Skip to main content

Koveria SDK

Welcome to the Koveria SDK documentation! This guide will help you build compliant AI agents using Python.

What is the Koveria SDK?​

The Koveria SDK is a Python library that enables you to:

  • Build AI agents with a simple, declarative API
  • Ensure compliance with automatic risk assessment and guardrails
  • Track memory across conversations with hot and durable storage
  • Monitor observability with built-in logging, metrics, and tracing
  • Deploy easily to teams with integrated CI/CD workflows

Quick Example​

from koveria import Agent, agent_action

class CustomerSupportAgent(Agent):
    """Handle customer support inquiries."""

    @agent_action(action_type="process")
    async def handle_inquiry(self, message: str) -> str:
        """Process a customer inquiry and return a response."""
        response = await self.llm.complete(
            messages=[
                {"role": "system", "content": "You are a helpful support agent."},
                {"role": "user", "content": message}
            ],
            model="gpt-4o"
        )
        return response.content

# Usage
agent = CustomerSupportAgent()  # Loads config from environment
result = await agent.handle_inquiry("How do I reset my password?")
print(result)

Getting Started​

Choose your path:

🎓 Learning Path (Recommended for beginners)​

  1. Getting Started - Install and configure (5 min)
  2. First Agent Tutorial - Build your first agent (10 min)
  3. Core Concepts - Understand the architecture (15 min)

🛠️ Task-Oriented Path (For experienced developers)​

📚 Reference Path (Quick lookups)​

Key Features​

🔒 Compliance-First​

Every agent is automatically assessed for EU AI Act compliance:

  • Risk classification (prohibited, high-risk, limited-risk, minimal-risk)
  • Automatic guardrail assignment based on risk level
  • Evidence collection and audit trail
  • Agent prefilling reduces manual work by 70%

💾 Built-in Memory​

Agents can remember state across conversations:

  • Hot storage (Redis) for fast access
  • Durable storage (PostgreSQL) for long-term retention
  • Automatic memory management with TTLs
  • Team-level memory isolation

📊 Observability​

Monitor your agents with comprehensive observability:

  • Logs: Structured logging with 4 levels (INFO, DEBUG, AUDIT, TRACE)
  • Metrics: Prometheus metrics for cost, latency, errors
  • Traces: Distributed tracing with Jaeger/OpenTelemetry

🚀 Team-Based Deployment​

Deploy agents to teams and promote across environments:

  • Development → Quality → Production
  • Multi-team agent reuse (one agent, multiple teams)
  • Independent version control per team

Next Steps​

Ready to build your first agent? Start with our Getting Started Guide.

Have questions? Join our Discord community or check out GitHub Discussions.