Patent Drafting Assistant

专利撰写助手 — AI-powered patent disclosure drafting

Features • Quick Start • Usage • Structure • Customize • Contributing

Input your core technical concept and let AI generate a complete 10-chapter Chinese patent disclosure. Supports streaming generation, per-chapter editing, auto figure prompts, version history, and DOCX/Markdown export — all through a clean local web UI.

Features

• AI-driven drafting — Generate a complete patent disclosure from a single technical concept
• 10-chapter standard structure — Follows Chinese patent disclosure conventions
• Concept-first workflow — Analyze your concept first, then generate chapters based on structured extraction
• Technical feature analysis — AI extracts key elements, innovation points, and expected effects
• Streaming generation — Watch each chapter generate in real time via SSE
• Per-chapter editing — Edit any chapter manually, switch between Markdown edit and preview modes
• Auto figure prompts — AI inserts [Figure N: description] markers; right panel shows the figure checklist
• Version history — Every save creates a snapshot; rollback anytime
• DOCX export — Professional Word documents with correct Chinese fonts (Song Ti / Hei Ti)
• Markdown export — Plain text for other editing tools
• Zero-config LLM — Auto-reads API keys from Claude Code settings or environment variables

Quick Start

Prerequisites

• Python 3.10+
• An Anthropic-compatible LLM API (e.g. DeepSeek, OpenAI, Claude)

Install

git clone https://github.com/Dyp130/Patent-assistant.git
cd Patent-assistant
pip install -r requirements.txt

Configure

Copy the example env file and set your credentials:

cp .env.example .env
# Edit .env with your API key, base URL, and model

Or export as environment variables:

export ANTHROPIC_BASE_URL="https://api.deepseek.com"
export ANTHROPIC_AUTH_TOKEN="your-api-key"
export ANTHROPIC_MODEL="deepseek-chat"

If you have Claude Code installed, the app automatically reads API config from ~/.claude/settings.json.

Launch

python run.py

Open http://127.0.0.1:8000

Usage

Workflow

Screenshots

Screenshots coming soon — run python run.py to see it in action.

Chapter Structure

Customizing

Edit config/prompts.yaml to adjust AI generation behavior for each chapter — no code changes needed. Variables use {placeholder} syntax.

chapter_5_technical_solution:
  system: |
    你是一位资深的中国专利代理人。
    请根据以下核心技术构思，撰写"技术方案"章节。
    ...
  user: |
    发明名称：{title}
    核心技术构思：{technical_concept}
    ...

Project Structure

Patent-assistant/
├── config/
│   ├── settings.py           # Configuration & env reading
│   ├── prompts.yaml          # AI prompt templates
│   └── chapter_schema.yaml   # Chapter definitions
├── src/
│   ├── main.py               # FastAPI application entry
│   ├── db/                   # Database setup (SQLAlchemy + SQLite)
│   ├── routes/               # Project, chapter, generation, export APIs
│   ├── models/               # ORM models and Pydantic schemas
│   ├── services/             # AI generation, analysis, export
│   ├── templates/            # Jinja2 HTML templates
│   └── static/               # CSS & vanilla JavaScript
├── .env.example              # API configuration template
├── requirements.txt
├── run.py                    # Launch script
└── README.md

Tech Stack

Contributing

Contributions are welcome — see CONTRIBUTING.md.

1. Fork the repo
2. Create a branch (git checkout -b feat/your-feature)
3. Commit your changes
4. Push and open a Pull Request

License

MIT — see LICENSE.