Documentation — NETWORQIX

Installation

Install the NETWORQIX SDK for your preferred language. A running bus instance is required — use our cloud sandbox or self-host with Docker.

# Python
pip install networqix

# TypeScript / Node.js
npm install @networqix/sdk

# Go
go get github.com/networqix/sdk-go

# Start local bus (Docker)
docker run -p 4222:4222 networqix/bus:latest
            

Quick Start

Connect two agents and send a message in under 30 lines of code.

from networqix import Agent, Bus

# Connect to the bus
bus = Bus.connect("nqx://localhost:4222")

# Create an agent
agent = Agent(
    bus=bus,
    name="researcher",
    capabilities=["web_search", "summarize"]
)

# Subscribe to tasks
@agent.on("tasks.research")
async def handle_research(msg):
    result = await search_and_summarize(msg.payload)
    await msg.reply(result)

# Publish a task from another agent
await bus.publish("tasks.research", {
    "query": "Latest advances in multi-agent RL"
})
            

Core Concepts

Bus: The central communication hub. All agents connect to a bus instance (or cluster). Messages are routed through the bus — agents never communicate directly.

Agent: A registered participant on the bus with a name, capabilities, and message handlers. Agents can be long-running services or ephemeral task workers.

Topic: A named channel for messages (e.g., tasks.research, state.shared). Supports wildcards and hierarchical namespacing.

Context Pipeline: A specialized transport for large payloads. Automatically chunks, compresses, and reassembles context data between agents.

Messaging Patterns

NETWORQIX supports multiple messaging patterns out of the box:

# Pub/Sub — broadcast to all subscribers
await bus.publish("events.agent.ready", payload)

# Request/Reply — send and wait for response
response = await bus.request("tasks.analyze", payload, timeout="5s")

# Fan-Out — parallel dispatch to multiple agents
results = await bus.fanout("tasks.review", payload, agents=["reviewer-1", "reviewer-2"])
            

State Synchronization

Share mutable state across agents with CRDT-backed synchronization. No locks, no conflicts — eventual consistency at wire speed.

# Create shared state
state = bus.state("project-alpha", schema=ProjectSchema)

# Read and update from any agent
state.tasks.append({"id": "task-42", "status": "pending"})
await state.commit()

# Subscribe to changes
@state.on_change
async def on_update(diff):
    print(f"State updated: {diff}")
            

Deployment

For production deployments, we recommend running a 3-node bus cluster with persistent storage. See our Architecture page for topology options.