Horizon

Overview

Horizon is an open-source AI-powered news radar that gathers stories from multiple sources, filters and ranks them with configurable profiles, and produces daily briefings in English and Chinese. It adds summaries, background, and community discussion, then publishes or delivers briefings through Markdown, GitHub Pages, email, webhooks, and WeChat.

Key Features

• Collects items from RSS/Atom, Hacker News, Reddit, Telegram, Twitter/X, GitHub, OpenBB, OSS Insight, GDELT, and Google News.
• Uses profiles to define scoring rubrics, thresholds, deduplication, summaries, context, solutions, and takeaways for different content types.
• Generates English and Chinese daily briefings with ranked items, summaries, background, and community discussion.
• Caps digest length and category share so one busy topic does not crowd out other interests.
• Delivers briefings as Markdown, GitHub Pages, email, webhook notifications, and WeChat.
• Supports CLI runs, Docker deployment, a setup wizard, an MCP server, and an AI assistant skill for configuration and operation.

Use Cases

• Following technology news with impact and community discussion using the tech-news profile.
• Tracking engineering deep dives for background, solution, and practical takeaways using the tech-blog profile.
• Curating AI creator material with summary, timely hooks, and content angles using the ai-creator profile.
• Running a self-hosted daily news briefing delivered to email, GitHub Pages, or chat webhooks.

Getting Started

• Install locally by cloning the repository and running uv sync, or build the Docker image with docker compose build.
• Configure Horizon with the interactive wizard (uv run horizon-wizard) or copy .env.example and data/config.example.json for manual setup, then add API keys and sources.
• Run the pipeline with uv run horizon locally or docker compose run --rm horizon in Docker.
• Optionally schedule daily runs with the GitHub Actions workflow template.

Deployment & Requirements

• A local Python environment using uv or pip is supported, with Docker and Docker Compose as alternatives.
• An AI provider must be configured, with API key environment variables such as OPENAI_API_KEY or GOOGLE_API_KEY.
• Node.js 22.20 or newer is required only for the Agent Skill command npx skills add.
• The optional OpenBB financial-news source needs its extra, and the Twitter/X extra requires a Playwright browser and system packages that the current Dockerfile does not install.

Before You Adopt

• License: MIT. Review its terms before using, modifying, or distributing the project.
• WeChat delivery has platform reply limits.
• The optional Twitter/X source requires a Playwright browser and system packages that the current Dockerfile does not install.
• The project describes itself as maintained in spare time.