raoulbia/tagui-qwen-gguf

TagUI-Qwen (GGUF)

Fine-tuned Qwen2.5-Coder-14B-Instruct for TagUI browser automation script generation.

Model Description

This model is specialized for generating TagUI automation scripts. TagUI is an open-source RPA tool for automating web interactions.

Training Data

• ~3,000 curated TagUI examples covering:
  • Web navigation and form filling
  • Data extraction and scraping
  • Control flow (loops, conditionals)
  • Error handling patterns
  • File operations

Key Features

The model understands TagUI-specific syntax including:

• URL navigation: URLs written directly (no navigate command)
• Element interaction: click, type, select, hover
• Data operations: read, dump, write, table
• Control flow: for, if/else, while
• Helper functions: exist(), present(), count()

Usage with Ollama

# Download the GGUF file
wget https://huggingface.co/raoulbia/tagui-qwen-gguf/resolve/main/tagui-qwen-bf16.gguf

# Download the Modelfile
wget https://huggingface.co/raoulbia/tagui-qwen-gguf/resolve/main/Modelfile

# Create the Ollama model
ollama create tagui-qwen -f Modelfile

# Run the model
ollama run tagui-qwen "Write a TagUI script to search Google for 'weather today'"

Example Output

Prompt: "Write a TagUI script to log into a website and download a report"

Response:

https://example.com/login
wait 2

type username_field as myuser
type password_field as mypassword
click login_button

wait 3

if exist('dashboard')
    click reports_link
    wait 2
    click download_button
    wait 5
    echo Report downloaded successfully
else
    echo Login failed

Important TagUI Syntax Notes

1. No navigate command - URLs are written directly on their own line
2. No try/catch - Use if exist() or if present() for error handling
3. count() in loops - Always assign to a variable first:

   num_items = count('selector')
   for i from 1 to num_items
       // process item
   

License

Apache 2.0 (same as base model)

Citation

If you use this model, please cite:

• Base model: Qwen2.5-Coder-14B-Instruct
• TagUI: AI Singapore TagUI