Smart DIAN Agentic Invoicing System

Autonomous multi-agent system that audits fiscal data, resolves DIAN XML validation errors, and automates supplier invoice processing using Gemini.

Comment

Inspiration

Electronic invoicing with the Colombian tax authority (DIAN) is rigid and error-prone. Small and medium enterprises frequently face rejected XML documents, complex tax validations, and manual data entry bottlenecks. We built an autonomous multi-agent pipeline powered by Gemini to transform static invoicing into a seamless, intelligent workflow.

What it does

Smart DIAN Agentic Invoicing System handles backend tax operations autonomously:

  • Pre-submission Tax Auditor: An agent in second-plan inspects JSON payload fields (VAT, withholding rates, tax regime) against DIAN rules before submission.
  • DIAN XML Error Resolver: Interprets cryptic error codes returned by DIAN servers, providing natural language resolution steps and auto-fixing formatting issues.
  • Supplier Invoice Parser: Multi-modal extraction of incoming supplier PDFs/images, automatically registering line items and tax figures into the database.

How we built it

  • Backend: Node.js, TypeScript, and Docker containerization.
  • AI & Multi-Agent Architecture: Google Gemini API (@google/genai) and Google ADK concepts for background task execution.
  • Database & API: PostgreSQL/MySQL schema with secure multi-tenant isolation and DIAN electronic invoicing integration routines.

Challenges we faced

Mapping raw tax regulations and unstructured XML error logs into structured JSON schemas that Gemini could reliably parse without high latency during high-volume invoice processing.

What we learned

Implementing background agent harnesses significantly improves resilience in critical fintech workflows, drastically reducing manual troubleshooting time.

What's next

Adding automated tax reporting agents and deeper integration with ERP systems for real-time ledger synchronization.

Share this project:

Updates