🧠 AI Metadata Generator & Applier for Markdown Files

This project automates metadata generation and application for Markdown (.md) files using a local language model (LM Studio). It first generates meaningful metadata (like title, category, summary, and tags), then safely applies it back to the files.

---

📦 Project Structure

.
├── generate\_metadata.py         # Uses LM Studio to extract metadata from Markdown content
├── apply\_metadata.py            # Applies generated metadata to frontmatter of files
├── config.example.json          # Example config file (copy & rename to config.json before use)
├── metadata\_output.json         # Generated metadata (intermediate output)
└── archive/                     # Backup of original Markdown files (created on first run)

---

🔧 Requirements

• Python 3.x
• LM Studio running locally (default API: http://localhost:1234/v1/chat/completions)
• No external Python libraries needed (uses standard library only)

---

⚙️ Configuration

Before running, copy config.example.json to config.json and edit the path:

cp config.example.json config.json

Edit config.json:

{
  "MY_PATH": "path/to/your/markdown/files"
}

Replace "path/to/your/markdown/files" with the absolute or relative path to your Markdown directory.

---

🚀 Usage

Step 1: Generate Metadata Using LM Studio

python generate_metadata.py

• Scans all .md files in the configured folder.

• Uses LM Studio to generate:

  • title
  • summary
  • tags
  • category

• Saves results to metadata_output.json.

Step 2: Apply Metadata to Files

python apply_metadata.py

• Reads metadata_output.json.
• Updates each file's frontmatter only if the field exists and is empty.
• Tags are normalized by replacing spaces with underscores.
• Creates backups of original files in archive/ folder before applying changes.

---

📝 Example: metadata_output.json

[
  {
    "filepath": "notes/example.md",
    "filename": "example.md",
    "title": "Benefits of Morning Walks",
    "summary": "Walking early in the morning improves mental clarity and physical health.",
    "tags": ["health", "exercise", "morning"],
    "category": "health"
  }
]

---

🔐 Safe by Design

• Original .md files are backed up to archive/ before changes.
• Only existing and empty frontmatter fields are updated.
• Tags are automatically cleaned ("artificial intelligence" → "artificial_intelligence").
• <think>...</think> blocks (if any) are stripped from input text before analysis.

---

📌 Notes

• LM Studio must be running locally and accessible at the defined API port.
• The script reads folder path from config.json. Make sure to rename and edit the example config before running.
• This system does not add new frontmatter fields if they don't already exist — it fills in missing values only.

---

🛠️ Customization

• You can tweak the prompt behavior in generate_metadata.py:

prompts = {
    "category": "Return one simple category label...",
    ...
}

• You can modify apply_metadata.py to allow adding new fields or change archiving behavior.

---

📃 License

MIT — Use freely, modify safely.

---