Skip to content

WaiLyanPyae/rag-vs-mcp

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 

Repository files navigation

📡 RAG + MCP News Classifier & Alert System

A project exploring real-time news classification, RAG-enhanced summarization, and MCP-inspired decision logic.


📖 Overview

This project implements a real-time news pipeline that:

  • Ingests live news articles (RSS feeds).

  • Classifies them into categories: hate_speech, misinformation, emergency, neutral.

  • Summarizes articles using Retrieval-Augmented Generation (RAG).

  • Applies MCP-inspired agent logic to decide system actions:

    • Trigger alerts (email/SMS/webhook).
    • Archive non-critical news.
    • Summarize misinformation/emergency items for context.
  • Exposes results via a REST API + live dashboard.

The system provides a comparative framework:

  • Baseline: Transformer-only (DistilBERT + BART).
  • Proposed: Transformer + RAG + MCP decision layer.

📂 Project Structure

news-rag-mcp/
├─ ingestion/        # RSS poller
├─ backend/          # FastAPI backend (REST + streaming)
├─ models/           # Transformers (classification)
├─ rag/              # FAISS index + RAG summarizer
├─ agent/            # MCP-inspired decision engine
├─ eval/             # Evaluation scripts (metrics, latency)
├─ dashboard/        # React.js frontend (alerts, trends)
├─ infra/            # Postgres + Docker configs
├─ data/             # Datasets and indexes
│   ├─ raw/
│   ├─ processed/
│   └─ indexes/
├─ notebooks/        # Experiments / EDA
└─ scripts/          # Utility scripts

🔍 Research Questions

  • RQ1: Effectiveness of transformer-based classification (DistilBERT, RoBERTa).
  • RQ2: Improvement in summarization via Retrieval-Augmented Generation.
  • RQ3: Structuring MCP-inspired logic for context-aware decision-making.

📈 Evaluation Criteria

Aspect Baseline (Transformer Only) Proposed (RAG + MCP)
Classification Precision, Recall, F1, Accuracy Same
Summarization ROUGE, BLEU (static model) ROUGE, BLEU + human eval
Retrieval N/A Cosine similarity relevance
Decision Logic Rule-based MCP-inspired modular agent
Alert Timeliness Ingestion → alert latency measured Same (<5s target)

🚧 Roadmap

  • RSS ingestion → Postgres
  • Transformer baseline classifier
  • Baseline summarization (BART)
  • RAG summarization with FAISS
  • MCP-inspired rule-based decision logic
  • Full integration pipeline
  • React dashboard
  • Evaluation experiments
  • Dissertation writing

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors