// the find
sansan0/TrendRadar
⭐AI-driven public opinion & trend monitor with multi-platform aggregation, RSS, and smart alerts.🎯 告别信息过载，你的 AI 舆情监控助手与热点筛选工具！聚合多平台热点 + RSS 订阅，支持关键词精准筛选。AI 智能筛选新闻 + AI 翻译 + AI 分析简报直推手机，也支持接入 MCP 架构，赋能 AI 自然语言对话分析、情感洞察与趋势预测等。支持 Docker ，数据本地/云端自持。集成微信/飞书/钉钉/Telegram/邮件/ntfy/bark/slack 等渠道智能推送。
TrendRadar is a self-hosted news aggregator that pulls trending topics from Chinese social platforms (Weibo, Zhihu, Bilibili, etc.) plus RSS, filters by keyword or AI interest matching, and pushes digests to 9+ notification channels. It has grown significantly in scope — now including an MCP server, AI analysis, HTML reports, and a visual config editor — making it a reasonable full stack for personal media monitoring if you live in that ecosystem.
The notification channel implementation is thorough: batching logic for size-limited platforms (Feishu 30KB, DingTalk 20KB) is handled per-channel rather than globally, which actually matters in practice. The two-stage AI filtering pipeline — extract structured tags from natural-language interest descriptions, then batch-score headlines against those tags — is a sensible design that degrades gracefully to keyword matching on failure. The timeline.yaml scheduling system is genuinely flexible: per-period AI vs keyword mode, weekday/weekend differentiation, and conflict detection between overlapping time windows. Switching from a hand-rolled AI client to LiteLLM in v5.3.0 was the right call — it removed a maintenance burden and opened 100+ provider support without code changes.
The core data feed depends on newsnow's third-party API, which the README itself flags as a goodwill dependency — if that changes, the platform coverage collapses unless you self-host newsnow. Every major version has had breaking config changes (v2, v4, v5, v6 all broke config.yaml), and the upgrade path is manual file copy-paste rather than a migration system; the author explicitly left main.py as a monolith to make this easier, which is an honest admission that the architecture is not upgrade-safe. GitHub Actions deployment requires periodic manual renewal ('签到续期') to keep the scheduled job alive — that's a real operational burden that isn't obvious from the quick-start docs. The project is almost entirely Chinese-platform focused; international users without custom RSS feeds get very little out of the box, but the marketing presents it as a general-purpose tool.