Deng-m1/MaliangAINovalWriter

马良写作 AINoval — AI novel writer for Chinese webnovels（网文）| 多智能体AI小说创作平台：三级大纲、知识图谱一致性、30万字连载

📚 Maliang AI Novel Writer (AINoval)

What it is – An open‑source, web‑based platform that helps Chinese web‑novel authors (and the operators of such platforms) write long‑form fiction with the assistance of large language models. It bundles a rich‑text editor, AI‑driven outline/scene generation, world‑building tools, and a full‑featured admin console for managing users, models, costs and content moderation.

🎯 Core capabilities

Feature	What you get
AI‑powered writing assistant	One‑click continuation, expansion, polishing, translation, character and plot generation. Supports OpenAI, Anthropic Claude, Google Gemini and any OpenRouter‑compatible model via private API keys.
Hierarchical content model	Works with a four‑level hierarchy – Work → Volume → Chapter → Scene – making it easy to organise novels that run into hundreds of thousands of words.
World‑building engine	Structured “setting tree” (characters, locations, items, factions, etc.) that can be generated or edited by AI, versioned with snapshots and visualised as a relationship graph.
Prompt & preset management	Users store system‑level and personal prompts, bundle them with model parameters into reusable presets that appear in the editor toolbar.
LLM observability & cost tracking	Every model call is logged (prompt, response, latency, token usage). Admins can query logs, see per‑user/model statistics and audit expenses.
Admin dashboard	Separate Flutter‑Web admin UI for user/RBAC management, public model pool configuration, pricing tables, subscription/credit handling, content‑review workflows and knowledge‑extraction tasks.
AI‑driven plot‑outline generator (Next Outline)	Generates multiple possible continuations for the current scene, lets you re‑generate individual options, and apply the chosen outline directly to the manuscript.
Knowledge‑extraction / “book‑splitting”	Uses AI to analyse existing novels (via upload or a third‑party “Tomato Novel” service) and extracts style, plot, character, world‑building and other meta‑knowledge into a personal knowledge base.
Analytics & dashboards	Real‑time word‑count, writing streaks, token‑consumption trends and feature‑usage charts help authors monitor productivity and budget.

🛠️ Technology stack

Layer	Tech
Frontend (author UI)	Flutter Web, Flutter Quill rich‑text editor, flutter_bloc / Provider for state, fl_chart for graphs, Dio for HTTP, Hive & shared_preferences for local cache
Frontend (admin UI)	Separate Flutter Web app (`lib/admin_main.dart`) with the same stack, plus custom widgets for tables, charts and dialogs
Backend	Spring Boot 3 (WebFlux, reactive), Java 21, Spring Security + JWT, RabbitMQ for async jobs
AI orchestration	LangChain4j (Java wrapper for LLM chains), supports OpenAI, Anthropic, Gemini, OpenRouter and custom endpoints
Databases	MongoDB (reactive) for core data, Chroma vector store for embedding‑based knowledge extraction
Storage / services	Alibaba Cloud OSS & SMS (configurable), Docker‑compose deployment, optional vector‑store, RabbitMQ
Containerisation	Docker image bundles the compiled Spring‑Boot JAR and the pre‑built Flutter web assets; `docker‑compose.yml` also spins up MongoDB.

🚀 Getting started (Docker one‑click)

Clone the repo and cd to the project root.
Copy the example env file:
```
cp deploy/open/production.env.example deploy/open/production.env
```
Edit the file – set your MongoDB URI, JWT secret, and any LLM API keys you want to use.
Place the server JAR (download from the releases page) into deploy/dist/ainoval-server.jar.

Build and run:

docker compose -f deploy/open/docker-compose.yml build
docker compose -f deploy/open/docker-compose.yml up -d

Create the admin account (run the init script once the containers are up):
```
# Windows
.\deploy\init-admin.bat
# Linux/macOS
chmod +x deploy/init-admin.sh && ./deploy/init-admin.sh
```
Default credentials – admin / 123456 – should be changed immediately.
Open http://localhost:18080/ for the author UI or http://localhost:18080/admin for the admin console.

👥 Who might use it?

Audience	Why it helps
Solo novelists	AI‑assisted drafting, polishing and outline generation speeds up writing and keeps long‑term story consistency.
Writing teams / studios	Shared setting tree and knowledge base, plus RBAC, let multiple writers collaborate on a single serial.
Platform operators	The admin panel gives full control over model pools, cost tracking, subscription/credit systems and content moderation.
Researchers / hobbyists	Open‑source code, Docker deployment and a Java‑centric AI stack (LangChain4j) make it a good sandbox for experimenting with LLM‑driven creative tools.

🤝 Contributing

The project follows the usual fork‑branch‑PR workflow. Issues are tracked on GitHub, and a QQ group (ID 1062403092) is listed for community chat. The code is licensed under Apache 2.0.

📌 Bottom line

Maliang AINoval is a full‑stack, production‑ready AI‑novel writing platform that combines a modern Flutter‑Web editor with a Spring‑Boot backend, multi‑model LLM orchestration, structured world‑building, and extensive admin/observability tools. It is a genuine software project squarely in the AI‑assisted content‑creation space.