OpenSpec

Enterprise-Grade AI Platform for Professional Long-Document Generation

Turn your knowledge base into standards-compliant, publication-ready documents in minutes — not days.

English | 中文

Live Demo • Video Introduction • Quick Start • Contact Us

---

Why OpenSpec?

Generic AI (ChatGPT, Claude, Qwen, etc.) falls short on professional long documents:

Problem	Generic AI	OpenSpec
Documents over 50 pages	Context lost, inconsistencies	Chapter-by-chapter generation, fully coherent
Industry standards & regulations	Hallucination, makes things up	RAG retrieval from your own knowledge base
Document formatting & templates	Cannot control layout	PDF / Markdown / AutoCAD export
Quality assurance	Single-pass, no review	Three-agent workflow (Researcher + Generator + Auditor)

OpenSpec is not another AI writing tool. It is a document engineering platform built for professionals who need accuracy, compliance, and scale.

How It Works

                          ┌───────────────────────────────────────┐
                          │          Knowledge Base (RAG)         │
                          │   (Standards, Cases, Project Docs)    │
                          └──┬──────────────┬──────────────┬─────┘
                             │              │              │
                          query on       query on       query on
                          demand         demand         demand
                             │              │              │
┌──────────┐     ┌──────────▼──┐     ┌─────▼──────┐     ┌▼───────────┐     ┌──────────┐
│  User    │────▶│  Researcher │────▶│  Generator │────▶│  Auditor   │────▶│  Human-AI│
│  Input   │     │  (Research  │     │  (Write    │     │  (Review & │     │  Editing │
│          │     │   & Gather) │     │   Content) │     │   Verify)  │     │          │
└──────────┘     └─────────────┘     └────────────┘     └──────┬─────┘     └────┬─────┘
                                           ▲                   │                │
                                           └───── revise ──────┘                ▼
                                                                        ┌──────────────┐
                                                                        │  One-Click    │
                                                                        │  Export       │
                                                                        │  (PDF/MD/CAD) │
                                                                        └──────────────┘

The system is powered by a three-agent workflow, where each agent can autonomously query the knowledge base when it needs more context:

1. Researcher Agent — Gathers relevant standards, historical cases, and reference materials from the knowledge base to build a solid research foundation
2. Generator Agent — Produces standards-compliant professional content chapter by chapter, based on the research context
3. Auditor Agent — Reviews the generated content for compliance, consistency, and accuracy; queries the knowledge base for cross-validation and sends revisions back to the Generator if needed
4. Human-AI Collaboration — Review, rewrite, and refine at the chapter level
5. One-Click Export — PDF, Markdown, AutoCAD title blocks, and more

Use Cases

Domain	Typical Documents
Architecture Design	Construction drawing design notes, feasibility study reports
Automotive Repair	Repair technical manuals, fault diagnosis reports
Healthcare	Clinical trial reports, diagnosis & treatment standard documents
Bidding & Tendering	Technical proposals, tender document preparation
Any Industry	Any structured long document generated from a knowledge base

UI Showcase

Video Demo (Architecture Design Scenario): Watch on Bilibili

Live Demo

Try OpenSpec online: https://archspec.aizzyun.com/

• Email: test@qq.com
• Password: test123456

Tech Stack

Layer	Technology
Frontend	Vue 3, TypeScript, Vite
Backend	Spring Boot 3, Java 17
AI Agent	Python, LangGraph, LangChain
Knowledge Retrieval	RAGFlow
Observability	Langfuse (LLM tracing & cost analytics)
Database	PostgreSQL
Deployment	Docker, Docker Compose

Quick Start

Requirements

Component	Version
Docker	>= 20.10
Docker Compose	>= 2.0

One-Click Deployment

# 1. Clone the repository
git clone https://github.com/zhuzhaoyun/OpenSpec.git
cd OpenSpec

# 2. Copy and edit environment variables
cp deploy/docker/.env.example deploy/docker/.env
# Edit .env and fill in required configurations (RAGFlow, LLM API Key, etc.)

# 3. Start all services
cd deploy/docker
docker compose up -d

Once started, visit http://localhost to use the platform.

Environment Variables

Variable	Description	Required
RAGFLOW_API_KEY	RAGFlow API key	Yes
RAGFLOW_BASE_URL	RAGFlow service URL	Yes
DASHSCOPE_API_KEY	LLM API key (default: Qwen)	Yes
LANGFUSE_SECRET_KEY	Langfuse secret key (Prompt management)	No
LANGFUSE_PUBLIC_KEY	Langfuse public key	No
LANGFUSE_BASE_URL	Langfuse service URL	No

For the full list of environment variables, see deploy/docker/.env.example.

Local Development

Expand to view local development guide

Frontend

cd apps/web
npm install
npm run dev
# Visit http://localhost:5173

AI Agent

cd apps/agent
pip install -r requirements.txt
cp ../../deploy/docker/.env.example .env
# Edit .env and fill in required configurations
uvicorn app:app --reload --port 5000 --host 0.0.0.0

Backend

cd apps/backend
mvn spring-boot:run

Open Source vs Enterprise

Feature	Open Source	Enterprise
Intelligent Retrieval & Chapter Generation	✅	✅
Human-AI Collaborative Editing	✅	✅
PDF / Markdown Export	✅	✅
AutoCAD Plugin (File Import)	❌	✅
Training & Q&A Support	❌	✅
Data Processing Consulting	❌	✅

Contributing

We welcome contributions! Feel free to:

• Star this project to show your support
• Submit Issues for bug reports and feature requests
• Open Pull Requests for improvements

Star History

Contact Us

For enterprise edition details or business cooperation, feel free to reach out:

• Email: dlutyaol@qq.com

• WeChat: Scan the QR code below

  

License

Copyright (c) 2024-2026 OpenSpec Contributors.

Licensed under The GNU General Public License v3.0 (GPLv3).