tags: - ai-news-aggregator - daily-briefing - content-curation - multi-language - open-source

Horizon: An AI-Powered News Radar for Personalized Daily Briefings¶

Introduction¶

Horizon is an open-source AI-powered news radar that aggregates, curates, and delivers personalized daily briefings in both English and Chinese. In an era where information overload is commonplace, Horizon helps users filter through noise from multiple sources including Hacker News, Reddit, RSS feeds, Telegram channels, Twitter/X, GitHub releases, and OpenBB financial news. By combining artificial intelligence with human-curated preferences, Horizon creates a streamlined reading list that prioritizes relevance and context.

Core Features and Functionality¶

Multi-Source Content Aggregation¶

Horizon consolidates content from diverse sources into a single pipeline:

Hacker News: Top stories by score with community comments
RSS/Atom Feeds: Any standard web feed
Reddit: Subreddit posts and user content with top comments
Telegram: Public channel messages
Twitter/X: Tweets from specific accounts
GitHub: User events and repository releases
OpenBB: Financial company news by watchlist/provider

AI-Powered Scoring and Filtering¶

The platform uses advanced AI models to evaluate and rank content:

Supports multiple AI providers including Claude, GPT, Gemini, DeepSeek, Doubao, MiniMax, and Ollama
Scores each item on a 0-10 scale based on relevance and importance
Allows customizable scoring thresholds to filter content
Configurable through a single JSON configuration file

Content Enrichment and Deduplication¶

Horizon enhances the reading experience through intelligent processing:

Deduplication: Merges repeated stories across different platforms
Background Context: Adds web-researched context for unfamiliar concepts, companies, and technical terms
Community Discussion: Collects and summarizes comments from supported sources

Multi-Language Support¶

The system generates daily briefings in both English and Chinese from the same source set, making it accessible to bilingual users and those seeking content in either language.

Deployment and Delivery Options¶

Multiple Output Channels¶

Horizon supports various delivery methods to suit different preferences:

GitHub Pages: Publishes generated Markdown as a daily-updated briefing site
Email: Self-hosted SMTP/IMAP newsletter with automatic subscription management
Webhooks: Pushes results to Feishu/Lark, DingTalk, Slack, Discord, or custom endpoints
MCP Server: Exposes pipeline steps as tools for AI assistant integration

Installation Methods¶

Users can deploy Horizon through two primary approaches:

Local Installation: Using uv package manager or pip
Docker: Containerized deployment using Docker Compose

Automation Support¶

The platform integrates seamlessly with GitHub Actions for automated daily briefing generation and deployment to GitHub Pages.

Configuration and Customization¶

Setup Wizard¶

Horizon includes an interactive setup wizard (horizon-wizard) that asks about user interests and automatically generates personalized source configurations, eliminating the need for manual setup.

Manual Configuration¶

For advanced users, direct configuration is possible through:

.env file for API keys and environment variables
data/config.json for detailed source, filtering, and delivery settings
Environment variable references using ${VAR_NAME} syntax

Configuration Sections¶

The system allows customization across multiple dimensions:

AI provider settings and model selection
Source configurations for each supported platform
Filtering thresholds and scoring parameters
Delivery channel settings (email, webhooks, etc.)
Language preferences for bilingual output

Technical Architecture¶

The processing pipeline follows a structured workflow:

Configuration: Define sources, thresholds, models, and outputs
Fetch: Concurrently pull content from all configured sources
Deduplicate: Merge items pointing to the same story or URL
Score & Filter: Use AI to rank and filter based on thresholds
Enrich: Add context and collect community discussion
Summarize: Generate structured Markdown briefing
Deliver: Publish to configured output channels

Getting Started¶

Quick Setup Process¶

Installation: Choose between local installation or Docker deployment
Configuration: Use the interactive wizard or manual setup
Execution: Run Horizon with default 24-hour window or custom timeframes
Automation: Optionally set up GitHub Actions for daily automation

System Requirements¶

Python 3.10+ for local installation
Docker for containerized deployment
Various API keys for AI providers and integrated services

Project Status and Future Development¶

Horizon currently supports the complete daily briefing loop with all major features implemented. Planned improvements include additional source types (such as Discord), custom scoring prompts per source, and direct PyPI package publication.

The project maintains an active community with 5.3k GitHub stars and 727 forks, demonstrating strong interest in automated content curation solutions.

Conclusion¶

Horizon represents a sophisticated approach to managing information overload through AI-powered content curation. By combining multiple data sources, intelligent filtering, and flexible delivery options, it provides users with personalized daily briefings that cut through the noise. Its open-source nature, combined with extensive customization options and multi-language support, makes it a versatile tool for professionals, researchers, and content enthusiasts seeking to stay informed efficiently.