AI-powered article generator in Italian.
- Article generation from press releases using AI
- HTML formatting for web publishing
- Modular, extensible codebase
- Clone the repo
- Install requirements:
pip install -r requirements.txt - Set your OpenAI API key as an environment variable.
- Run the demo in tester.py
- In tester.py, change the "topic" value and the "article_type" parameter ('editoriale', 'cronaca', 'intervista') to see different outputs.
- Project modularization.
- Asynchronous calls to OpenAI API for efficiency
- Use different models according to the task:
- GPT-4.1 for articles as it can handle higher token limits.
- GPT-4o-mini for date extraction due to its cost-effectiveness.
- Use of Jinja2 for dynamic prompt engineering.
- Use of Pydantic for structured data handling.
- Attention to detail: handling of dates and deictics, and interruptions in interviews.
- HTML formatting for easy web integration.
- Use of RAG or tool use (via MCP) to fetch relevant news articles for context. Editorial and Interview genres would particularly benefit by setting a vector knowledgebase.
- Hosting an API for make the service accessible.
- Handle several languages or add translation microservice.
- Handle more article genres.
- Add file import for input the article sources.