GitHub radar
A framework for a site that finds and writes its own news digests
AIHOT is an open-source engine for a self-made media site: it gathers sources, scores the news, and publishes a ready digest every day.
It's website code that scans chosen sources daily, uses a language model to pick what matters, and turns that into headlines and short digests.
When it helps
Useful if you want your own topic-focused news site (law, HR, finance, AI, etc.) without hiring an editorial team. Skip it if manual news monitoring is enough or you have no LLM API access.
Pros
- Runs on an ordinary computer, no GPU needed
- Automatically merges news from different sources into one card
- Customizable for any topic and your own selection criteria
- A cheap language model is enough, no need for an expensive one
Cons
- Site and prompts default to Chinese — you'll need to translate them
- Requires a paid LLM API key, even if a cheap one
- First news processing run takes about half an hour
How to set it up — step by step
- 1Install Docker and Node.js 24 — the site won't start without them.
- 2Clone the repo and launch it via Docker using the command from the README, with your model's API key.
- 3Get an LLM API key with an OpenAI-compatible interface, e.g. DeepSeek, Qwen, or Zhipu GLM.
- 4Open the site at http://localhost:3000, the admin panel is at /admin.
- 5Send the agent the text from the block below to configure sources and language for your topic.
- 6Check the daily digest the next day — the first processing run takes about half an hour.
Text for your agent
Copy this and send it to your agent — Claude Code, Codex, any of them:
Deploy AIHOT from https://github.com/KKKKhazix/AIHOT — bring it up via Docker following the README, tune it to my topic (I'll tell you which sources matter and what I care about), and show me the daily briefing at http://localhost:3000. You'll need a key for an OpenAI-compatible LLM API — ask me for it if needed.
Light: a regular computer with Docker and Node.js is enough, no GPU needed. All the model work runs through an API key for any OpenAI-compatible LLM — DeepSeek, Qwen or Zhipu GLM, for instance.
Open on GitHub▌ More finds