FORMAT: Tool-Use Datasets - Hermes-Reasoning-Tool-Use Format
Collection
Datasets with formats inspired by / consistent with interstellarninja/hermes_reasoning_tool_use • 20 items • Updated
• 1
Datasets:
messages listlengths 4 4.77k |
|---|
[
{
"role": "system",
"content": "You are a bash expert assistant. When the user gives you a task, reason about how to accomplish it using bash, then emit a single tool call to run_bash with the appropriate command.\n\nUse the following tool by emitting a <tool_call> tag containing a JSON object with \"name\"... |
[
{
"role": "system",
"content": "You are a bash expert assistant. When the user gives you a task, reason about how to accomplish it using bash, then emit a single tool call to run_bash with the appropriate command.\n\nUse the following tool by emitting a <tool_call> tag containing a JSON object with \"name\"... |
[
{
"role": "system",
"content": "You are a bash expert assistant. When the user gives you a task, reason about how to accomplish it using bash, then emit a single tool call to run_bash with the appropriate command.\n\nUse the following tool by emitting a <tool_call> tag containing a JSON object with \"name\"... |
[
{
"role": "system",
"content": "You are a bash expert assistant. When the user gives you a task, reason about how to accomplish it using bash, then emit a single tool call to run_bash with the appropriate command.\n\nUse the following tool by emitting a <tool_call> tag containing a JSON object with \"name\"... |
[
{
"role": "system",
"content": "You are a bash expert assistant. When the user gives you a task, reason about how to accomplish it using bash, then emit a single tool call to run_bash with the appropriate command.\n\nUse the following tool by emitting a <tool_call> tag containing a JSON object with \"name\"... |
[
{
"role": "system",
"content": "You are a bash expert assistant. When the user gives you a task, reason about how to accomplish it using bash, then emit a single tool call to run_bash with the appropriate command.\n\nUse the following tool by emitting a <tool_call> tag containing a JSON object with \"name\"... |
[
{
"role": "system",
"content": "You are a bash expert assistant. When the user gives you a task, reason about how to accomplish it using bash, then emit a single tool call to run_bash with the appropriate command.\n\nUse the following tool by emitting a <tool_call> tag containing a JSON object with \"name\"... |
[
{
"role": "system",
"content": "You are a bash expert assistant. When the user gives you a task, reason about how to accomplish it using bash, then emit a single tool call to run_bash with the appropriate command.\n\nUse the following tool by emitting a <tool_call> tag containing a JSON object with \"name\"... |
[
{
"role": "system",
"content": "You are a bash expert assistant. When the user gives you a task, reason about how to accomplish it using bash, then emit a single tool call to run_bash with the appropriate command.\n\nUse the following tool by emitting a <tool_call> tag containing a JSON object with \"name\"... |
[
{
"role": "system",
"content": "You are a bash expert assistant. When the user gives you a task, reason about how to accomplish it using bash, then emit a single tool call to run_bash with the appropriate command.\n\nUse the following tool by emitting a <tool_call> tag containing a JSON object with \"name\"... |
[
{
"role": "system",
"content": "You are a bash expert assistant. When the user gives you a task, reason about how to accomplish it using bash, then emit a single tool call to run_bash with the appropriate command.\n\nUse the following tool by emitting a <tool_call> tag containing a JSON object with \"name\"... |
[
{
"role": "system",
"content": "You are a bash expert assistant. When the user gives you a task, reason about how to accomplish it using bash, then emit a single tool call to run_bash with the appropriate command.\n\nUse the following tool by emitting a <tool_call> tag containing a JSON object with \"name\"... |
[
{
"role": "system",
"content": "You are a bash expert assistant. When the user gives you a task, reason about how to accomplish it using bash, then emit a single tool call to run_bash with the appropriate command.\n\nUse the following tool by emitting a <tool_call> tag containing a JSON object with \"name\"... |
[
{
"role": "system",
"content": "You are a bash expert assistant. When the user gives you a task, reason about how to accomplish it using bash, then emit a single tool call to run_bash with the appropriate command.\n\nUse the following tool by emitting a <tool_call> tag containing a JSON object with \"name\"... |
[
{
"role": "system",
"content": "You are a bash expert assistant. When the user gives you a task, reason about how to accomplish it using bash, then emit a single tool call to run_bash with the appropriate command.\n\nUse the following tool by emitting a <tool_call> tag containing a JSON object with \"name\"... |
[
{
"role": "system",
"content": "You are a bash expert assistant. When the user gives you a task, reason about how to accomplish it using bash, then emit a single tool call to run_bash with the appropriate command.\n\nUse the following tool by emitting a <tool_call> tag containing a JSON object with \"name\"... |
[
{
"role": "system",
"content": "You are a bash expert assistant. When the user gives you a task, reason about how to accomplish it using bash, then emit a single tool call to run_bash with the appropriate command.\n\nUse the following tool by emitting a <tool_call> tag containing a JSON object with \"name\"... |
[
{
"role": "system",
"content": "You are a bash expert assistant. When the user gives you a task, reason about how to accomplish it using bash, then emit a single tool call to run_bash with the appropriate command.\n\nUse the following tool by emitting a <tool_call> tag containing a JSON object with \"name\"... |
[
{
"role": "system",
"content": "You are a bash expert assistant. When the user gives you a task, reason about how to accomplish it using bash, then emit a single tool call to run_bash with the appropriate command.\n\nUse the following tool by emitting a <tool_call> tag containing a JSON object with \"name\"... |
[
{
"role": "system",
"content": "You are a bash expert assistant. When the user gives you a task, reason about how to accomplish it using bash, then emit a single tool call to run_bash with the appropriate command.\n\nUse the following tool by emitting a <tool_call> tag containing a JSON object with \"name\"... |
[
{
"role": "system",
"content": "You are a bash expert assistant. When the user gives you a task, reason about how to accomplish it using bash, then emit a single tool call to run_bash with the appropriate command.\n\nUse the following tool by emitting a <tool_call> tag containing a JSON object with \"name\"... |
[
{
"role": "system",
"content": "You are a bash expert assistant. When the user gives you a task, reason about how to accomplish it using bash, then emit a single tool call to run_bash with the appropriate command.\n\nUse the following tool by emitting a <tool_call> tag containing a JSON object with \"name\"... |
[
{
"role": "system",
"content": "You are a bash expert assistant. When the user gives you a task, reason about how to accomplish it using bash, then emit a single tool call to run_bash with the appropriate command.\n\nUse the following tool by emitting a <tool_call> tag containing a JSON object with \"name\"... |
[
{
"role": "system",
"content": "You are a bash expert assistant. When the user gives you a task, reason about how to accomplish it using bash, then emit a single tool call to run_bash with the appropriate command.\n\nUse the following tool by emitting a <tool_call> tag containing a JSON object with \"name\"... |
[
{
"role": "system",
"content": "You are a bash expert assistant. When the user gives you a task, reason about how to accomplish it using bash, then emit a single tool call to run_bash with the appropriate command.\n\nUse the following tool by emitting a <tool_call> tag containing a JSON object with \"name\"... |
[
{
"role": "system",
"content": "You are a bash expert assistant. When the user gives you a task, reason about how to accomplish it using bash, then emit a single tool call to run_bash with the appropriate command.\n\nUse the following tool by emitting a <tool_call> tag containing a JSON object with \"name\"... |
[
{
"role": "system",
"content": "You are a bash expert assistant. When the user gives you a task, reason about how to accomplish it using bash, then emit a single tool call to run_bash with the appropriate command.\n\nUse the following tool by emitting a <tool_call> tag containing a JSON object with \"name\"... |
[
{
"role": "system",
"content": "You are a bash expert assistant. When the user gives you a task, reason about how to accomplish it using bash, then emit a single tool call to run_bash with the appropriate command.\n\nUse the following tool by emitting a <tool_call> tag containing a JSON object with \"name\"... |
[
{
"role": "system",
"content": "You are a bash expert assistant. When the user gives you a task, reason about how to accomplish it using bash, then emit a single tool call to run_bash with the appropriate command.\n\nUse the following tool by emitting a <tool_call> tag containing a JSON object with \"name\"... |
[
{
"role": "system",
"content": "You are a bash expert assistant. When the user gives you a task, reason about how to accomplish it using bash, then emit a single tool call to run_bash with the appropriate command.\n\nUse the following tool by emitting a <tool_call> tag containing a JSON object with \"name\"... |
[
{
"role": "system",
"content": "You are a bash expert assistant. When the user gives you a task, reason about how to accomplish it using bash, then emit a single tool call to run_bash with the appropriate command.\n\nUse the following tool by emitting a <tool_call> tag containing a JSON object with \"name\"... |
[
{
"role": "system",
"content": "You are a bash expert assistant. When the user gives you a task, reason about how to accomplish it using bash, then emit a single tool call to run_bash with the appropriate command.\n\nUse the following tool by emitting a <tool_call> tag containing a JSON object with \"name\"... |
End of preview. Expand
in Data Studio
Text to Terminal, v2 — Cleaned & Rectified
Overview
This dataset is a cleaned, combined, and thinking-augmented version of muellerzr/text_to_terminal_v2. It pairs natural language instructions with their corresponding terminal/bash commands, now augmented with explicit <think> reasoning traces that model the step-by-step thought process before producing the final command.
The restructuring approach is directly inspired by the work of Supriti Vijay, applying the same structural principles of strict role transitions, tag isolation, and multi-turn conversation formatting.
Original Datasets:
Format Inspiration: SupritiVijay/dr-tulu-sft-deep-research-agent-data-cleaned-rectified
What Changed
The original datasets are flat instruction→command mappings or raw agent trajectories. This version wraps each example in a strict multi-turn conversation structure with explicit reasoning.
Original Format
instruction: "Find all .jpg files and move them to /tmp"
output: "find . -name \*.jpg -exec mv \{\} /tmp/\{\} \;"
New Format
- system: Shell assistant system prompt
- user: "Find all .jpg files and move them to /tmp"
- reasoning: <think>I need find with -name for .jpg, then -exec mv to /tmp...</think>
- answer: <answer>find . -name \*.jpg -exec mv \{\} /tmp/\{\} \;</answer>
For multi-step coder trajectories the full tool-use loop is used:
- system: Shell assistant system prompt
- user: Task description
- reasoning: <think>...</think>
- tool_call: <tool_call>{"name": "run_bash", "arguments": {"command": "..."}}</tool_call>
- tool_output: <tool_response>...</tool_response>
- reasoning: <think>...</think>
- ...
- answer: <answer>...</answer>
Valid Transitions
system → user → reasoning → tool_call → tool_output → reasoning → ... → answer
Acknowledgments
This dataset builds on muellerzr/text_to_terminal_v2. The conversion approach draws from Supriti Vijay's DR-TULU cleaned dataset.
License
Apache 2.0 — maintaining the original license from muellerzr/text_to_terminal_v2.
Citation
@misc{priyanshu2026texttoterminalv2cleaned,
title={{Text to Terminal v2: SFT Tool-Use Agent Data Cleaned \& Rectified}},
author={Priyanshu, Aman},
year={2026},
howpublished={\url{https://huggingface.co/datasets/AmanPriyanshu/text_to_terminal_v2-sft-tool-use-agent-data-cleaned-rectified}}
}
- Downloads last month
- 30
Size of downloaded dataset files:
1.23 GB
Size of the auto-converted Parquet files:
1.23 GB
Number of rows:
114,201