial-intelligence, developer-tools, tutorials
- platform: python, self-hosted
- tags: mcp-server, hybrid-search, rerank, bm25, vector-search, image-captioning, ragas-evaluation, streamlit-dashboard, skill-driven-development, interview-preparation, pluggable-architecture, retrieval-augmented-generation, natural-language-processing, search, docker, web-server

## Member repositories
- jerry-ai-dev/MODULAR-RAG-MCP-SERVER (main) score 47

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:37.194729+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 2026-08-30T06:43:35.302939+00:00, confidence not recorded.
  - readme: https://github.com/jerry-ai-dev/MODULAR-RAG-MCP-SERVER (fetched 2026-08-28T04:03:37.194729+00:00, sha 03edc17cd11a)
- Data as of 2026-08-30T08:39:29.467469+00:00.