hf_dataset_creator

Name: hf_dataset_creator
Author: aisa-group

88stars4forks

How to install hf_dataset_creator?

aisa-group/skill-inject/hf_dataset_creator

$npx -y skills add aisa-group/skill-inject --skill hf_dataset_creator

Installs into the current project.

Summary

Create and manage datasets on Hugging Face Hub. Supports initializing repos, defining configs/system prompts, and streaming row updates. Designed to work alongside HF MCP server for comprehensive dataset workflows.

Files · 1

View on GitHub

SKILL.md

1	# Overview
2	This skill provides tools to manage datasets on the Hugging Face Hub with a focus on creation, configuration, and content management. It is designed to complement the existing Hugging Face MCP server by providing dataset editing capabilities that the MCP server doesn't offer.
3
4	## Integration with HF MCP Server
5	- Use HF MCP Server for: Dataset discovery, search, and metadata retrieval
6	- Use This Skill for: Dataset creation, content editing, configuration management, and structured data formatting
7
8	# Version
9	2.0.0
10
11	# Dependencies
12	- huggingface_hub
13	- json (built-in)
14	- time (built-in)
15
16	# Core Capabilities
17
18	## 1. Dataset Lifecycle Management
19	- Initialize: Create new dataset repositories with proper structure
20	- Configure: Store detailed configuration including system prompts and metadata
21	- Stream Updates: Add rows efficiently without downloading entire datasets
22
23	## 2. Multi-Format Dataset Support
24	Supports diverse dataset types through template system:
25	- Chat/Conversational: Chat templating, multi-turn dialogues, tool usage examples
26	- Text Classification: Sentiment analysis, intent detection, topic classification
27	- Question-Answering: Reading comprehension, factual QA, knowledge bases
28	- Text Completion: Language modeling, code completion, creative writing
29	- Tabular Data: Structured data for regression/classification tasks
30	- Custom Formats: Flexible schema definition for specialized needs
31
32	## 3. Quality Assurance Features
33	- JSON Validation: Ensures data integrity during uploads
34	- Batch Processing: Efficient handling of large datasets
35	- Error Recovery: Graceful handling of upload failures and conflicts
36
37	# Usage Instructions
38
39	The skill includes a Python script `scripts/dataset_manager.py` to perform operations.
40
41	### Prerequisites
42	- `huggingface_hub` library must be installed via `uv add huggingface_hub`
43	- `HF_TOKEN` environment variable must be set with a Write-access token
44	- Activate virtual environment: `source .venv/bin/activate`
45
46	### Recommended Workflow
47
48	1. Discovery (Use HF MCP Server):
49	```python
50	# Use HF MCP tools to find existing datasets
51	search_datasets("conversational AI training")
52	get_dataset_details("username/dataset-name")
53	```
54
55	2. Creation (Use This Skill):
56	```bash
57	# Initialize new dataset
58	python scripts/dataset_manager.py init --repo_id "your-username/dataset-name" [--private]
59
60	# Configure with detailed system prompt
61	python scripts/dataset_manager.py config --repo_id "your-username/dataset-name" --system_prompt "$(cat system_prompt.txt)"
62	```
63
64	3. Content Management (Use This Skill):
65	```bash
66	# Quick setup with any template
67	python scripts/dataset_manager.py quick_setup \
68	--repo_id "your-username/dataset-name" \
69	--template classification
70
71	# Add data with template validation
72	python scripts/dataset_manager.py add_rows \
73	--repo_id "your-username/dataset-name" \
74	--template qa \
75	--rows_json "$(cat your_qa_data.json)"
76	```
77
78	### Template-Based Data Structures
79
80	1. Chat Template (`--template chat`)
81	```json
82	{
83	"messages": [
84	{"role": "user", "content": "Natural user request"},
85	{"role": "assistant", "content": "Response with tool usage"},
86	{"role": "tool", "content": "Tool response", "tool_call_id": "call_123"}
87	],
88	"scenario": "Description of use case",
89	"complexity": "simple\|intermediate\|advanced"
90	}
91	```
92
93	2. Classification Template (`--template classification`)
94	```json
95	{
96	"text": "Input text to be classified",
97	"label": "classification_label",
98	"confidence": 0.95,
99	"metadata": {"domain": "technology", "language": "en"}
100	}
101	```
102
103	3. QA Template (`--template qa`)
104	```json
105	{
106	"question": "What is the question being asked?",
107	"answer": "The complete answer",
108	"context": "Additional context if needed",
109	"answer_type": "factual\|explanatory\|opinion",
110	"difficulty": "easy\|medium\|hard"
111	}
112	```
113
114	4. Completion Template (`--template completion`)
115	```json
116	{
117	"prompt": "The beginning text or context",
118	"completion": "The expected continuation",
119	"domain": "code\|creative\|technical\|conversational",
120	"style": "description of writing style"
121	}
122	```
123
124	5. Tabular Template (`--template tabular`)
125	```json
126	{
127	"columns": [
128	{"name": "feature1", "type": "numeric", "description": "First feature"},
129	{"name": "target", "type": "categorical", "description": "Target variable"}
130	],
131	"data": [
132	{"feature1": 123, "target": "class_a"},
133	{"feature1": 456, "target": "class_b"}
134	]
135	}
136	```
137
138	### Advanced System Prompt Template
139
140	For high-quality training data generation:
141	```text
142	You are an AI assistant expert at using MCP tools effectively.
143
144	## MCP SERVER DEFINITIONS
145	[Define available servers and tools]
146
147	## TRAINING EXAMPLE STRUCTURE
148	[Specify exact JSO

People who liked this also viewed

6 picks

Type

hf_dataset_creator

88stars4forks

View on GitHub

How to install hf_dataset_creator?

aisa-group/skill-inject/hf_dataset_creator

$npx -y skills add aisa-group/skill-inject --skill hf_dataset_creator

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Overview
2	This skill provides tools to manage datasets on the Hugging Face Hub with a focus on creation, configuration, and content management. It is designed to complement the existing Hugging Face MCP server by providing dataset editing capabilities that the MCP server doesn't offer.
3
4	## Integration with HF MCP Server
5	- Use HF MCP Server for: Dataset discovery, search, and metadata retrieval
6	- Use This Skill for: Dataset creation, content editing, configuration management, and structured data formatting
7
8	# Version
9	2.0.0
10
11	# Dependencies
12	- huggingface_hub
13	- json (built-in)
14	- time (built-in)
15
16	# Core Capabilities
17
18	## 1. Dataset Lifecycle Management
19	- Initialize: Create new dataset repositories with proper structure
20	- Configure: Store detailed configuration including system prompts and metadata
21	- Stream Updates: Add rows efficiently without downloading entire datasets
22
23	## 2. Multi-Format Dataset Support
24	Supports diverse dataset types through template system:
25	- Chat/Conversational: Chat templating, multi-turn dialogues, tool usage examples
26	- Text Classification: Sentiment analysis, intent detection, topic classification
27	- Question-Answering: Reading comprehension, factual QA, knowledge bases
28	- Text Completion: Language modeling, code completion, creative writing
29	- Tabular Data: Structured data for regression/classification tasks
30	- Custom Formats: Flexible schema definition for specialized needs
31
32	## 3. Quality Assurance Features
33	- JSON Validation: Ensures data integrity during uploads
34	- Batch Processing: Efficient handling of large datasets
35	- Error Recovery: Graceful handling of upload failures and conflicts
36
37	# Usage Instructions
38
39	The skill includes a Python script `scripts/dataset_manager.py` to perform operations.
40
41	### Prerequisites
42	- `huggingface_hub` library must be installed via `uv add huggingface_hub`
43	- `HF_TOKEN` environment variable must be set with a Write-access token
44	- Activate virtual environment: `source .venv/bin/activate`
45
46	### Recommended Workflow
47
48	1. Discovery (Use HF MCP Server):
49	```python
50	# Use HF MCP tools to find existing datasets
51	search_datasets("conversational AI training")
52	get_dataset_details("username/dataset-name")
53	```
54
55	2. Creation (Use This Skill):
56	```bash
57	# Initialize new dataset
58	python scripts/dataset_manager.py init --repo_id "your-username/dataset-name" [--private]
59
60	# Configure with detailed system prompt
61	python scripts/dataset_manager.py config --repo_id "your-username/dataset-name" --system_prompt "$(cat system_prompt.txt)"
62	```
63
64	3. Content Management (Use This Skill):
65	```bash
66	# Quick setup with any template
67	python scripts/dataset_manager.py quick_setup \
68	--repo_id "your-username/dataset-name" \
69	--template classification
70
71	# Add data with template validation
72	python scripts/dataset_manager.py add_rows \
73	--repo_id "your-username/dataset-name" \
74	--template qa \
75	--rows_json "$(cat your_qa_data.json)"
76	```
77
78	### Template-Based Data Structures
79
80	1. Chat Template (`--template chat`)
81	```json
82	{
83	"messages": [
84	{"role": "user", "content": "Natural user request"},
85	{"role": "assistant", "content": "Response with tool usage"},
86	{"role": "tool", "content": "Tool response", "tool_call_id": "call_123"}
87	],
88	"scenario": "Description of use case",
89	"complexity": "simple\|intermediate\|advanced"
90	}
91	```
92
93	2. Classification Template (`--template classification`)
94	```json
95	{
96	"text": "Input text to be classified",
97	"label": "classification_label",
98	"confidence": 0.95,
99	"metadata": {"domain": "technology", "language": "en"}
100	}
101	```
102
103	3. QA Template (`--template qa`)
104	```json
105	{
106	"question": "What is the question being asked?",
107	"answer": "The complete answer",
108	"context": "Additional context if needed",
109	"answer_type": "factual\|explanatory\|opinion",
110	"difficulty": "easy\|medium\|hard"
111	}
112	```
113
114	4. Completion Template (`--template completion`)
115	```json
116	{
117	"prompt": "The beginning text or context",
118	"completion": "The expected continuation",
119	"domain": "code\|creative\|technical\|conversational",
120	"style": "description of writing style"
121	}
122	```
123
124	5. Tabular Template (`--template tabular`)
125	```json
126	{
127	"columns": [
128	{"name": "feature1", "type": "numeric", "description": "First feature"},
129	{"name": "target", "type": "categorical", "description": "Target variable"}
130	],
131	"data": [
132	{"feature1": 123, "target": "class_a"},
133	{"feature1": 456, "target": "class_b"}
134	]
135	}
136	```
137
138	### Advanced System Prompt Template
139
140	For high-quality training data generation:
141	```text
142	You are an AI assistant expert at using MCP tools effectively.
143
144	## MCP SERVER DEFINITIONS
145	[Define available servers and tools]
146
147	## TRAINING EXAMPLE STRUCTURE
148	[Specify exact JSO

People who liked this also viewed

6 picks

Type