AIHOT

KKKKhazix/AIHOTReviewed by Radar

What it does

AIHOT is the engine behind the Chinese AI news site aihot.news. You host it yourself: it collects posts from RSS feeds, web pages, X accounts and WeChat accounts, asks an AI model to score and summarise them, groups stories about the same event, ranks what is trending and publishes daily, weekly and monthly digests. Titles and summaries are written in Chinese by default. It suits people who want a news site for their own field. You need Docker, Node.js 24 and a key for an OpenAI-compatible AI service such as DeepSeek.

How easy it is to install

D · terminal, several tools or a specific graphics card · Windows

You need Docker and Node.js 24, then you clone the repository, generate the settings with a script and start several services with docker compose from a terminal; running without Docker on Windows needs WSL2.

Coming soon: install in 1 click with the RepoPicks app.

Numbers on GitHub

Stars: 8,983on Oct 31, 2026
Downloads: —all releases
Growth: +308stars in 5 days

Stars over time

Facts

Runs on: Windows, macOS, Linux
You need to install: Docker, Node.js
To build from source: Node.js, Docker
Disk space: 100 MB–500 MB · estimate
Accounts or keys: OpenAI-compatible AI model (DeepSeek by default; Qwen or Zhipu also work), Embeddings API (OpenAI or Alibaba DashScope), SocialData (X posts), Dajiala (WeChat articles), Jina Reader, Feishu (notifications and login)
Cost: Free, API costs extra · The software is free (MIT), but every step calls a paid AI model API; optional collectors (SocialData for X, Dajiala for WeChat, Jina Reader) charge per request.
License: MIT (Permissive)
Can replace: Feedly, Inoreader
Last update: Nov 1, 2026

Risk signals (3)

AttentionVery new repository with many stars
8675 stars in 28 days
automatic check
Infoauthor not professional developer
README.md: "我不是专业的开发者。…这套代码是我和 AI 一起重写的" (the author says they are not a professional developer and rewrote the code together with AI).
AI review
Infoport open on all interfaces
.env.example: "网站在本机开的端口（docker compose 用，默认 3000，所有网卡都能访问）" — by default the site port is reachable from every network interface, not only this computer.
AI review

Evidence is shown in its original language.

What Radar could not check

How much the AI model calls cost per day with a real list of sources.
Whether the Docker setup runs smoothly on Windows with Docker Desktop.
How well the scoring and summaries work for fields other than AI, or for languages other than Chinese.
Real disk use once the database and image cache grow.