What Is TrendRadar by sansan0? A Complete Guide to the Open-Source News Aggregation Tool

TrendRadar is an open-source Python tool that aggregates, filters, and analyzes trending news from 11+ Chinese and international platforms, then delivers personalized reports via multiple notification channels or interactive HTML dashboards.

TrendRadar by sansan0 transforms raw hot-search data from platforms like Zhihu, Bilibili, Weibo, and the Wall Street Journal into concise, actionable intelligence. Whether you need daily digests, real-time alerts, or AI-curated content streams, this modular tool automates the entire pipeline from crawling to delivery.

Core Capabilities of TrendRadar

Multi-Platform News Aggregation

TrendRadar ships with built-in crawlers for 11+ platforms and a plugin-style API for adding custom sources. The crawler layer in trendradar/crawler/fetcher.py handles HTTP requests with configurable proxy support and request throttling via REQUEST_INTERVAL.

# From trendradar/__main__.py — pipeline orchestration

analyzer = NewsAnalyzer(config=ctx.config)
analyzer.run()  # Triggers: crawl → analyze → generate HTML → notify

RSS and Atom Feed Support

The trendradar/crawler/rss.py module provides optional RSS/Atom ingestion that shares the same keyword-based filtering logic as hot-search crawlers. Each feed supports max_age_days for freshness filtering and max_items for volume control.

# Example RSS feed configuration structure

new_feed = {
    "id": "myblog",
    "name": "My Blog",
    "url": "https://example.com/rss",
    "max_items": 30,
    "enabled": True,
    "max_age_days": 7
}

Dual Filtering Modes: Keyword and AI-Driven

TrendRadar offers two complementary filtering approaches controlled via config/config.yaml:

Method	Implementation	Use Case
Keyword matching	frequency_words.txt groups with required/normal words	Fast, deterministic filtering by topic
AI classification	trendradar/ai/filter.py (AIFilter) with LLM APIs	Natural-language interest descriptions, semantic understanding

The keyword engine in trendradar/core/analyzer.py computes news weights using rank, frequency, and hotness contributions via calculate_news_weight().

# Switch to AI-only filtering in config/config.yaml

FILTER:
  METHOD: ai
AI_FILTER:
  ENABLED: true
  INTERESTS_FILE: my_interests.txt
  MIN_SCORE: 0.6
  BATCH_SIZE: 200

Three Push Modes for Different Schedules

The trendradar/core/scheduler.py module implements three execution modes controlled by config/timeline.yaml:

• daily: Full-day summary with complete statistics
• current: Latest snapshot of hot searches
• incremental: Only newly-appearing items since last run

The scheduler guarantees once-per-period execution via storage-backed tracking in ResolvedSchedule.

Rich Notification Channel Support

TrendRadar's trendradar/notification/dispatcher.py builds channel-specific payloads and handles batch splitting for oversized messages. Supported channels include:

• Enterprise WeChat, personal WeChat
• Telegram, Feishu, DingTalk
• Slack, ntfy, Bark
• Email, generic webhooks

The optional AI translator (AITranslator) can translate hot-list content when AI_TRANSLATION.ENABLED is true.

Interactive HTML Reports and Visual Editor

TrendRadar generates a self-contained index.html suitable for GitHub Pages hosting. The report includes:

• Visual configuration editor accessible via "配置编辑器" button
• Direct loading and editing of config/config.yaml and config/timeline.yaml
• Real-time preview of platform lists, keyword groups, and schedule templates

Deployment Options for TrendRadar

Docker Container Deployment

The official image wantcat/trendradar supports environment-variable configuration:

docker run -d \
  -e TIMEZONE=Asia/Shanghai \
  -e USE_PROXY=true \
  -v $(pwd)/config:/app/config \
  -v $(pwd)/output:/app/output \
  wantcat/trendradar

GitHub Actions CI/CD

TrendRadar includes ready-to-use workflows for automated execution without local infrastructure.

Command-Line Options

# Standard full pipeline

python -m trendradar

# Skip crawling, use existing data

python -m trendradar --no-crawl

# Install dependencies

pip install -r requirements.txt

Summary

TrendRadar by sansan0 is a production-ready, extensible news aggregation system with these key strengths:

• Modular architecture: Clean separation between crawling, analysis, AI, scheduling, and notification layers
• Dual filtering: Keyword precision plus AI semantic understanding
• Flexible deployment: Docker, GitHub Actions, or direct Python execution
• Rich output: Interactive HTML, multi-channel notifications, and visual configuration
• Extensible design: Plugin API for new platforms and notification channels

The codebase in sansan0/TrendRadar demonstrates best practices for maintainable Python automation tools, with centralized context management via AppContext and storage-agnostic data persistence.

Frequently Asked Questions

What platforms does TrendRadar support out of the box?

TrendRadar includes built-in crawlers for 11+ platforms including Zhihu, Bilibili, Weibo, 36Kr, Juejin, Wall Street Journal, and others. The PLATFORMS configuration in config/config.yaml controls which sources are active, and the plugin-style API in trendradar/crawler/fetcher.py enables adding custom platforms.

How does the AI filtering work in TrendRadar?

The AI filter in trendradar/ai/filter.py (AIFilter class) classifies news items against natural-language interest descriptions stored in ai_interests.txt. It uses LLM APIs (OpenAI, Gemini, DeepSeek) with configurable MIN_SCORE thresholds and BATCH_SIZE for efficient processing. The system caches results and detects interest file changes via hashing to avoid redundant API calls.

Can TrendRadar run without crawling new data?

Yes. TrendRadar supports a --no-crawl flag (handled in trendradar/__main__.py) that skips the DataFetcher entirely. This mode loads existing data from SQLite storage, regenerates the HTML report, and sends notifications—useful for scheduled re-reports or when data is collected by external systems.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}