📝 Blogen - Community Edition Models

This repository contains the quantized AI models used by Blogen Community Edition, a self-hosted, privacy-focused AI blogging assistant.

These models are optimized for local inference (CPU) using GGUF format.

📦 Included Models

Model Type	Model Name	Filename	Size	Description
LLM	Google Gemma 3 12B IT	gemma-3-12b-it-q4_0.gguf	~4.7 GB	Instruction-tuned model for generating blog posts, titles, and SEO metadata. Quantized to 4-bit (Q4_0).
Image Gen	Stable Diffusion v1.5	stable-diffusion-v1-5-pruned-emaonly-Q4_1.gguf	~2.0 GB	Text-to-Image model for generating blog cover images. Quantized to Q4_1 for stable-diffusion.cpp.

✨ New Capabilities (v1.1)

These models power the latest version of Blogen, enabling:

• 🌍 Multilingual Blogging: Native support for generating content in Spanish, French, German, and 50+ languages via Gemma 3 instructions.
• 🎨 High-Fidelity Images: Optimized Stable Diffusion pipeline with 30-step generation for clearer, artifact-free cover images.
• 🛡️ Enterprise Grade: Ready for secure, air-gapped deployments with Ed25519 license verification.

🚀 Usage

These models are designed to be automatically downloaded by the Blogen Docker container upon startup.

Manual Download & Run

If you prefer to download them manually (e.g., to save bandwidth on re-deployments):

1. Download the files to a local folder (e.g., ./models).
2. Run Blogen Community Edition:

   docker run -d \
     -p 3000:3000 \
     -v $(pwd)/models:/app/models \
     -v $(pwd)/data:/app/data \
     ghcr.io/org-runink/blogen/server:free
   

⚖️ License & Acknowledgments

• Blogen CE: Apache 2.0
• Gemma 3: Gemma Terms of Use (Google)
• Stable Diffusion: CreativeML Open RAIL-M (RunwayML / Stability AI)

These files are quantized redistributions of the original models cited above.