Create New Dataset

Learn to create datasets to do experimentations on them

Using SDK

Import your data using Future AGI SDK.

Upload File

Upload CSV, JSON or JSONL files.

Synthetic Data Generation

Synthetically generate data.

Manual Creation

Manually create dataset from scratch.

Import from Hugging Face

Import datasets directly from Hugging Face.

From Existing Dataset

Create a subset from an existing dataset.

---

Using SDK

Use SDK to import your data to Future AGI.

1. Assign Dataset Name

   Assign a name to your dataset and click on “Next” to proceed.

   
2. Add Rows via SDK

   You will be greeted with a screen containing code snippet to add rows to your dataset.

   Python

   # pip install futureagi
   
   import os
   from fi.datasets import Dataset
   from fi.datasets.types import (
       Cell,
       Column,
       DatasetConfig,
       DataTypeChoices,
       ModelTypes,
       Row,
       SourceChoices,
   )
   
   # Set environment variables
   os.environ["FI_API_KEY"] = "<fi_api_key>"
   os.environ["FI_SECRET_KEY"] = "<fi_secret_key>"
   
   # Get existing dataset
   config = DatasetConfig(name="my-dataset", model_type= ModelTypes.GENERATIVE_LLM)
   dataset = Dataset(dataset_config=config)
   dataset = Dataset.get_dataset_config("my-dataset")
   
   # Define columns
   columns = [
       Column(
           name="user_query",
           data_type=DataTypeChoices.TEXT,
           source=SourceChoices.OTHERS
       ),
       Column(
           name="response_quality",
           data_type=DataTypeChoices.INTEGER,
           source=SourceChoices.OTHERS
       ),
       Column(
           name="is_helpful",
           data_type=DataTypeChoices.BOOLEAN,
           source=SourceChoices.OTHERS
       )
   ]
   
   # Define rows
   rows = [
       Row(
           order=1,
           cells=[
               Cell(column_name="user_query", value="What is machine learning?"),
               Cell(column_name="response_quality", value=8),
               Cell(column_name="is_helpful", value=True)
           ]
       ),
       Row(
           order=2,
           cells=[
               Cell(column_name="user_query", value="Explain quantum computing"),
               Cell(column_name="response_quality", value=9),
               Cell(column_name="is_helpful", value=True)
           ]
       )
   ]
   
   try:
       # Add columns and rows to dataset
       dataset = dataset.add_columns(columns=columns)
       dataset = dataset.add_rows(rows=rows)
       print("✓ Data added successfully")
   
   except Exception as e:
       print(f"Failed to add data: {e}")
   

   Click here to access API Key and Secret Key.

---

Upload File

1. Upload CSV, JSON or JSONL files to create a dataset.

   

---

Using Synthetic Data Generation

Synthetically generate data and perform experimentations on it.

1. Add Details

   Provide basic details about the dataset you want to generate.

   

   Property	Description
   Name	Name of the dataset
   Knowledge Base (optional)	Select which knowledge base you want to use.
   Description	Describe the dataset you want to generate
   Objective (optional)	Use case of the dataset
   Pattern (optional)	Style, tone or behavioral traits of the generated dataset
   No. of Rows	Row count of the generated dataset (min 10 rows)

2. Add Column Properties

   Define column types and properties

   

   Property	Description
   Column Name	Name of the column
   Column Type	Choose the type of the column (available types: text, boolean, integer, float, json, array, datetime)

3. Add Description

   Now add description for each column. Describe in detail what values you want in this column.

   
4. Create Dataset

   Click on “Create Dataset” button to generate the dataset. Your synthetic dataset will be generated in a few seconds and will be available in your dataset dashboard.

   If you are not satisfied with the generated dataset, you can click on “Configure Synthetic Data” button. It will allow you to edit the fields and generate the dataset again.

   

---

Manually Create Dataset

Manually create dataset from scratch.

1. Provide Basic Details

   To proceed with creating dataset manually from scratch, provdide the name you want to assign and the number of columns and rows you want.

   

   You will be greeted with an empty dataset with the name you assigned and with empty rows and columns.

   
2. Populating the dataset

   You can populate the dataset by double-tapping over the empty cell you want to populate. It will open an editor where you can provide the details you want to fill in that cell.

   

Importing from Hugging Face

1. Search Hugging Face Dataset

   Search for the dataset you want to import from Hugging Face. You can even refine the search by using flters given on left side.

   
2. Import Dataset

   Once you have selected the dataset you want to import, click on that dataset and it will open a panel where you can select what subset and split you want to import.

   You can also select the number of rows you want to import. By default, it will import all the rows.

   

   Click on “Start Experimenting” button and it willl start importing the dataset and you will be able to see it in your dataset dashboard.

---

From Existing Dataset

You can create a subset from an existing dataset.

1. Choose the existing dataset

   Assign a name to this dataset and choose the existing dataset from the dropdown you want to create a subset from.

   

   It allows you to import the dataset in two ways:

   1. Import Data: It will only import the original columns from the existing dataset.
   2. Import Data and Prompt Configuration: Along with original column, it will also import the prompt columns from that dataset.
2. Map the columns

   You can choose what columns you want to use from that existing dataset and also you can assign a new name to the columns you want to use.

   
3. Import the dataset

   Click on “Add” button and it will create a new dataset in your dataset dashboard.