Blog Writer Agent

A powerful Streamlit-based blog writing application powered by the Agno framework and OpenAI. Generate technical blog posts with AI-assisted outlines, brand-aligned hero images, inline illustrations, and multi-format exports.

Features

Content Generation

• AI-Powered Outlines: Generate structured blog post outlines from a topic or URL
• Draft Writing: Streamed content generation with the technical-blog-writer skill
• Markdown Editor: Edit drafts in-browser with streamlit-ace before image generation
• URL Content Fetching: Extract and rewrite content from existing articles

Image Generation

• Hero Images: Wide landscape images (1536x1024) for page headers with multiple variants
• Inline Images: Square illustrations (1024x1024) for blog sections
• Brand Integration: Multi-brand system with configurable palettes, styles, and tones
• Vision Support: GPT-4o Vision for image analysis and validation

Workflow

1. Input: Enter a topic idea or provide a URL to rewrite
2. Outline: Review and edit the generated outline
3. Draft: Streamed generation with live preview
4. Edit: Refine content in the markdown editor
5. Hero Images: Generate and select hero image variants
6. Inline Images: Generate illustrations for key sections
7. Export: Save with embedded images, generate social snippets

Archive & Organization

• Session-Based Files: All files saved with consistent timestamps (YYYYMMDD_HHMMSS)
• Archive Page: Browse posts and related images grouped by session
• Auto-Save: Drafts automatically saved after generation
• Prompt History: Image generation prompts saved to YAML files

Installation

# Clone or navigate to the project
cd blog-agent

# Install dependencies (recommended: use uv)
uv pip install -r requirements.txt

# Or with pip
pip install -r requirements.txt

# Copy environment template
cp .env.example .env

# Edit .env with your API keys
# - OPENAI_API_KEY (required for image generation and LLM)

Usage

# Run the Streamlit application
streamlit run Home.py

The app will open at http://localhost:8501

Project Structure

blog-agent/
├── Home.py                    # Main Streamlit application
├── config.yaml                # App configuration (LLM, image models, vision)
├── config.py                  # Configuration loader with defaults
├── requirements.txt           # Python dependencies
├── .env                       # Environment variables (API keys)
├── agent/                     # Agent orchestration and tools
│   ├── orchestrator.py       # Agno agent setup
│   ├── tools.py              # File saving, URL fetching tools
│   └── workflow.py           # Blog generation workflow
├── pages/                     # Streamlit multi-page setup
│   ├── 2_⚙️_Config.py        # Configuration UI
│   └── 3_📁_Archive.py       # Archive browser
├── skills/                    # Specialized skills
│   ├── image_gen.py          # DALL-E / GPT image generation
│   ├── vision.py             # GPT-4o Vision analysis
│   ├── brand_visual.py       # Brand validation
│   └── technical_blog_writer/ # Blog writing skill
├── brands/                    # Brand configuration files
│   ├── anthropic.yaml        # Example brand
│   └── corpusanalytica.yaml  # Your custom brand
├── output/                    # Generated content
│   ├── posts/                # Saved blog posts (.md)
│   ├── images/               # Generated images (.png)
│   └── prompts/              # Image prompts (.yaml)
└── templates/                # Export templates

Configuration

config.yaml

active_brand: corpusanalytica

llm:
  provider: openai
  model: gpt-4o-mini

image:
  hero:
    model: gpt-image-2      # or dall-e-3
    size: 1536x1024
    quality: high
    num_variants: 1
    output_format: png
  inline:
    model: gpt-image-2
    size: 1024x1024
    quality: high
    max_per_session: 2
    output_format: png

vision:
  model: gpt-4o             # Vision model for analysis
  detail: auto              # auto | low | high
  max_image_size: 2048

Brand Configuration

Create custom brand files in brands/:

name: CorpusAnalytica
style: Minimalist monochrome abstract composition...
tone: particle simulation, neural network aesthetics...
palette:
  primary: '#1C1917'
  accent: '#E27E03'
  background: '#FAF8F5'
logo_path: https://example.com/logo.png
reference_image: https://example.com/reference.png
validation:
  enabled: true
  min_score: 7
  max_retries: 2

Image Models Supported

• GPT Image Models: gpt-image-2, gpt-image-1.5, gpt-image-1, gpt-image-1-mini

  • Returns base64 encoded images
  • Supports output_format and background parameters

• DALL-E Models: dall-e-3, dall-e-2

  • Returns image URLs
  • Supports hd and standard quality

Vision Configuration

Vision models analyze generated images for brand alignment:

Model	Quality	Best For
gpt-4o	High	Detailed analysis, brand validation
gpt-4o-mini	High	Cost-conscious (higher vision token cost)
gpt-4.1	High	Balanced performance
gpt-4.1-mini	Medium	Fast processing
gpt-4.1-nano	Basic	Quick checks

Detail levels:

• low: 512x512 fixed (~85 tokens)
• high: Full resolution
• auto: Let OpenAI decide

Output File Naming

All files use consistent session timestamps:

output/
├── posts/
│   ├── blog-20250110_143022.md          # Final saved post
│   └── blog-20250110_143022-auto.md     # Auto-saved draft
├── images/
│   ├── hero-20250110_143022-v1.png      # Hero variant 1
│   ├── hero-20250110_143022-v2.png      # Hero variant 2
│   ├── inline-20250110_143022-1.png     # Inline image 1
│   └── inline-20250110_143022-2.png     # Inline image 2
└── prompts/
    └── prompts-20250110_143022.yaml      # Generation prompts

Archive Page

Browse all generated content by session:

• Sessions: Grouped by timestamp
• Posts: Preview content, download markdown
• Images: View hero and inline images per session
• Stats: Total sessions, posts, and images

Dependencies

• agno ≥2.6.5 — Agent framework
• streamlit ≥1.40 — Web UI
• openai ≥1.0 — Image generation and LLM
• streamlit-ace ≥0.1.1 — Markdown editor
• beautifulsoup4, readability-lxml — URL content extraction
• python-docx, markdown — Export formats
• Pillow — Image processing
• PyYAML ≥6.0 — Configuration files

Environment Variables

OPENAI_API_KEY=sk-...           # Required
# Optional: Other API keys if extending

License

MIT License - See repository for details.

Credits

• Built with Agno framework
• Images generated via OpenAI API
• UI powered by Streamlit