Version 0.1.2

GitHub Release: WizardSData v0.1.2

Overview

Version 0.1.2 of WizardSData introduces new templates for financial retail banking scenarios, expanded profile datasets, and bug fixes to improve dataset generation.

What's New

Added

• New Financial Retail Banking Template: Enhanced conversation generation with specialized templates for diverse banking scenarios, available in /templates/financial_retail_banking/prompts/
• Expanded Profile Dataset: Added 160 comprehensive profiles for retail banking covering multiple topics like mortgages, investments, savings, and insurance in /templates/financial_retail_banking/profiles/
• Complete Generated Dataset: Included a full dataset with 1700 rows spanning 320 unique conversations in /templates/financial_retail_banking/outputs/

Fixed

• Topic Storage: Topic information is now properly stored in generated datasets, making categorization and filtering easier
• Sequence Numbering: Fixed sequence number progression in dataset generation for more accurate conversation tracking

Getting Started

To use the new templates, update to the latest version:

pip install --upgrade wizardsdata

Documentation

For detailed usage of the new templates, please refer to the documentation at: https://peremartra.github.io/WizardSData/

Feedback

If you encounter any issues or have suggestions for improvement, please file an issue on the GitHub repository.

---

Thank you for using WizardSData! This release continues our mission to provide high-quality synthetic conversation datasets for training and evaluating language models.

WizardSData 0.1.1 - First Public Release

WizardSData 0.1.1

We're excited to announce the first public release of WizardSData, a Python library for generating conversation datasets using language models.

Overview

WizardSData provides a simple interface for configuring, generating, and saving conversations between different roles (such as a financial advisor and a client). This makes it easy to create realistic conversation datasets for training, testing, or demonstration purposes.

Key Features

• Simple Configuration: Configure conversation parameters with a straightforward API
• Template System: Use Jinja2 templates to customize conversation prompts
• Profile Management: Define conversation profiles with different characteristics
• Automatic Generation: Generate realistic conversations using OpenAI models
• Structured Output: Save conversations in a well-organized JSON format
• Command-line Interface: Use WizardSData directly from the command line

Documentation

Full documentation is available at: https://yourusername.github.io/wizardsdata/

Installation

pip install wizardsdata