Article Generation Prompts

This document explains the prompt engineering strategy used to generate high-quality, structured articles using AI.

Overview

The article generation process is a two-step pipeline designed to ensure structural integrity and content quality. It supports varying levels of detail via the depth parameter (standard vs deep-dive).

Structure Generation: The AI first plans the article skeleton (JSON).
Content Generation: The AI then fills in the content based on the plan (Markdown/HTML).

1. Structure Generation Prompt

Goal: To create a valid JSON object defining the article's metadata and sections.

Prompt Template:

`You are an expert article writer... topic: "${topic}".

**Mode: ${depth === 'deep-dive' ? "DEEP DIVE" : "Standard"}**
// If deep-dive: "Aim for 8-12 comprehensive sections. Include advanced concepts..."

Please provide the output in strict JSON format...`

Key Features:

Strict JSON: Ensures the output can be parsed programmatically.
Layout & Aspect Ratio Control: The layout field dictates how the section is rendered, while imageAspectRatio ensures images are generated in the correct dimensions (e.g., 16:9, 4:3, 1:1).
Depth Control: Adjusts section count and complexity based on user request.

1.5 Asset Generation Logic

Goal: To generate visual assets (cover image and section graphics) with high quality and consistency.

Background Image (Cover):

Prompt: Strictly instructed to have NO TEXT, typography, or words.
Configuration: Forced to 16:9 aspect ratio and 2K resolution for professional aesthetics.

Section Images:

Prompt: Contextually enhanced with the article topic.
Configuration: Uses the dynamic imageAspectRatio predicted during Step 1.

2. Content Generation Prompt

Goal: To generate the final Semantic HTML content with embedded styling.

Prompt Template:

`You are an expert web content creator...

Structure: ...
Image URLs: ...

Instructions:
...
// If deep-dive: "Do not be superficial. Each section must be substantial (300-500 words)..."
...`

Key Features:

Semantic HTML: Outputs clean <section>, <figure>, <h2> tags.
Depth Control: Enforces word count minimums and technical depth for "deep-dive" requests.
Asset Integration: Images are placed contextually.

Workflow

User Input: Topic received via API.
Step 1 (AI): Generate JSON Structure (including layouts and aspect ratios).
Step 1.5 (AI & Storage):
- Generate high-res background (cover) image (16:9, no text).
- Iterate through sections. If imagePrompt exists, generate image using the predicted imageAspectRatio, upload to Storage, and get URL.
Step 2 (AI): Generate HTML content using the Structure + Image URLs.
Save: Store the final Article object in the database.