Introduction
============

Welcome to **Dataset Generator**, a flexible, extendable pipeline for creating realistic synthetic ticket datasets using AI assistants. Whether you need sample support tickets for testing, machine-learning training, or analytics, Dataset Generator lets you configure every step—from subject generation to multilingual translation and first-response simulation.

Key Concepts
------------

- **Graph-based pipeline**
  Each ticket is generated by a directed graph of AI “nodes” (assistants). Outputs of one node feed into the next, enabling complex workflows such as:
  1. Topic generation
  2. Email text creation
  3. Tag assignment
  4. Paraphrasing & translation
  5. First-response drafting

- **Modular Assistants**
  Every step is powered by a configurable Assistant class. You can swap models, prompts, or even injection-decorate runs for cost tracking.

- **Rich Ticket Attributes**
  Generate tickets with full metadata:
  - **Subject** & **Body**
  - **Type** (Incident, Request, etc.)
  - **Queue** (Technical Support, Billing, etc.)
  - **Priority** (low, medium, high)
  - **Language** (EN, DE, FR, ES, PT)
  - **Tags** (4–8 relevant topics)
  - **First-answer** simulation

- **Cost Analysis**
  Automatically measure token usage and per-token costs across runs, aggregate by Assistant, and output summaries in your chosen currency.

Getting Started
---------------

1. **Configure your project** in `config/config.py`:
   - Number of tickets
   - Translation nodes per ticket
   - Models, prompts, and providers

2. **Build the docs**
   ```bash
   cd docs
   make html

`Dataset Generator on GitHub <https://github.com/Softoft/dataset-generator/tree/dev>`_