Build with NANDA

This tutorial outlines the steps to build, test, and deploy a minimal autonomous agent using the Python nanda-core library.

---

1. Environment Setup

Install the primary development package:

pip install nanda-core

---

2. Basic Agent Implementation

Create a Python file named incident_agent.py:

import asyncio
from nanda import Agent, SwarmMesh, ActionConstraint

class IncidentTriageAgent(Agent):
    name = "IncidentTriage"
    role = "MunicipalRouter"
    constraints = [
        ActionConstraint.VERIFIABLE_LOGGING,
        ActionConstraint.PRIVACY_REDACTION
    ]

    async def on_message(self, event, context):
        # Triage and categorize incoming civic report
        classification = await self.reason(
            prompt=f"Categorize severity and municipal department for: {event.payload}"
        )

        # Broadcast decision across the Swarm Mesh
        await self.broadcast(
            channel="municipal.routing",
            data={
                "incident_id": event.id,
                "classification": classification,
                "status": "ROUTED"
            }
        )

async def main():
    mesh = SwarmMesh(endpoint="grpc://127.0.0.1:50051")
    agent = IncidentTriageAgent()

    async with mesh.connect():
        await mesh.register(agent)
        print("Agent successfully registered to Swarm Mesh. Awaiting events...")
        await agent.run_forever()

if __name__ == "__main__":
    asyncio.run(main())

---

3. Local Test Execution

To simulate events locally before production rollout:

# Launch the NEST test harness
nanda-nest start --preset=civic-testbed

# In a separate terminal, start your agent
python incident_agent.py

---

4. Key Agent APIs

Memory Graph Access

await agent.memory.store(key="incident:101", value={"sector": 4, "severity": "HIGH"})
records = await agent.memory.query("incidents in sector 4")

Cryptographic Action Attestation

proof = await agent.generate_proof(action="DISPATCH_MEDICAL")
print(f"Attestation Hash: {proof.hash}")