Meridian provides AI-powered daily news briefings, filtering and analyzing hundreds of sources to deliver personalized intelligence to curious individuals. It aims to replicate presidential-level briefings by offering context, implications, and deep analysis beyond typical headlines.

How It Works

Meridian employs a multi-stage AI pipeline orchestrated by Cloudflare Workers and Python. It scrapes RSS feeds, extracts content (using direct fetching or browser rendering for paywalls), and processes articles with Gemini models for analysis. Related articles are clustered using embeddings (multilingual-e5-small), UMAP, and HDBSCAN. Finally, LLMs synthesize these clusters into a coherent, analytical daily brief, presented via a Nuxt 3 frontend. This approach leverages scalable cloud infrastructure and cost-effective AI models (Gemini Flash) for efficient, large-scale news processing.

Quick Start & Requirements

• Install via git clone and pnpm install.
• Requires Node.js v22+, pnpm v9.15+, Python 3.10+, PostgreSQL, a Cloudflare account, and a Google AI API key.
• Setup involves configuring .env files, database migration via pnpm --filter @meridian/database migrate, and manual deployment/execution via Wrangler and Python notebooks.
• Official documentation and setup guides are available within the repository.

Highlighted Details

• Utilizes a multi-stage LLM processing pipeline for article and cluster analysis.
• Employs embeddings, UMAP, and HDBSCAN for smart article clustering.
• Features a Nuxt 3 frontend for a clean user interface.
• Leverages Cloudflare Workers and Workflows for infrastructure.

Maintenance & Community

The project highlights significant AI collaboration, with Claude 3.7 Sonnet assisting in architecture and prompt refinement, and Gemini 2.5 Pro aiding in code review and analysis tone. Gemini 2.0 Flash is noted as the cost-effective workhorse for the intelligence pipeline. No specific community links or roadmap are provided in the README.

Licensing & Compatibility

Released under the MIT License, permitting commercial use and integration with closed-source projects.

Limitations & Caveats

The core pipeline is functional, but automated brief generation is currently a manual process requiring execution of a Python notebook. Scraping robustness requires further improvement.