TAXIA

Korean Tax AI Library with Graph-RAG

RAG (Retrieval-Augmented Generation) + Graph-RAG library for Korean tax consultation and question-answering.

All answers are provided with clear legal citations and complete audit trail support.

Current Status: v1.0.0 Official Release ✅

All development phases completed (Phase 0-7)

• ✅ Qdrant vector search integration
• ✅ Claude/OpenAI LLM support
• ✅ Neo4j Graph-RAG support
• ✅ CLI tools (taxia index, taxia ask, taxia health)
• ✅ FastAPI REST API server (7 endpoints)
• ✅ Korean tax law data loader
• ✅ 40+ tests (100% pass rate)
• ✅ Comprehensive documentation (7 markdown + official documentation site)
• ✅ Public GitHub repository
• ✅ PyPI deployment complete
• ✅ GitHub Pages documentation site deployment
• ✅ Complete audit trail support

What is TAXIA?

TAXIA is a production-ready AI library designed specifically for Korean tax law consultation. It combines:

• Vector Search (Qdrant): Fast semantic search across tax law documents
• Graph-RAG (Neo4j): Relationship-aware retrieval using law article connections
• LLM (Claude/GPT): Natural language generation with mandatory legal citations
• Audit Trail: Complete tracking of every query and response

Key Features

1. Mandatory Citations

Every answer includes at least 2 legal citations: - Law articles - Enforcement decrees - Enforcement rules - Official interpretations

2. Complete Audit Trail

• Unique trace ID per query
• Full context logging
• Citation provenance tracking
• Timestamp recording

3. Production Ready

• Comprehensive test suite (40+ tests)
• Type-safe Python with py.typed
• FastAPI REST API
• CLI tools
• Docker support

4. Korean Tax Law Coverage

• Income Tax Act (2015-2025)
• Corporate Tax Act (2015-2025)
• Value Added Tax Act (2015-2025)
• 242 files total across 11 years

Installation

pip install taxia-core

Quick Example

from taxia import TaxiaEngine

# Initialize engine
engine = TaxiaEngine()

# Ask a question
result = engine.answer("What is the corporate tax filing deadline?")

# Access results
print(result.answer)
# "The corporate tax filing deadline is within 3 months from the end of the fiscal year."

print(result.citations)
# ["Corporate Tax Act Article 60", "Enforcement Decree Article 132"]

print(result.trace_id)
# "trace-2025-01-23-abc123"

Architecture

┌─────────────────────────────────────┐
│         User Interface              │
│  CLI / REST API / Python SDK        │
└─────────────┬───────────────────────┘
              │
       ┌──────▼──────┐
       │ TaxiaEngine │
       └──────┬──────┘
              │
    ┌─────────┼─────────┐
    │         │         │
┌───▼───┐ ┌──▼───┐ ┌──▼────┐
│Qdrant │ │Neo4j │ │ LLM   │
│Vector │ │Graph │ │Claude │
└───────┘ └──────┘ └───────┘

Use Cases

Tax Consultation Automation

Automate responses to common tax questions with full legal backing.

Compliance Checking

Verify tax compliance requirements with citation trails.

Research Assistant

Quick lookup of tax law provisions across years.

Chatbot Backend

Power tax consultation chatbots with reliable answers.

Components

Core Engine

• TaxiaEngine: Main orchestration layer
• TaxiaConfig: Configuration management
• QueryResult: Type-safe results

Retrieval

• VectorStore: Qdrant integration
• GraphStore: Neo4j integration
• DocumentLoader: Tax law data loading

LLM Integration

• ClaudeConnector: Anthropic Claude
• OpenAIConnector: OpenAI GPT
• Prompt templates for tax law

API & CLI

• FastAPI REST server (7 endpoints)
• Rich CLI with health checks
• Type-safe client libraries

Data Distribution

Tax law data distributed via Hugging Face:

Repository: xaikorea/taxia-data

• 242 JSON files (2015-2025)
• Automatic download on first use
• Version controlled updates

Requirements

• Python 3.9+
• Qdrant (vector search)
• Neo4j (optional, for Graph-RAG)
• Anthropic API key or OpenAI API key

Development Status

Links

• GitHub: xaikorea/taxia
• PyPI: taxia-core
• Documentation: GitHub Pages
• Data: Hugging Face

License

Apache 2.0 - See LICENSE

Citation

@software{taxia2025,
  title={TAXIA: Korean Tax AI Library with Graph-RAG},
  author={XAI Korea},
  year={2025},
  url={https://github.com/xaikorea/taxia}
}