Auto-News: An Automatic News Aggregator with LLM

A personal news aggregator to pull information from multi-sources + LLM (ChatGPT) to help us read efficiently with less noise, the sources including Tweets, RSS, YouTube, and Articles.

Why need it?

In the world of this information explosion, we live with noise every day, it becomes even worse after the generative AI was born. Time is a precious resource for each of us, How to use our time more efficiently? It becomes more challenging than ever. Think about how much time we spent on pulling/searching/filtering content from different sources, how many times we put the article/paper or long video as a side tab, but never got a chance to look at, and how much effort to organize the information we have read. We need a better way to get rid of the noises, focus on reading the information efficiently based on our interests, and stay on track with the goals we defined.

See this Blog post for more details.

Goals

The Auto-News was born for the following goals:

• Automatically pull feed sources, including RSS and Tweets.
• Support clip content from source directly, later generate summary and translation, including random web articles, YouTube videos
• Filter content based on personal interests and remove 80%+ noises
• A unified/central reading experience (e.g., RSS reader-like style, Notion based)
• Weekly/Monthly top-k aggregations

Architecture

• UI: Notion-based, cross-platform (Web browser, iOS/Android app)
• Backend: Runs on Linxu/MacOS

Backend System Requirements

Component	Requirement
OS	Linux, MacOS
Memory	6GB
Disk	20GB+

Installation

Preparison

• [Required] Notion token
• [Required] OpenAI token
• [Required] Docker
• [Optional] Notion Web Clipper
• [Optional] Twitter Developer Tokens

[UI] Create Notion Entry Page

Go to Notion, create a page as the main entry (For example Readings page), and enable Notion Integration for this page

[Backend] Create Environment File

Checkout the repo and copy .env.template to build/.env, then fill up the environment vars:

• NOTION_TOKEN
• NOTION_ENTRY_PAGE_ID
• OPENAI_API_KEY
• [Optional] Vars with TWITTER_ prefix

[Backend] Build Services

make deps && make build && make deploy && make init

[Backend] Start Services

make start

Now, the services are up and running, it will pull sources every hour.

[UI] Set up Notion Tweet and RSS list

Go to the Notion entry page we created before, and we will see the following folder structure has been created automatically:

Readings
├── Inbox
│   ├── Inbox - Article
│   └── Inbox - YouTube
├── Index
│   ├── Index - Inbox
│   ├── Index - ToRead
│   ├── RSS_List
│   └── Tweet_List
└── ToRead
    └── ToRead

• Go to RSS_List page, and fill in the RSS name and URL
• Go to Tweet_List page, and fill in the Tweet screen names

[UI] Set up Notion database views

Go to Notion ToRead database page, all the data will flow into this database later on, create the database views for different sources to help us organize flows easier. E.g. Tweets, Articles, YouTube, RSS, etc

Now, enjoy and have fun.

Operations

[Monitoring] Control Panel

For troubleshooting, we can use the URLs below to access the services and check the logs and data

Service	Role	Panel URL
Airflow	Orchestration	http://localhost:8080
Milvus	Vector Database	http://localhost:9100
Adminer	DB accessor	http://localhost:8070

Stop/Restart Services

In case we want, apply the following commands from the codebase folder.

# stop
make stop

# restart
make stop && make start

Redeploy .env and DAGs

make stop && make init && make start

Upgrade to the latest code

make upgrade && make stop && make init && make start