MERIDIAN 🔍

An autonomous agent that monitors FCA publications, classifies by urgency, and delivers structured briefings.

Regulatory agencies publish hundreds of documents a year and nobody has time to read it all. Meridian is a working agent that solves it by fetching, classifying, and briefing you on what actually matters. As an agentic system, it scores each document from 1 (least urgent) to 5 (most urgent), based on document type and document title. If a document scores below 4, MERIDIAN stores the metadata only. If it scores 4 or 5, it fetches the full document, summarises it, and flags it for the briefing.

---

How It Works

initialise db
↓
fetch FCA feed
↓
for each document:
    check if seen → skip if yes
    classify document type
    score urgency on document type
    score urgency on title keywords
    ↓
    if urgency >= threshold (4):
        fetch full document
        summarise document
    
    store document in db
↓
generate briefing
↓
email briefing

Tech Stack

Component	Technology
Language	Python 3.14
LLM	Qwen 2.5 via Ollama
Database	SQLite
Feed Source	FCA RSS
Key Libraries	feedparser, requests, hashlib, datetime, schedule, pypdf
Email	Resend

Getting Started

Prerequisites

• Python 3.14+
• Ollama installed and running
• Resend make an account and get an API key

# clone repo
git clone https://github.com/RachelBurman/MERIDIAN.git
cd MERIDIAN

# install dependencies
pip install requests feedparser schedule pypdf python-dotenv resend

# pull the LLM
ollama pull qwen2.5:latest

# create data directory
mkdir data

# run MERIDIAN
python main.py

Create a .env file in the project root:

RESEND_API_KEY=re_xxxxxxxxxxxx
RESEND_FROM=you@yourdomain.com
RESEND_RECIPIENT=your@email.com

Example Reports

Here is an example brief taken from 14/05/2026 View example briefing

Roadmap

• Core agentic pipeline — fetch, classify, summarise, brief
• Title-based keyword urgency classification
• Error handling and timeouts
• README + example briefing
• Logging
• Twice daily scheduling
• Domain tag matching — completes urgency scoring
• PDF handling — richer summaries from policy documents
• Email delivery — production ready briefings
• Migrate email to Resend
• Run as background service
• Library cross-reference — SCALPEL integration

About

Hi, I'm Rachel, a data scientist with a passion for AI and machine learning, currently working as a research assistant in data science and clinical trials.

GitHub • Personal Website