AI Article Writer

A multi-agent article writing system built with the OpenAI Agents SDK and Gradio. Enter any topic and a coordinated pipeline of AI agents researches the web, generates an image, writes a polished article, and checks it through an editorial guardrail — all in one click.

---

Examples

• View Markdown Healthcare
• View Markdown n8n
• View Markdown Firefighting

Live Demo on Hugging Face

AI Article Writer

How It Works

The system uses six specialised agents coordinated by an orchestrator:

Agent	Role	Model
Orchestrator	Coordinates all agents; chooses writer style	o4-mini
Research Agent	Searches the web and extracts content from sources	gpt-4.1-mini
Image Generator	Crafts a DALL-E prompt and generates the article image	gpt-4.1-mini + DALL-E 3
Writer A	Writes in a formal, academic style	gpt-4.1-mini
Writer B	Writes in an engaging, conversational style	gpt-4.1-mini
Guardrails Agent	Reviews the article for safety, accuracy, and quality	gpt-4.1-mini

Pipeline flow:

1. The orchestrator calls the Research Agent twice with slightly different queries and selects the richer brief.
2. The Image Generator Agent creates a relevant DALL-E 3 image.
3. The orchestrator hands off to Writer A (technical topics) or Writer B (general-interest topics).
4. The Guardrails Agent checks the article before it is returned. If it fails, the run is blocked.
5. The article (.md), image (.png), and combined page (.html) are saved to outputs/ — all sharing the same timestamp.

---

Features

• Live web research — DuckDuckGo search + full-page text extraction via trafilatura (minimum 6 sources)
• AI image generation — DALL-E 3 produces a 1024×1024 article image
• Dual writing styles — formal/academic or conversational, chosen automatically by the orchestrator
• Editorial guardrail — a dedicated agent checks safety, coherence, completeness, and quality
• Timestamped exports — .png, .md, and .html files share a timestamp so runs never overwrite each other
• Gradio UI — runs locally or deploys directly to Hugging Face Spaces

---

Quickstart

1. Clone and install

git clone https://github.com/<your-username>/ai-article-writer.git
cd ai-article-writer
pip install -r requirements.txt

2. Set your API key

Create a .env file in the project root:

OPENAI_API_KEY=sk-...

3. Run

python app.py

Open http://127.0.0.1:7860 in your browser, enter a topic, and click Write Article.

⏱️ Each run takes 2–4 minutes — the research agent queries multiple web sources before writing begins.

---

Outputs

Every run saves three files to the outputs/ folder, all sharing the same timestamp:

outputs/
  article_date_time.png   ← DALL-E 3 image
  article_date_time.md    ← raw Markdown article
  article_date_time.html  ← styled HTML page with image embedded

The .html file references the .png by relative filename, so keep them in the same folder when sharing.

---

Project Structure

ai-article-writer/
├── app.py                            # Gradio app — all agents, tools, and UI
├── requirements.txt
├── outputs/                          # Generated articles and images (git-ignored)
├── Article_Writer_Agentic_System.png # Architecture diagram
└── .env                              # API key (not committed)

---

Deploying to Hugging Face Spaces

1. Create a new Space → Gradio SDK.
2. Push this repository to the Space.
3. Add OPENAI_API_KEY as a Space Secret (Settings → Variables and secrets).
4. The Space will install requirements.txt and launch app.py automatically.

---

Requirements

Package	Purpose
openai-agents	OpenAI Agents SDK — agent orchestration and tool use
openai	DALL-E 3 image generation
gradio	Web UI
ddgs	DuckDuckGo web search
trafilatura	Web page text extraction
markdown	Markdown-to-HTML conversion for exports
requests	Downloading generated images
python-dotenv	Loading .env for local development
pydantic	Structured guardrail output model

---

Configuration

Variable	Where to set	Description
OPENAI_API_KEY	.env or HF Secret	Required — your OpenAI API key
MODEL	app.py line 22	Base model for all agents (default: gpt-4.1-mini)
EXAMPLE_TOPIC	app.py line 23	Pre-filled placeholder in the UI

---

License

MIT