Functions

A function processes messages from one or more Pulsar or Kafka topics and can write results to an output topic. The registry Functions API supports both messaging protocols; the function's connection selects the protocol. Use functions for lightweight filtering, transformation, enrichment, or routing that does not need an agent loop.

Get your registry endpoint

Function configuration fields

Function create and update are multipart requests. Supply one package source through an uploaded data part, a package url, or the matching field in functionConfig.

Choose a messaging protocol

Both protocols use /apis/cloud.sn.io/v1/functions, ork functions, and orca.cloud.functions. Set connection to a Pulsar or Kafka connection in your Workspace; there is no separate protocol field on the function. The connection supplies the broker endpoint and authentication configuration. To change a function's connection or protocol, create a new function.

Kafka functions require the generic runtime. Set genericKind in the JSON-encoded customRuntimeOptions string to match your package, such as python for a Python function. Your Workspace must provide the corresponding generic runner image. Optional kafkaConsumerConfig and kafkaProducerConfig objects carry Kafka client properties such as auto.offset.reset and acks. Use these objects for Kafka client tuning instead of the Pulsar producerConfig field.

If the Kafka connection defines spec.kafka.schemaRegistry, the function also receives its Schema Registry URL. External connections can supply Basic or OAuth2 credential references there; internal connections use Workspace-managed authentication.

Create a Pulsar function

This example deploys the built-in word_count function. $ORCA_REGISTRY_URL points at the host root; each client appends /apis/cloud.sn.io/v1.

ork functions create \
  --name word-count \
  --function-type word_count \
  --inputs persistent://public/default/sentences \
  --output persistent://public/default/word-counts \
  --connection pulsar-main \
  --parallelism 1

For a custom package, upload it through Packages and set a package URL, or pass a local file through the CLI's --jar, --py, or --go flag.

Create a Kafka function

Create a Kafka connection named events-kafka as shown in Connections, and create the Kafka topics sentences and uppercase-sentences on that cluster. Save this function as uppercase.py in your current directory. It returns each input message in uppercase.

class Uppercase:
    def process(self, input, context):
        return input.upper()

Each example uploads that file and selects the Python generic runtime. In the TypeScript example, reuse the orca client initialized above.

ork functions create \
  --name kafka-uppercase \
  --py ./uppercase.py \
  --classname uppercase.Uppercase \
  --inputs sentences \
  --output uppercase-sentences \
  --connection events-kafka \
  --parallelism 1 \
  --processing-guarantees ATLEAST_ONCE \
  --custom-runtime-options '{"genericKind":"python","kafkaConsumerConfig":{"auto.offset.reset":"earliest"},"kafkaProducerConfig":{"acks":"all"}}'

Check the deployment with ork functions status kafka-uppercase -o json. Publish a message such as hello to sentences using a Kafka producer, then consume uppercase-sentences and check for HELLO.

Operate functions

The CLI reference lists package, schema, windowing, resource, retry, ordering, state, and lifecycle flags.

Check status

ork functions status word-count -o json

Permissions

Treat every credential that can call the Workspace's cloud.sn.io routes as full access to its functions, and separate access with Workspaces. See Control registry access.

What's next