This is an experiment on using AI to generate news articles.
The system collects programming news from multiple sources, selects the relevant items and summarizes them.
You can add your own sources and adjust the prompts to make it write accordingly to your own editorial policy and biases.
You need to create a .env file with your OpenAI key, in this format
OPENAI_API_KEY=sk-...
The system performs the following steps:
1 - Scanner
Collects top items from the source URLs. Results are stored in a database and
each item is marked as "pending".
2 -
Revalance filter
Picks "pending" items and asks the AI to identify what are the most relevant
ones according to the target audience. Items are marked as "approved".
3 - Scrapper
Articles marked as "approved" are scraped. The resulting content is stored in
the database.
4 - Writer
Asks the AI write a summary about the scraped article.
5 - Publisher
Processed items are published to a static website using Lume.
The possible status transitions for each article:
scan -> create articles with status=pending
candidates -> submit status=pending articles to relevance check
approved ? either yes -> (status=approved) or no -> (status=rejected)
with approved: -> try scraping
scraped ? either yes -> (status=scraped) or no -> (status=error-scraping)
with each scraped: -> write -> (status=published)
Running make run will scan the sources and write the news into the database.
Running make build-pages will use the updated db to generate the static
website under _site
Running make serve will build the static website and serve it at port 3000.
Running sh cron.sh will scan and write website files. It runs every hour.