This project implements a modular pipeline to automate the scraping, rewriting, and review of online book chapters with human-in-the-loop collaboration. It integrates AI writing and reviewing agents, human feedback loops, reinforcement-based decision logic, and support for versioning and future semantic search.

Web scraping with screenshots: Extracts chapter content from a given URL using Playwright and BeautifulSoup.

AI Writing and Reviewing: Uses instruction-tuned large language models (LLMs) to rewrite ("spin") chapters and optionally refine them.

Human-in-the-loop editing: Enables users to accept, reject, or edit AI-generated content before finalizing.

Reward-based decision loop: Accept/reject actions are passed to a reward model for feedback, simulating reinforcement learning inference.

Versioning support: Saves chapter versions for traceability and future comparison.

Agentic flow: Each functional step is isolated and can be expanded into callable APIs or microservices.

Python 3.10+

Playwright for web scraping and screenshots

Hugging Face Transformers for LLM-based rewriting

ChromaDB (planned or stubbed) for vector-based content versioning and semantic search

Custom reward logic simulating an RL feedback loop