What it is

AI 写作助手 (AI Writing Assistant) is an open-source, Next.js-based writing tool that helps people organize style prompts and generate content through large language models. It is written in TypeScript under the MIT license and lives in the Node.js and Next.js ecosystem, distributed as source that runs on a local or self-hosted server. The project has been public for roughly two years, holds 644 stars and 112 forks, and carries only 3 open issues at the time of writing. A hosted instance is offered at writing.chixitown.com.

The concrete problem it solves is the gap between raw chat interfaces and controlled long-form drafting. Instead of asking a model for text in one vague prompt, the tool separates the work into two stages: a writing assistant that builds an article from a topic, keywords, and a target word count, and a text optimizer that rewrites existing drafts to read less like machine output. The style editor exposes language, structure, narrative viewpoint, and emotional expression as adjustable dimensions, so a user can pin down voice before generating rather than fixing it afterward. The optimizer targets the statistical signals that detectors such as GPTZero are reported to measure, specifically perplexity and burstiness, and ships presets including a human-writing-characteristics mode and an AI-rewrite-guidance mode.

Key capabilities

• Generates articles from a topic, a set of keywords, and a requested word count.
• Multi-dimensional style editor for language, structure, narrative perspective, and emotional expression, plus signature phrases and cultural references.
• Built-in editor for revising generated text in place, with one-click Markdown export.
• AI text optimizer that removes machine-writing traits and targets perplexity and burstiness statistics.
• Optimization presets for human writing characteristics and for AI rewrite guidance, plus custom rewrite instructions.
• Multi-LLM support covering OpenAI GPT-4 and GPT-3.5, xAI Grok-2, Ollama local models, DeepSeek, and custom OpenAI-compatible endpoints.
• Backend tuned for long requests, with timeouts of up to ten minutes for large text processing, plus responsive desktop and mobile UI and detailed error and debugging output.

Who uses it and how

• Writers and content creators drafting articles who want a fixed personal voice applied across many pieces.
• Students and professionals producing copy under a word-count target.
• Teams routing the same prompts through whichever provider is cheapest or fastest, by switching the API settings panel.
• Users running Ollama locally to keep text on their own hardware.
• Editors pasting existing drafts into the optimizer to reduce detector-flagged statistical patterns before publishing.

Getting started

Install Node.js 16.20.0 or higher,

npm install

npm run dev