TrendRadar is an AI-driven open-source project designed as a public opinion and trend monitoring tool that aggregates information from multiple platforms and RSS feeds, offering smart alerts and AI-powered analysis. Its primary goal is to combat information overload by providing a lightweight and easily deployable solution for filtering and delivering relevant news.

The core methodology of TrendRadar encompasses several sophisticated components:

1. Multi-Platform Data Aggregation:
   • The system sources data from a diverse set of mainstream platforms (e.g., Zhihu, Douyin, Bilibili, Baidu Hot Search, Wall Street News, Phoenix News, Toutiao, Weibo) and is designed to be extensible to include more.
   • It leverages the newsnow project's API for multi-platform data acquisition, with a note on respectful usage to avoid overwhelming the upstream service.
   • RSS/Atom Feed Integration (v4.5.0+): TrendRadar seamlessly integrates RSS/Atom feeds, allowing users to subscribe to personalized sources. These RSS items are processed and filtered by the same keyword mechanism as platform-based news and can be consolidated into a single push message.

2. Intelligent Content Filtering and Analysis:
   • Keyword-based Filtering: Users define frequency_words.txt to specify keywords. The system supports various syntax for precise control:
     • Basic keyword matching.
     • Advanced Regular Expressions (v4.7.0+): /pattern/ syntax enables sophisticated matching (e.g., $/\bai\b/$ for "AI" as a whole word).
     • Display Names (v4.7.0+): $=>$ syntax allows assigning user-friendly names to complex regex patterns for clearer display.
     • Required Words: Using +keyword ensures that specified words must be present in a news title for inclusion.
     • Exclusion/Global Filtering: !keyword or [GLOBAL_FILTER] section enables filtering out unwanted content (e.g., ads, low-quality news).
   • Customizable Hotness Algorithm: TrendRadar re-ranks aggregated news to present the most relevant and trending items. This algorithm incorporates configurable weights:
     • Platform popularity.
     • Timeliness (how recent the news is).
     • Duration (how long it has been trending).
     • This allows users to prioritize specific aspects of "hotness" beyond native platform algorithms.
   • Hot Trend Analysis: Beyond simple filtering, the system tracks the lifecycle of news topics:
     • Timeline Tracking: Records the full time span a news item appears, from its first to last detection.
     • Heat Change: Quantifies ranking fluctuations and appearance frequency over time.
     • New Detection: Highlights newly emerging topics with a "🆕" marker.
     • Cross-Platform Comparison: Identifies how the same news item performs across different platforms, indicating media attention divergence.
   • AI Analysis Push (v5.0.0+): A significant feature leveraging large language models (LLMs) for deep content analysis:
     • Integration: Supports DeepSeek (default), OpenAI, Google Gemini, and any OpenAI-compatible API endpoints.
     • Analysis Scope: Automatically generates reports on trend overviews, keyword popularity trends, cross-platform correlation, potential impact assessment, sentiment analysis (positive/negative/controversial), and strategic recommendations.
     • HTML Embedding (v5.2.0+): AI analysis results are directly embedded into HTML reports with rich styling.
     • Data Input: AI can access and analyze historical ranking changes, duration of hotness, cross-platform performance, and potentially predict trends.
     • Customizable Prompts: Users can define AI analysis roles and output formats via config/ai_analysis_prompt.txt.
   • AI Multi-language Translation (v5.2.0+): Enables real-time translation of aggregated content into any target language (e.g., English, Korean, Japanese), breaking language barriers for international news and RSS feeds. It intelligently batches translation requests to optimize API calls and costs.

3. Push Notification Strategy:
   • Multiple Push Modes:
     • daily: Summarizes all matching news for the day (may include previously pushed items). Suitable for daily reports.
     • current: Pushes the current hotlist at each execution, showing ongoing trends. Suitable for content creators.
     • incremental: Only pushes newly detected content since the last run, minimizing repetition. Ideal for investors/traders focused on new signals.
   • Configurable Push Time Windows: Users can set specific time ranges (e.g., 09:00-18:00) to receive notifications, avoiding disruptions outside work hours.
   • Content Order Configuration: Allows customizing the display order of "hot keywords statistics" and "new hot news."
   • Multi-Channel & Multi-Account Support (v3.5.0+):
     • Supports a wide array of notification channels: Enterprise WeChat (with personal WeChat forwarding option), Feishu, DingTalk, Telegram, Email, ntfy, Bark (iOS-specific), Slack, and a generic Webhook for custom integrations (e.g., Discord, Matrix).
     • Each channel can be configured with multiple accounts (e.g., $FEISHU_WEBHOOK_URL=url1;url2$), allowing parallel pushes to different groups or devices.
     • Handles message length limits by automatically batching and splitting content for platforms like DingTalk, Telegram, Feishu, and Slack.

4. Data Management and Storage Architecture (v4.0.0+):
   • Modular Storage Backend: Introduces a flexible storage module supporting:
     • Local SQLite Database: Default for Docker and local environments, providing full data control.
     • Remote Cloud Storage (S3-compatible): Default for GitHub Actions, supporting protocols like Cloudflare R2, ensuring data persistence without polluting the repository.
   • Automated Backend Selection: The system intelligently chooses the appropriate storage method based on the deployment environment.
   • Database Optimization: Refactored SQLite table structures for improved data efficiency and query performance.
   • Data Persistence: Addresses issues with remote storage data persistence and ensures accurate data merging.

5. Deployment Options:
   • GitHub Actions: Leverages GitHub's free resources for automated, scheduled crawling, relying on remote cloud storage for data. Requires periodic "check-ins" for continued service.
   • Docker Deployment: Provides containerized execution with multi-architecture support, allowing for local data storage. A dedicated wantcat/trendradar-mcp Docker image is available for the AI analysis service, enabling a dual-container architecture for scalability.
   • Local Execution: Direct execution on Windows, Mac, or Linux environments.
   • Built-in Web Server (v3.5.0+): Generates index.html reports accessible via a simple web server, configurable for local access or deployment to static hosting platforms like GitHub Pages/Cloudflare Pages.

TrendRadar caters to a diverse audience, including investors, self-media professionals, public relations specialists, and general users interested in current affairs. Typical use cases involve monitoring stock market trends, tracking brand public sentiment, following industry developments, and curating personal news feeds. The project actively encourages community contributions and financial support for its continued development and API costs.