on_event
React when something happens on the bus.
Client
await messenger.publish("order.created", payload)Server
class OrderService:
@on_event("order.created")
async def on_order(self, payload):
...
Build services, watch them live, and get alerted when something fails — in one small Python stack.
Kontiki is a compact open-source stack for teams that need the essentials — building, running and monitoring Python services — without the heavy ops stack.
Kontiki
Write the methods. Run the service.
A Python framework on AMQP — you write entrypoints, Kontiki runs the bus.
- Write only your business logic — Kontiki manages RabbitMQ connections, message routing (RPC, events, sessions, broadcast), service lifecycle, and configuration merge.
- Design interactions as messages — combine RPC, events, broadcast, and per-instance sessions to describe how services collaborate.
- Configuration-driven — merged YAML and a unified runner so services start the same way in development and production.
- Scale out — run as many service instances as you need by starting more processes.
@rpc
Call another service like a local method.
Client
await StockProxy(messenger).get_stock(sku)Server
class StockService:
@rpc
async def get_stock(self, sku):
return self.delegate.stock_for(sku)@http
Expose a simple HTTP API.
class HealthService:
@http("/health", "GET")
async def health(self, request):
return {"ok": True}@task
Run a periodic task.
class PollService:
@task(interval=60)
async def poll(self):
...
kontiki-tui
Services tab
Events tab
Watch your services live
A Textual TUI over what the Kontiki registry already tracks — plus local logs.
- Services — status, pid/host, version, and local stats.
- Events & exceptions — inspect what the registry recorded, with local filters.
- Logs — read configured log files without leaving the terminal.
kontiki-monitor
Alert when a service goes degraded
Get alerted when something fails
A small ops suite for Kontiki platforms — complete enough to run, simple enough to own.
- One complete ops loop — build, observe, get alerted — no patchwork of tools to glue together.
- Alerts that matter from day one — fleet gaps, bad service state, full disks: signals you can act on immediately.
- Platform-native, low overhead — built on the Registry and the bus you already run, without a separate observability stack.
Get started
Install from PyPI
pip install kontiki kontiki-tui kontiki-monitor