afterimage 0.13.0

Create synthetic datasets based on your documents

AfterImage

AfterImage is a flexible Python framework for generating synthetic conversational datasets using State-of-the-Art Large Language Models (LLMs), primarily Google Gemini and OpenAI-compatible APIs.

It is designed to be highly customizable, enabling tailored instruction generation, persona-based simulation, diverse document ingestion, and context-aware conversation generation.

🚀 Getting Started

To get started with AfterImage, please refer to the Quickstart Guide. It covers:

• Installation: How to set up the library.
• Basic Usage: Generating simple conversations.
• RAG & Context: Using documents to drive questions.
• Personas: Creating varied user personas for realistic datasets.

📖 Documentation

• Quickstart Guide: The best place to start.
• Design & Architecture: Understanding the core concepts and codebase structure.
• Examples: Examples of how to use AfterImage.

📦 Installation

pip install git+https://github.com/altaidevorg/afterimage.git

Optional extras (see pyproject.toml):

• embeddings-local — sentence-transformers for ProcessEmbeddingProvider, QdrantRetriever (by model name), and QualityChecker semantic checks.
• server — FastAPI app (afterimage-server).
• training — Torch/TRL stack for examples/demo_ui/training_scripts/train.py.

pip install "afterimage[embeddings-local]@git+https://github.com/altaidevorg/afterimage.git"

✨ Key Features

• Async-First: High-performance parallel generation.
• Persona Simulation: Realistic user diversity.
• Context-Aware: Grounds conversations in your documents (RAG).
• Multi-Modal: Support for text and potentially other modalities in future.
• Monitoring: Real-time generation metrics and alerts.