PlotPilot is an AI-driven long-form creative writing platform designed to automate content generation, manage narrative elements through a knowledge graph, and analyze authorial style. It targets creators and technical users seeking to streamline the production of extensive written works like novels, screenplays, or technical documents, offering automated chapter generation and structured world-building tools.

How It Works

The platform employs a DDD (Domain-Driven Design) four-layer architecture, integrating context management, a knowledge system, consumption components, and state awareness across over 20 prompt nodes. Its core innovation lies in the "Autopilot" mode, a background daemon that continuously generates chapters in stages—macro planning, beat planning, chapter generation, and review—supporting SSE real-time streaming. A unified post-chapter pipeline uses a single LLM call to generate summaries, extract key events and character triplets, track narrative hooks, and build local vector indexes. Multi-layer memory, including a Story Bible for characters/settings, chapter summaries, vector search, and a plot tracker, feeds into continuous monitoring of chapter tension and authorial style drift, enabling directed rewriting.

Quick Start & Requirements

For Windows users, an executable launcher (tools/aitext.bat) automates environment setup, dependency installation (using domestic mirror sources), and backend startup, requiring only the placement of python-3.11.9-embed-amd64.zip in the tools/ directory. Developers require Python 3.9+ and Node.js 18+. Manual setup involves creating a Python virtual environment, installing backend dependencies (pip install -r requirements.txt), configuring LLM credentials in .env, and running the FastAPI backend (uvicorn). The frontend requires npm install and npm run dev. Supported AI models include OpenAI compatible, Anthropic Claude, and Volcano Ark Doubao. Local FAISS is used for vector storage, and SQLite serves as the primary database.

Highlighted Details

• Full-hosted Autopilot: Continuously generates content to target word count without manual chapter-by-chapter triggering, featuring fault tolerance and manual review nodes.
• Unified Post-Chapter Pipeline: Automates summarization, event extraction, triplet generation, hook tracking, and local vector indexing in a single LLM pass.
• Multi-layer Memory & Monitoring: Integrates structured world-building (Story Bible), semantic search, plot tracking, and monitors chapter tension and style drift with directed rewriting capabilities.
• Workbench & Prompting: Centralized configuration for 20+ prompt nodes allows strategy customization, with a workbench integrating writing, preview, status, knowledge graph, and LLM controls.

Maintenance & Community

The project actively engages with its community through live streams on Douyin (search 91472902104), typically nightly around 21:00 UTC, covering live writing, bug fixing, PR reviews, and deployment Q&A. Contributions are welcomed via forking, feature branching, and submitting Pull Requests following Conventional Commits guidelines.

Licensing & Compatibility

PlotPilot is licensed under the Apache License 2.0 with an added Commons Clause restriction. This permits learning, modification, and non-commercial internal deployment. However, it strictly prohibits any commercial use, including offering it as a paid SaaS, selling the source code, or integrating it as a value-add service in commercial products.

Limitations & Caveats

The primary limitation is the strict non-commercial use clause imposed by the Commons Clause, preventing its adoption in revenue-generating products or services. While a user-friendly launcher exists for Windows, developer setup requires specific Python and Node.js versions and manual configuration of LLM API keys.