README.md

metadata

title: n8n Workflow Generator
emoji: 🚀
colorFrom: red
colorTo: green
sdk: gradio
sdk_version: 4.16.0
app_file: app.py
pinned: false
license: apache-2.0

n8n Workflow Generator

Generate n8n workflows using natural language! This app uses a fine-tuned Qwen2.5-Coder-1.5B model to convert plain English descriptions into working n8n workflows.

🎯 Performance

• Overall Test Score: 92.4%
• Training Examples: 247 curated workflows
• Validation Examples: 44

🚀 Features

• Natural Language Input: Describe workflows in plain English
• Visual Preview: See your workflow nodes and connections
• TypeScript DSL: Get clean, production-ready code
• n8n JSON Export: Import directly into your n8n instance
• Adjustable Parameters: Control creativity and output length

📊 Test Results by Category

Category	Score
Basic Workflows	100%
Complexity	96%
Error Handling	80%
Loops	67%
Branching	67%
Overall	92.4%

💡 Example Prompts

Try these prompts to see what the model can do:

• "Create a webhook that sends data to Slack"
• "Schedule that runs daily and backs up database to Google Drive"
• "Webhook receives form data, validates email, saves to Airtable"
• "Monitor RSS feed and post new items to Twitter"
• "Fetch GitHub issues, if priority is high send to Slack, else email"

🛠️ How It Works

1. Input: You describe your workflow in natural language
2. Generation: Fine-tuned Qwen2.5-Coder-1.5B generates TypeScript DSL
3. Conversion: Code is converted to n8n JSON format
4. Visualization: Workflow structure is displayed visually
5. Export: Copy and import into your n8n instance

📝 Model Details

• Base Model: Qwen/Qwen2.5-Coder-1.5B-Instruct
• Fine-tuning Method: LoRA (Low-Rank Adaptation)
• Dataset: Curated n8n workflows from GitHub
• Training Framework: Transformers + PEFT

🎓 Training Details

• LoRA Rank: 16
• LoRA Alpha: 32
• Learning Rate: 2e-4
• Training Strategy: Early stopping with validation monitoring
• Hardware: NVIDIA Tesla T4 GPU

📖 Usage Tips

• Be specific: More details = better results
• Use n8n terminology: Mention specific nodes like "webhook", "Slack", "HTTP Request"
• Describe the flow: "When X happens, do Y, then Z"
• Adjust temperature: Lower (0.1-0.3) for consistency, higher (0.5-0.8) for creativity

🔧 Limitations

• Works best with common n8n patterns
• May struggle with very complex branching (>5 conditions)
• Advanced error handling might need manual refinement
• Custom node configurations may need adjustment

📄 License

Apache 2.0

🙏 Acknowledgments

Built with:

• Qwen2.5-Coder
• Hugging Face Transformers
• PEFT
• Gradio