aswansong/novelagent
🤖 AI Agentaswansong
A human-in-the-loop multi-agent system for generating consistent, long-form web novels using RAG and LLMs.
NovelAgent is a sophisticated Python-based framework built on FastAPI, LangChain, and LangGraph, specifically engineered for long-form narrative generation. It addresses the 'context window' and 'logical drift' challenges inherent in standard LLM text generation by utilizing a multi-agent architecture. Each agent is assigned specific roles—such as plot planning, character management, and scene drafting—which work in concert to maintain narrative integrity over extended lengths.
The system integrates Retrieval-Augmented Generation (RAG) to keep track of world-building details, character arcs, and previous plot developments, effectively acting as a 'long-term memory' for the AI. A key innovation is the human-in-the-loop design, which forces the system to pause at critical junctures, providing the user with an interface to audit, edit, or steer the story direction before the agents proceed. This hybrid approach ensures that while the AI handles the heavy lifting of drafting, the creative control remains firmly with the human author.
💡Highlights
- ├─Multi-agent narrative orchestration
- ├─RAG-based long-term story memory
- └─Human-in-the-loop control flow
🎯For
- ├─Creative writers
- ├─AI researchers
- └─Narrative designers