Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
A command-line interface to generate textual and conversational datasets with LLMs.
| Date | Stars |
|---|---|
| 2026-07-31 | 300 |
| 2026-08-06 | 300 |
Today
— stars today
This week
— stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.00%/day
# datasetGPT
`datasetGPT` is a command-line interface and a Python library for inferencing Large Language Models to generate textual datasets.
Possible use cases may include:
- Constructing textual corpora to train/fine-tune detectors for content written by AI.
- Collecting datasets of LLM-produced conversations for research purposes, analysis of AI performance/impact/ethics, etc.
- Automating a task that a LLM can handle over big amounts of input texts. For example, using GPT-3 to summarize 1000 paragraphs with a single CLI command.
- Leveraging APIs of especially big LLMs to produce diverse texts for a specific task and then fine-tune a smaller model with them.
> This tool is distributed freely and doesn't imply any restrictions on the downstream use cases.
> However, you should make sure to follow the **Terms of use** of the backend APIs (OpenAI, Cohere, Petals, etc.) in your specific context.
## Installation
```
pip install datasetGPT
```
Most of the generation features rely on third-party APIs. Install their respective packages:
```
pip install openai cohere petals
```
## Usage examples
### Inference LLMs at scale
```bash
export OPENAI_API_KEY="..."
export COHERE_API_KEY="..."
datasetGPT texts \
--prompt "If {country} was a planet in the Star Wars universe it would be called" \
--backend "openai|text-davinci-003" \
--backend "cohere|medium" \
--temperature 0.9 \
--option country Germany \
--option country France \
--max-length 50 \
--num-samples 1 \
--single-file
```
The command above should produce a dataset file with 4 texts. Each possible combination of options is used for each of the backend LLMs. Check out the [CLI reference](#cli-reference) for more details. A dataset file looks like this:
```json
[
{
"sample_id": 0,
"backend": "openai|text-davinci-003",
"max_length": 50,
"temperature": 0.9,
"country": "Germany",
"prompt": "If Germany was a planet in the Star Wars universe it would be called",
"output": " Euron. The planet would be home to a powerful and diverse species of aliens, known as the Eurons, that have evolved to a higher level of understanding and technological advancement compared to many of the other planets in the galaxy. The planet would be"
},
{
"sample_id": 0,
"backend": "openai|text-davinci-003",
"max_length": 50,
"temperature": 0.9,
"country": "France",
"prompt": "If France was a planet in the Star Wars universe it would be called",
"output": " The Empire of Liberty. It would be a peaceful, democratic planet with a strong sense of justice and equality. The planet would be home to many different species of aliens but the majority of its population would be humans. It would have a strong military and"
},
{
"sample_id": 0,
"backend": "cohere|medium",
"max_length": 50,
"temperature": 0.9,
"country": "Germany",
"prompt": "If Germany was a planet in the Star Wars universe it would be called",
"output": " the Hoth of the universe.\nAfter the Soviet invasion of Eastern Europe and the subsequent Western anti-Soviet sentiment, Germany's arms manufacturers went into hyperdrive and the country churned out guns at a frightening pace. By the early 1930"
},
...
]
```
Alternatively, you can use our [`TextsGenerator`](https://github.com/radi-cho/datasetGPT/blob/main/datasetGPT/texts.py#L27) to produce texts in Python.
### Generate conversations with the ChatGPT API
```bash
export OPENAI_API_KEY="..."
datasetGPT conversations \
--length 4 \
--length 5 \
--agent1 "You're a shop assistant in a pet store. Answer to customer questions politely." \
--agent2 "You're a customer in a pet store. You should behave like a human. You want to buy {n} pets. Ask questions about the pets in the store." \
--temperature 0.1 \
--temperature 0.2 \
--option n 2 \
Excerpt of 13,185 characters
Read on GitHubWould you bet a product on this? Bounded 0–100 and slow moving.
matched fp:64698fed54e172fb, topic:large-language-models