Query Status from Plugins#

Other conda plugins can use Broker to make runtime decisions. Query methods do not start the broker.

from conda_broker import Broker

service = Broker.current().service("package-cache")
check = service.check()

if check.ready:
    query_local_metadata()
else:
    query_repodata_directly()

status = check.status

Service.check() returns a compact report for plugin CLIs and JSON output. When the broker is running, it distinguishes a known stopped service from an unknown service:

check = Broker.current().service("package-cache").check()

if check.available:
    print(check.to_dict())
else:
    use_inline_fallback(reason=check.reason)

When the broker is stopped, check() returns reason="broker-unavailable" without loading provider entry points. That keeps hot-path conda hooks cheap even when many providers are installed.

Service.status() returns None when the service or broker is unavailable, so optional integrations can safely fall back. Use Broker.current().status("service-name") when you explicitly want provider discovery and a strict query that errors for unknown services.

For services that expose a local API, check readiness and read the endpoint:

from conda_broker import Broker

service = Broker.current().service("package-cache")

if endpoint := service.endpoint(ready=True):
    query_local_metadata(endpoint.url)
else:
    query_repodata_directly()

Service.endpoint(ready=True) never starts the broker. It returns None unless the service is already ready and the endpoint has a resolved URL.

Only explicit startup calls can start the broker:

from conda_broker import Broker

broker = Broker.current()
broker.start()
broker.service("package-cache").start()
broker.service("package-cache").wait(start=True)

For scripts that need a temporary service and want automatic cleanup, use a context manager:

from conda_broker import Broker

with Broker.current().service("package-cache").started(wait=True) as service:
    if endpoint := service.endpoint(ready=True):
        query_local_metadata(endpoint.url)

The context manager stops the service on exit only when it started the service on entry. If it had to start the broker too, it stops that broker on exit as well. With wait=True, failure to reach readiness raises ServiceNotReadyError after cleanup.

Use startup calls in user-visible commands, not in hooks that run for every conda invocation.

Show Source