README.md

---
license: mit
tags:
- n8n
- workflow
- automation
- no-code
- low-code
- axolotl
- fine-tuning
- text-generation
task_categories:
- text-generation
language:
- en
size_categories:
- 1K<n<10K
---

# n8n Workflow Templates Dataset

A curated collection of n8n workflow automation templates formatted for LLM fine-tuning in multiple formats.

## Dataset Description

This dataset contains **2,737** n8n workflow templates from the official n8n template library, formatted in multiple formats for fine-tuning LLMs to generate n8n workflows.

### Available Formats

1. **Alpaca Format** (`train` split)
   - Traditional instruction-input-output format
   - Compatible with Axolotl and other Alpaca-based training frameworks

2. **OpenAI Messages Format** (`train_openai` split)
   - Chat-based format with role-based messages
   - Compatible with OpenAI fine-tuning API and similar chat-based training

### Data Fields

#### Alpaca Format
| Field       | Description                               |
| ----------- | ----------------------------------------- |
| instruction | System prompt for n8n workflow generation |
| input       | User's workflow requirements/description  |
| output      | Complete n8n workflow JSON                |

#### OpenAI Messages Format
| Field    | Description                               |
| -------- | ----------------------------------------- |
| messages | Array of message objects with role and content |

Each message object contains:
- `role`: "user" or "assistant"
- `content`: The message content (instruction or workflow JSON)

### System Prompt

All records share the same instruction:

> You are an expert n8n workflow generation assistant. Your goal is to create valid, efficient, and error-free n8n workflow JSONs based on the user's requirements. Always output ONLY the valid JSON workflow.

## Dataset Structure

### Alpaca Format Example
```json
{
  "instruction": "You are an expert n8n workflow generation assistant...",
  "input": "Create a workflow that retrieves Google Analytics data...",
  "output": "{"id": "...", "nodes": [...], "connections": {...}}"
}
```

### OpenAI Messages Format Example
```json
{
  "messages": [
    {
      "role": "user",
      "content": "You are an expert n8n workflow generation assistant...\n\nCreate a workflow that retrieves Google Analytics data..."
    },
    {
      "role": "assistant",
      "content": "{"id": "...", "nodes": [...], "connections": {...}}"
    }
  ]
}
```

## Loading the Dataset

### Using Hugging Face Datasets

```python
from datasets import load_dataset

# Load Alpaca format
alpaca_dataset = load_dataset("mbakgun/n8nbuilder-n8n-workflows-dataset", split="train")

# Load OpenAI format
openai_dataset = load_dataset("mbakgun/n8nbuilder-n8n-workflows-dataset", split="train_openai")

# Load both
full_dataset = load_dataset("mbakgun/n8nbuilder-n8n-workflows-dataset")
```

### Using Hugging Face Hub

```python
from huggingface_hub import hf_hub_download
import json

# Download Alpaca format
alpaca_file = hf_hub_download(
    repo_id="mbakgun/n8nbuilder-n8n-workflows-dataset",
    filename="train.jsonl",
    repo_type="dataset"
)

# Download OpenAI format
openai_file = hf_hub_download(
    repo_id="mbakgun/n8nbuilder-n8n-workflows-dataset",
    filename="train_openai.jsonl",
    repo_type="dataset"
)
```

## Fine-tuning

### With Axolotl (Alpaca Format)

```yaml
datasets:
  - path: mbakgun/n8nbuilder-n8n-workflows-dataset
    type: alpaca
    split: train
```

### With OpenAI Fine-tuning API (OpenAI Format)

```python
from openai import OpenAI

client = OpenAI()

# Prepare data from Hugging Face
dataset = load_dataset("mbakgun/n8nbuilder-n8n-workflows-dataset", split="train_openai")

# Convert to OpenAI format and upload
training_file = client.files.create(
    file=open("training_data.jsonl", "rb"),
    purpose="fine-tune"
)
```

### With Other Frameworks

Both formats can be easily converted to other training formats as needed.

## Use Cases

* Fine-tuning LLMs to generate n8n workflows from natural language
* Training models to understand workflow automation patterns
* Building AI assistants for no-code/low-code automation
* Research on code generation and workflow automation

## Statistics

| Metric          | Value          |
| --------------- | -------------- |
| Total Workflows | 2,737          |
| Formats         | 2 (Alpaca, OpenAI Messages) |
| Splits          | train (Alpaca), train_openai (OpenAI) |

## License

MIT License

## Acknowledgments

This dataset is currently maintained by n8nbuilder.dev — an AI-powered n8n workflow generation tool.

### Data Sources & Attribution

* **Template Source**: All workflow templates in this dataset are sourced from n8n's public template gallery. Template creators retain all rights to their workflows.
* **Indexing**: Templates were indexed using n8n-mcp by @czlonkowski.
* **n8n**: n8n is the workflow automation platform that powers these templates.

If you are a template creator and have concerns about your template being included, please open an issue.

## Citation

If you use this dataset, please cite:

```bibtex
@dataset{n8nbuilder_n8n_workflows_dataset_2025,
  title={n8nbuilder - n8n workflow automation templates dataset},
  year={2025},
  url={https://huggingface.co/datasets/mbakgun/n8nbuilder-n8n-workflows-dataset},
  note={Curated by n8nbuilder.dev}
}
```