Thysrael/Horizon

What it solves

Horizon addresses information overload by acting as a personalized AI-powered news radar. It aggregates content from disparate sources—such as RSS feeds, Hacker News, Reddit, GitHub, and financial news—and transforms them into a curated, deduplicated, and context-rich daily briefing tailored to your specific interests.

How it works

The project uses a pipeline architecture where various data sources are ingested and processed according to "profiles." Each profile contains editorial rules that define how content should be filtered, scored, and summarized. The system uses LLMs to analyze the content, perform deduplication, and generate summaries based on the specific goals of the profile (e.g., extracting "practical takeaways" for engineering blogs vs. "impact and discussion" for tech news). The final output can be delivered via email, webhooks, or published to a personal site.

Who it’s for

It is designed for professionals, researchers, and developers who need to stay informed across multiple domains without manually checking dozens of feeds. It is particularly useful for those who want to automate the synthesis of news into a structured format that highlights why a story matters.

Highlights

• Multi-Source Aggregation: Supports a wide range of platforms including X (Twitter), Telegram, GitHub, and OpenBB financial news.
• Profile-Driven Curation: Allows users to define custom scoring rubrics and output styles for different types of content.
• Balanced Digest: Prevents any single topic from dominating your briefing by allowing you to set category-specific limits.
• Flexible Delivery: Supports automated publishing to GitHub Pages, email delivery, and notifications via platforms like Feishu, DingTalk, and WeChat.
• Extensible: Includes an MCP (Model Context Protocol) server, enabling AI assistants to trigger pipeline stages.