Generation modes

IANews provides two distinct article generation modes -- Gateway and Agent. Gateway mode is optimized for speed, producing drafts in about 30 seconds from cached sources. Agent mode runs the full discovery-to-generation process in approximately 75 seconds, delivering publication-ready articles with rich citations. This guide explains how each mode works, their parameters, and when to choose one over the other.

Comparison at a glance

How Gateway mode works

Gateway mode bypasses the discovery and content extraction phases entirely. It queries the existing content index for relevant passages, then sends them to the AI for generation.

curl -N -X POST https://api.earlyforge.ai/api/articles/generate \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "topic": "AI regulation in Europe",
    "mode": "gateway",
    "tone": "journalistic",
    "length": "medium",
    "language": "en"
  }'

How Agent mode works

Agent mode runs the complete two-phase process: discovery followed by on-demand generation. This produces the most comprehensive, well-sourced articles.

Real-time streaming events

Both modes stream results in real time. Connect to the stream endpoint to receive live updates as the generation progresses.

const eventSource = new EventSource(`/articles/${articleId}/stream`);
 
eventSource.addEventListener("progress", (e) => {
  // Generation phase: "discovering", "extracting", "indexing", "generating"
  const { stage, message } = JSON.parse(e.data);
});
 
eventSource.addEventListener("sources", (e) => {
  // Discovered sources with metadata, sent before content generation begins
  const { sources } = JSON.parse(e.data);
});
 
eventSource.addEventListener("content", (e) => {
  // Article text chunks — append each to the editor
  const { text } = JSON.parse(e.data);
});
 
eventSource.addEventListener("citations", (e) => {
  // Citation metadata for inline [N] references
  const { citations } = JSON.parse(e.data);
});
 
eventSource.addEventListener("done", (e) => {
  // Generation complete — close the connection
  const { article_id, word_count } = JSON.parse(e.data);
  eventSource.close();
});

Generation parameters

Both modes accept the same request body fields:

Choosing the right mode

Use Gateway when:

• You need a quick first draft in under 30 seconds
• The topic has been covered before and the content index contains relevant material
• You are iterating on tone, structure, or length and need fast feedback loops
• You plan to edit the article heavily in the AI editor afterward

Use Agent when:

• You are covering a topic for the first time and need fresh source discovery
• The topic is trending and benefits from the very latest sources
• You need publication-quality articles with 8--15 inline citations
• Source depth and reliability scoring matter more than generation speed

What's next?

• AI editor -- Refine generated articles with AI-powered editing tools
• Citations and sources -- Understand inline citations and reliability scoring
• SEO optimization -- Optimize meta tags, headings, and keywords before publishing