muhammadravi251001/restructured-glaive-function-calling-v2

---
language:
- en
tags:
- function-calling
- tool-use
- llm
- instruction-tuning
- synthetic
task_categories:
- text-generation
size_categories:
- 100K<n<1M
license: apache-2.0
---

# Glaive Function Calling V2 (Structured)

This dataset is a **cleaned and structured version** of the original
[Glaive Function Calling V2](https://huggingface.co/datasets/glaiveai/glaive-function-calling-v2).

The goal of this dataset is to make the conversations **easier to use for training tool-calling / function-calling language models**, such as:

- Llama
- Qwen
- Mistral
- DeepSeek
- other OpenAI-compatible tool calling models

The original dataset stores conversations as raw text.
This version converts them into a **structured message format** suitable for modern LLM training pipelines.

---

# Dataset Features

Each row contains the following fields:

| Column | Description |
|------|-------------|
| `system` | System prompt defining assistant behavior |
| `messages` | Structured conversation messages |
| `tools` | List of tool schemas available to the assistant |
| `num_tools` | Number of tools available |
| `num_messages` | Number of messages in the conversation |
| `num_tool_calls` | Total number of tool calls made by the assistant |
| `has_tools` | Whether tools are available in the environment |
| `has_tool_call` | Whether the assistant used a tool |

---

# Example Row

```json
{
"system": "You are a helpful AI assistant...",
"tools": [
{
"name": "get_weather",
"description": "Get weather information",
"parameters": {
"type": "object",
"properties": {
"city": {"type": "string"}
}
}
}
],
"messages": [
{"role": "user", "content": "What's the weather in Tokyo?"},
{
"role": "assistant",
"tool_calls": [
{
"name": "get_weather",
"arguments": {"city": "Tokyo"}
}
]
},
{"role": "tool", "content": "{\"temperature\": 25}"}
],
"num_tools": 1,
"num_messages": 3,
"num_tool_calls": 1,
"has_tools": true,
"has_tool_call": true
}
````

---

# Source Dataset

Original dataset:

**GlaiveAI Function Calling V2**

[https://huggingface.co/datasets/glaiveai/glaive-function-calling-v2](https://huggingface.co/datasets/glaiveai/glaive-function-calling-v2)

All credit for the data goes to the original authors.

This version only restructures the dataset for easier machine learning usage.

---

# Notes

This dataset performs the following preprocessing steps:

* extraction of tool schemas from system prompts
* conversion of raw chat text into structured messages
* detection of tool calls
* generation of additional metadata fields

The goal is to provide a **cleaner dataset format for LLM training pipelines**.