ADRI Documentation

Agent Data Readiness Index – Stop AI agents from breaking on bad data

ADRI is an open-source data quality validation framework built for AI agents. Generate a standard from good data once, wrap your functions with @adri_protected, and block dirty payloads before they crash your agents.

ADRI in your stack

ADRI is a data quality gate for agent workflows. It complements related standards you may already use:

• ADRI: Validate inputs and enforce a quality contract before tools/actions run (fail-fast, warn, or continue).
• MCP: Agent-to-tool connectivity (standard way to connect agents to tools, APIs, and resources). See https://modelcontextprotocol.io/
• A2A: Agent-to-agent interoperability (standard messaging between agents across frameworks/vendors). See https://a2a-protocol.org/latest/

Use ADRI with or without MCP/A2A — the goal is to stop bad data from breaking agents right at the boundary.

Choose Your Path

🚀 Put ADRI to Work

Package consumer documentation – ship reliable agents fast

pip install adri
adri setup --guide
adri generate-standard examples/data/invoice_data.csv \
  --output examples/standards/invoice_data_ADRI_standard.yaml
adri assess examples/data/test_invoice_data.csv \
  --standard examples/standards/invoice_data_ADRI_standard.yaml

from adri import adri_protected

@adri_protected(standard="invoice_data_standard", data_param="invoice_rows")
def your_agent_function(invoice_rows):
    return agent_pipeline(invoice_rows)

📚 User Documentation:

• Getting Started – Installation, walkthrough, and first success
• FAQ – Answers for agent engineers, data teams, and compliance
• Framework Playbooks – LangChain, CrewAI, LlamaIndex, LangGraph, Semantic Kernel
• Adoption Journey – When to switch on Verodat-managed data supply
• API Reference – Complete decorator, CLI, and configuration details
• Why Open Source – Strategy and licensing

🛠️ Contribute to ADRI Community

Developer documentation – improve ADRI itself

🔧 Contributor Documentation:

• Development Workflow – Local testing and CI setup
• Framework Extension Pattern – Adding new framework support
• Code Style Guide – Contribution guidelines
• GitHub Repository – Source code and issues

Key Features

• 🛡️ One-Decorator Protection – Add @adri_protected to any function
• 🤖 Framework Agnostic – Works with LangChain, CrewAI, AutoGen, LlamaIndex, etc.
• 🚀 Smart Defaults – Zero-config start with optional tuning
• 📊 Five Dimensions – Validity, completeness, consistency, plausibility, freshness
• 📋 Flexible Modes – Fail-fast, warn, or continue for selective flows
• ⚡ Enterprise Ready – Local-first with a clear path to Verodat MCP

---

Ready to start? Hit the getting started guide, then follow the adoption journey when you need shared compliance logging and managed data supply.