Summarization¶

MyNews uses the ruby_llm gem for unified access to multiple LLM providers.

Supported Providers¶

Provider	Env Variable	Example Model
Anthropic	`ANTHROPIC_API_KEY`	`claude-sonnet-4-20250514`
OpenAI	`OPENAI_API_KEY`	`gpt-4o`
Google	`GEMINI_API_KEY`	`gemini-2.0-flash`

Configuration¶

In defaults.yml:

llm:
  provider: anthropic
  model: claude-sonnet-4-20250514
  max_tokens: 300

How It Works¶

The summarizer loads unsummarized articles from the database
Each article's markdown content is sent to the configured LLM
A system prompt styled as an Economist editor guides the summary format
The concise summary is stored on the Article record

# Internal flow
chat = RubyLLM.chat(model: config.llm_model)
chat.with_instructions(system_prompt)
response = chat.ask("Summarize this article concisely:\n\n#{markdown}")
article.update(summary: response.content)

System Prompt¶

The summarization prompt is stored in prompts/economist_editor.md and instructs the LLM to produce concise, informative summaries in the style of The Economist.

Running¶

my_news summarize

This processes all articles that don't yet have a summary.