SignalForge

SignalForge

AI-Powered Trending News Intelligence Platform

Filtering signal from noise - monitor what matters to you

---

📋 Table of Contents

• Overview
• Key Features
• Quick Start
• Configuration
• MCP Server
• Deployment Options
• Notification Channels
• License

---

🎯 Overview

SignalForge is an intelligent news aggregation and monitoring platform that helps you stay informed about topics that matter to you. Instead of drowning in information overload, SignalForge filters trending news from multiple platforms and delivers only what's relevant based on your custom keywords.

Why SignalForge?

• 🎯 Focused Monitoring: Define your own keywords and get notifications only for relevant news
• 📊 Multi-Platform Aggregation: Collects trending topics from 10+ news platforms
• 🤖 AI-Powered Analysis: Built-in MCP server for AI-assisted news analysis
• 🔔 Smart Notifications: Multiple notification channels (Telegram, Email, Feishu, DingTalk, etc.)
• ⚡ Lightweight & Fast: Designed for easy deployment and minimal resource usage

---

✨ Key Features

📰 News Aggregation

• Multi-Platform Support: Aggregates from 10+ popular news platforms
• Smart Filtering: Filter news by custom keywords defined in config/frequency_words.txt
• Real-Time Updates: Configurable crawling intervals
• Weighted Ranking: Customizable algorithm to prioritize news sources

🔔 Notification System

• Multiple Channels: Telegram, Email, Feishu, DingTalk, WeCom, ntfy
• Smart Batching: Automatic message batching for large updates
• Time Windows: Configure quiet hours to avoid notifications during off-hours
• Push Modes:
  • Daily: Daily summary of all matching news
  • Current: Latest trending topics from current rankings
  • Incremental: Only notify when new matching topics appear

🤖 MCP Server (AI Integration)

• 13 Powerful Tools for news query, analytics, and search
• FastMCP 2.0: Modern MCP protocol implementation
• Stdio & HTTP Transports: Flexible deployment options
• AI-Ready: Designed for integration with Claude, ChatGPT, and other AI assistants

MCP Tools Categories:

1. Data Query: Latest news, historical news, trending topics
2. Analytics: Trend analysis, sentiment analysis, platform comparisons
3. Search: Keyword search, fuzzy search, entity recognition
4. System: Configuration management, status monitoring, manual triggers

---

🚀 Quick Start

Prerequisites

• Python 3.10 or higher
• pip package manager

Installation

1. Clone the repository

git clone https://github.com/ruslanmv/SignalForge.git
cd SignalForge

2. Install dependencies

pip install -r requirements.txt
pip install -e .

3. Configure your keywords

Edit config/frequency_words.txt and add keywords you want to monitor (one per line):

AI
Machine Learning
Tesla
SpaceX
Climate Change

4. Configure notifications (optional)

Edit config/config.yaml and add your webhook URLs:

notification:
  webhooks:
    telegram_bot_token: "your-bot-token"
    telegram_chat_id: "your-chat-id"
    email_from: "your-email@example.com"
    email_to: "recipient@example.com"

⚠️ Security Warning: Never commit webhook URLs to public repositories. Use environment variables or GitHub Secrets for sensitive data.

5. Run the crawler

python main.py

---

⚙️ Configuration

Main Configuration File: config/config.yaml

Crawler Settings

crawler:
  request_interval: 1000  # Request interval in milliseconds
  enable_crawler: true    # Enable/disable crawling
  use_proxy: false        # Enable proxy if needed

Report Modes

report:
  mode: "daily"  # Options: "daily" | "current" | "incremental"

• daily: Daily summary of all matching news
• current: Current top trending topics
• incremental: Only new matches (reduces notification spam)

Notification Settings

notification:
  enable_notification: true
  push_window:
    enabled: false     # Enable time-based push control
    time_range:
      start: "09:00"   # Start time (24h format)
      end: "18:00"     # End time
    once_per_day: true # Only one notification per day in window

---

🤖 MCP Server

SignalForge includes a powerful MCP server for AI-assisted news analysis.

Start MCP Server

Stdio mode (for local AI assistants):

signalforge --transport stdio

HTTP mode (for production):

signalforge --transport http --host 0.0.0.0 --port 3333

Available MCP Tools

Category	Tool	Description
Data Query	get_latest_news	Get most recent news batch
	get_news_by_date	Query news by date
	get_trending_topics	Get trending keyword statistics
Analytics	analyze_topic_trend	Analyze topic trends over time
	analyze_data_insights	Platform comparisons and insights
	analyze_sentiment	Sentiment analysis of news
	find_similar_news	Find related/similar news items
	generate_summary_report	Generate daily/weekly reports
Search	search_news	Unified search (keyword/fuzzy/entity)
	search_related_news_history	Historical related news
System	get_current_config	View system configuration
	get_system_status	Check system status
	trigger_crawl	Manually trigger news crawl

MCP Client Configuration

For Claude Desktop, add to your config:

{
  "mcpServers": {
    "signalforge": {
      "command": "signalforge",
      "args": ["--transport", "stdio"],
      "env": {
        "SIGNALFORGE_ENV": "production"
      }
    }
  }
}

---

🐳 Docker Deployment

Using Docker Compose

1. Navigate to docker directory

cd docker

2. Start the service

docker-compose up -d

The crawler will run automatically on an hourly schedule using supercronic.

Environment Variables

Variable	Description	Default
SIGNALFORGE_ENV	Environment (dev/staging/production)	production
SIGNALFORGE_LOG_LEVEL	Logging level	INFO

---

📬 Notification Channels

SignalForge supports multiple notification channels:

Telegram

webhooks:
  telegram_bot_token: "your-bot-token"
  telegram_chat_id: "your-chat-id"

Email

webhooks:
  email_from: "sender@example.com"
  email_password: "app-password"
  email_to: "recipient@example.com"
  email_smtp_server: ""  # Auto-detect if empty
  email_smtp_port: ""    # Auto-detect if empty

Feishu / DingTalk / WeCom

webhooks:
  feishu_url: "https://open.feishu.cn/open-apis/bot/v2/hook/..."
  dingtalk_url: "https://oapi.dingtalk.com/robot/send?access_token=..."
  wework_url: "https://qyapi.weixin.qq.com/cgi-bin/webhook/send?key=..."

ntfy

webhooks:
  ntfy_server_url: "https://ntfy.sh"  # Or your self-hosted server
  ntfy_topic: "your-topic"
  ntfy_token: ""  # Optional for private topics

---

📊 Data Storage

All crawled news is stored in the output/ directory:

output/
  ├── 2025-11-24/
  │   └── txt/
  │       ├── 00-30.txt
  │       ├── 01-30.txt
  │       └── ...
  └── 2025-11-25/
      └── txt/
          └── ...

Each file contains news from all configured platforms at that timestamp.

---

🔧 Advanced Configuration

Platform Weighting

Adjust the importance of different ranking factors:

weight:
  rank_weight: 0.6       # Position in platform rankings
  frequency_weight: 0.3  # Appearance frequency
  hotness_weight: 0.1    # Platform-specific hotness score

Supported Platforms

Current platforms (configurable in config/config.yaml):

• Toutiao (Jinri Toutiao)
• Baidu Hot Search
• Weibo
• Zhihu
• Douyin
• Bilibili
• And more...

---

📝 License

This project is licensed under the GPL-3.0 License - see the LICENSE file for details.

---

🤝 Contributing

Contributions are welcome! Please feel free to submit pull requests.

---

⚠️ Disclaimer

This tool is for personal news monitoring purposes only. Please respect the terms of service of all news platforms. The developers are not responsible for any misuse of this software.

---

📧 Support

For issues and questions, please use the GitHub issue tracker.

---

SignalForge - Forging clear signals from noisy data

Built with ❤️ for the information age