About Rag Demo
Rag Demo is an MCP server published by juichiache in the RAG category: simple rag with fastMCP, azure ai search, azure openai.
The repository has 1 stars and 0 forks, with the last commit a year ago. Six months or more without a commit doesn't mean the server is broken, but check the open issues before depending on it in production.
Install
git clone https://github.com/juichiache/mcp-rag-demoThis server has no ConduID identity, so agent calls to it are not receipted. Pin the version you install and review the source before granting it credentials.
Ask AI
Ask AI about Rag Demo
Powered by Claude · Grounded in docs
Security checks
- ✓README presentThe repository ships documentation.
5 checks have not been run yet.
README
RAG Demo with Azure MCP
This repository demonstrates a simple Retrieval-Augmented Generation (RAG) workflow using Model Context Protocol (MCP) for Azure search and Azure OpenAI for generation.
Summary
This demo application:
- Spins up a simple SSE-based MCP search server (
mcp_server.py) that wraps your document index. - Provides a CLI client script (
demo.py) that:- Connects over SSE to the MCP server
- Sends a user query
- Retrieves the top N matching document chunks
- Feeds those chunks as context to Azure OpenAI
- Prints out a grounded, concise answer (or “I don’t know.”)
Getting Started
Prerequisites
- Python 3.8+
- Azure subscription with Azure AI Search and Azure OpenAI resources
- A working Azure AI Search index (text or hybrid with integrated vectorization configured)
- Environment variables
Setup
-
Clone the repo
git clone https://github.com/juichiache/mcp-rag-demo -
Install dependencies
pip install -r requirements.txt
Start MCP Server with environment variables loaded
In a new terminal, run mcp_server script
python mcp_server.py
Sample output from mcp_server.py:
Run RAG
python demo.py
Tech Stack Summary
- Language & Runtime: Python 3.8+ with asyncio
- Config: python-dotenv
- MCP Framework: mcp.server.fastmcp + mcp.client.sse
- Search: Azure Cognitive Search (azure-search-documents)
- LLM: Azure OpenAI (openai SDK)
- Transport: Server-Sent Events (SSE)
- Utilities: logging, json, pprint
[ demo.py CLI ] | | SSE → “search” { query, top:3 } ↓ [ MCP FastMCP server (mcp_server.py) ] | | Azure Search query → get top‐N hits ↓ Stream back JSONRPC “content” chunks over SSE | [ demo.py assembles chunks ] | | Azure OpenAI ChatCompletion(prompt with context) ↓ Final answer printed to console
Python & Asyncio
Runs the whole thing as async code (both the demo CLI and the SSE‐based search server). Uses asyncio for concurrency and streaming.
Client side (demo.py)
mcp.client.sse.sse_client for opening an SSE (Server-Sent-Events) connection mcp.client.session.ClientSession to talk the MCP “search” tool
Server side (mcp_server.py)
mcp.server.fastmcp.FastMCP to expose an SSE endpoint and route “search” calls
Azure AI Search
azure.search.documents.aio.SearchClient to query a search index azure.search.documents.indexes.aio.SearchIndexClient to build or update that index Powers the retrieval of document “chunks” (PDF pages, etc.)
Azure OpenAI
openai.AzureOpenAI client against your Azure OpenAI deployment (GPT-4.1) Takes the top N retrieved chunks, concatenates their "content" fields, and feeds them into a chat‐completion prompt
Server-Sent Events (SSE)
A lightweight streaming over HTTP mechanism so the client can send a prompt and incrementally receive back search hits
dotenv for configuration
python-dotenv to pull your Azure Search service name/key, Azure OpenAI endpoint/key, index name, etc. out of a .env file
Standard Python tooling
- logging for debug output
- pprint / json for introspecting responses
- asyncio.run(...) to kick off the async CLI
README copy captured when this server was indexed; it may lag the repository. The original is authoritative.