💡 摘要
澳大利亚税务优化器是一个AI平台,分析Xero数据以恢复未获得的税收利益。
🎯 适合人群
🤖 AI 吐槽: “看起来很能打,但别让配置把人劝退。”
风险:Medium。建议检查:是否执行 shell/命令行指令;是否发起外网请求(SSRF/数据外发);API Key/Token 的获取、存储与泄露风险;依赖锁定与供应链风险。以最小权限运行,并在生产环境启用前审计代码与依赖。
🇦🇺 Australian Tax Optimizer
AI-Powered Tax Recovery & Compliance Platform
Recover $200K-$500K in missed tax benefits per client through forensic Xero analysis
Getting Started • Documentation • API Reference • Architecture
🎯 What is ATO?
The Australian Tax Optimizer is an AI-powered platform that connects to your Xero accounting data (read-only) and performs forensic analysis to identify:
| Recovery Area | Potential Savings | Legislation | |--------------|-------------------|-------------| | 🔬 R&D Tax Incentive | Up to 43.5% offset | Division 355 ITAA 1997 | | 📊 Unclaimed Deductions | Variable | Section 8-1 ITAA 1997 | | 💰 Carry-Forward Losses | Tax offset on profits | Subdivision 36-A | | 🏦 Division 7A Compliance | Avoid deemed dividends | Division 7A ITAA 1936 | | 🎁 FBT Optimisation | Reduce FBT liability | FBTAA 1986 |
🏗 Architecture
flowchart TB subgraph Client["🖥️ Client Layer"] UI[Next.js Dashboard] Reports[PDF/Excel Reports] end subgraph API["⚡ API Layer"] Auth[OAuth 2.0 Auth] Audit[Audit Endpoints] Xero[Xero Proxy] end subgraph Core["🧠 Core Engine"] AI[Gemini AI Analyzer] RND[R&D Engine] DED[Deduction Engine] LOSS[Loss Engine] DIV7A[Division 7A Engine] end subgraph Agents["🤖 Agent Fleet"] A1[Tax Law Analyst] A2[Xero Auditor] A3[R&D Specialist] A4[Deduction Optimizer] end subgraph Data["💾 Data Layer"] SB[(Supabase)] XeroAPI[Xero API] Cache[Transaction Cache] end subgraph Validation["✅ Validation"] V1[Tax Calc Validator] V2[R&D Eligibility] V3[FY Validator] end UI --> Auth UI --> Audit Auth --> XeroAPI Audit --> AI AI --> RND & DED & LOSS & DIV7A RND & DED & LOSS & DIV7A --> Agents Agents --> SB XeroAPI --> Cache --> SB AI --> V1 & V2 & V3 Reports --> SB style AI fill:#4285F4,color:#fff style SB fill:#3FCF8E,color:#fff style XeroAPI fill:#13B5EA,color:#fff
✨ Features
🔍 Forensic Analysis
- AI-powered transaction scanning
- Pattern recognition for R&D activities
- Misclassification detection
- Multi-year analysis (FY2020-25)
🛡️ Enterprise Security
- Read-only Xero access
- AES-256-GCM token encryption
- Multi-tenant isolation
- Row-level security (RLS)
🤖 16 Specialised Agents
- Tax Law Analyst
- R&D Tax Specialist
- Deduction Optimizer
- Loss Recovery Agent
- Division 7A Compliance
- And 11 more...
📊 Validation System
- 10 specialised validators
- 90-95% confidence scores
- Automatic fix instructions
- Compliance verification
🤖 Multi-Agent Development Framework
This project employs a sophisticated multi-agent development framework that coordinates 4 specialist agents plus an orchestrator for structured, quality-gated development.
Agent Fleet
Framework Agents (Development Process):
- 🎯 Orchestrator - Task decomposition, work distribution, results synthesis, quality control
- 🏗️ Specialist A: Architect - System design, API schemas, database ERDs, architecture decision records (ADRs)
- 💻 Specialist B: Developer - Production code implementation, feature development, refactoring
- 🧪 Specialist C: Tester - Unit/integration/E2E testing, QA, coverage reports (≥80% target)
- 📚 Specialist D: Reviewer - Code review, technical documentation, API docs, user guides
Tax Domain Agents (Business Logic):
- 18 specialized agents for Australian tax analysis (R&D, deductions, losses, Division 7A, etc.)
- See .agent/AGENTS.md for complete agent registry
Development Workflow
graph TD A[👨💼 Developer] -->|Requirement| B[📊 Senior PM] B -->|Create Linear Issue| C[🎯 Orchestrator] C -->|Decompose & Assign| D[🏗️ Specialist A: Design] C -->|Decompose & Assign| E[💻 Specialist B: Implement] C -->|Decompose & Assign| F[🧪 Specialist C: Test] C -->|Decompose & Assign| G[📚 Specialist D: Document] D -->|Handoff Context| E E -->|Handoff Context| F F -->|Handoff Context| G G -->|Integrate Results| C C -->|Review & Report| B B -->|Approve & Deploy| A style A fill:#FFB74D,color:#000 style B fill:#4FC3F7,color:#000 style C fill:#BA68C8,color:#fff style D fill:#81C784,color:#000 style E fill:#64B5F6,color:#000 style F fill:#FFD54F,color:#000 style G fill:#A1887F,color:#fff
Quality Gates
Every phase transition requires passing automated quality gates:
| Gate | Criteria | Blocks Phase | |------|----------|--------------| | ✅ Design Complete | ADR created, diagrams complete, risks documented | Implementation | | ✅ Implementation Complete | Code compiles, linting passes, basic functionality works | Testing | | ✅ Testing Complete | ≥80% coverage, all tests pass, no critical bugs | Documentation | | ✅ Documentation Complete | API docs, changelog updated, examples provided | Integration | | ✅ Integration Complete | All outputs merged, system functional, no conflicts | Final Approval | | ✅ Final Approval | PM + Developer sign-off, ready for deployment | Deployment |
See lib/agents/quality-gates.ts for automated enforcement.
Linear Integration
All development work is tracked in Linear with automated updates:
- 📝 Task Creation - Parent issues created from Developer requests
- 🔄 Sub-Task Assignment - Specialists automatically assigned via Linear sub-issues
- 📊 Real-Time Updates - Status transitions (pending → in-progress → review → done)
- 🚫 Blocker Escalation - Automatic escalation with @mentions in comments
- 📈 Progress Tracking - Daily reports showing specialist workload and velocity
Quick Commands:
npm run agent:orchestrator -- --task "Add feature" --priority High npm run agent:daily-report # View daily status npm run agent:quality-gate -- --gate <name> # Check quality gate npm run linear:sync # Sync with Linear npm run linear:report # Generate comprehensive report
See MULTI_AGENT_ARCHITECTURE.md for complete framework specification.
🚀 Quick Start
- Node.js 20.x or higher
- npm 10.x or higher
- Xero Developer Account
- Supabase Project
- Google AI API Key (for Gemini)
# Clone the repository git clone https://github.com/CleanExpo/ATO.git cd ATO # Install dependencies npm install # Copy environment template cp .env.example .env.local # Configure your environment variables (see below)
# Supabase (Required) NEXT_PUBLIC_SUPABASE_URL=https://your-project.supabase.co NEXT_PUBLIC_SUPABASE_ANON_KEY=your-anon-key SUPABASE_SERVICE_ROLE_KEY=your-service-role-key # Xero OAuth (Required) XERO_CLIENT_ID=your-xero-client-id XERO_CLIENT_SECRET=your-xero-client-secret XERO_REDIRECT_URI=https://your-domain.com/api/auth/xero/callback # Google AI (Required for analysis) GOOGLE_AI_API_KEY=your-gemini-api-key # Security (Required for production) TOKEN_ENCRYPTION_KEY=your-32-byte-hex-key # Generate: openssl rand -hex 32
# Run migrations via Supabase Dashboard SQL Editor # Or use the migration script: npm run db:migrate
Migration files are located in supabase/migrations/.
# Development npm run dev # Production build npm run build npm start # Run tests npm test
📚 Documentation
| Document | Description | |----------|-------------| | CLAUDE.md | AI development guidelines (RTCC pattern) | | MULTI_AGENT_ARCHITECTURE.md | Multi-agent framework specification | | spec.md | Technical specification | | API_DOCUMENTATION.md | Complete API reference | | FORENSIC_AUDIT_GUIDE.md | User guide for audits | | AGENTS_README.md | Agent system documentation | | DATABASE_MIGRATIONS.md | Migration instructions |
🧪 Testing
# Run all tests npm test # Run with coverage npm run test:coverage # Run specific test file
优点
- 基于AI的分析,准确恢复税收
- 全面覆盖各种税收激励
- 为企业主提供用户友好的界面
- 对敏感数据采取强有力的安全措施
缺点
- 需要与Xero集成,限制用户基础
- 功能依赖外部API
- 新用户的复杂设置过程
- 对小企业可能成本较高
相关技能
免责声明:本内容来源于 GitHub 开源项目,仅供展示和评分分析使用。
版权归原作者所有 CleanExpo.
