“Let an LLM write and publish blog articles while you sleep.”

---

Overview

This project connects:

• a local LLM (llama.cpp)
• your Statix blog CLI
• a Python automation script

👉 Result: fully automated article generation + publishing.

---

1. Set up a Statix blog

• Statix

---

2. Configure Statix CLI

$ stx set-credentials --url URL --password TOKEN

---

3. Create a subject slot for AI articles

$ stx subject add "AutoBlog - Auto AI generated articles"

List subjects:

$ stx subjects

👉 Note the ID and set while you'll run the script with the --subject_id flag:

• Cf: run

---

4. Set up llama.cpp

$ sudo apt update
$ sudo apt install -y build-essential cmake
$ git clone https://github.com/ggml-org/llama.cpp
$ cd llama.cpp
$ cmake -B build -DGGML_NATIVE=ON
$ cmake --build build -j

---

5. Download a model

Download a GGUF model (example: LLaMA 3.1 8B Q4):
👉 https://huggingface.co/bartowski/Meta-Llama-3.1-8B-Instruct-GGUF

Recommended file:

Meta-Llama-3.1-8B-Instruct-Q4_K_M.gguf

Place it in a folder, e.g.:

~/models/

---

6. Start the LLM server

./build/bin/llama-server \
  -m ~/models/Meta-Llama-3.1-8B-Instruct-Q4_K_M.gguf \
  -c 1024 \
  -t 8

⚠️ Notes

• -c = context size → critical for RAM usage

• Recommended:

  • 1024 → safe (16GB RAM)
  • 2048 → higher quality but heavier

• -t is the number of CPU threads

---

7. Set up Python environment

$ python3 -m venv menv
$ source menv/bin/activate
(menv) $ pip install requests

---

8. Configure autoBlog.py

Project structure:

.
├── articles
└── autoBlog.py
└── prompt.txt
└── models

The prompt is stored in:

prompt.txt

The randomly chosen topics are in:

topics.txt

---

9. Run the automation

Example:

(menv) $ python3 autoBlog.py \ 
  --is_public true \
  --subject-id 24 \
  --timeout 6000 \
  --llm-url http://127.0.0.1:8080/completion \
  --topics-file topics.txt \
  --prompt-file prompt.txt

9.1 Side-effects

After running this command, a metadata Statix file will be created:

.statix_articles.json

That is the only side-effect

---

What happens

Loop:

1. Pick a topic
2. Generate article via local LLM
3. Extract title
4. Save markdown file
5. Create nickname
6. Publish via Statix

---

Output format

• Articles are generated in Markdown
• Title (# Title) is extracted and removed from content
• Body is saved in:

articles/<slug>.md

---

Example pipeline

LLM → Markdown
      ↓
extract title
      ↓
save file
      ↓
stx nickname create
      ↓
stx publish

---

Notes

• No external APIs required
• Fully local (LLM + CLI)
• Performance depends on CPU + RAM
• Quality depends heavily on prompt

---

Next steps

• Better prompt engineering
• Topic generation via LLM
• Deduplication / quality filtering
• Scheduling (cron)

---

Philosophy

No SaaS. No API keys. No black boxes.
Just local compute, full control, and automation.