🐳 Orca Python SDK

The Orca Python SDK enables developers to define and register Python-based algorithms into the Orca framework.

Orca exists to make it seamless to build scalable, production-grade ML or analytics pipelines on timeseries data.

🚀 Getting Started

Before using this SDK, you should install the Orca CLI and start Orca Core.

1. Install the Orca CLI Ensure that Docker is installed on your system.

Linux / macOS

curl -fsSL https://raw.githubusercontent.com/orc-analytics/orca/main/install-cli.sh | bash

Windows

Use WSL (Windows Subsystem for Linux) and run the above command inside your WSL shell.

Once installed, follow the instructions in the Orca documentation to start the Orca Core service.

2. Start the Orca Server

orca start

3. Print out the server details

orca status

4. Install the Orca sdk into your python project:

pip install orca-python

5. Start building out your algorithms

Write a file defining your algorithms and what windows trigger them:

# main.py
from orca_python import Processor

proc = Processor("ml")

@proc.algorithm("MyAlgo", "1.0.0", "MyWindow", "1.0.0")
def my_algorithm() -> dict:
return {"result": 42}

if __name__ == "__main__":
proc.Register()
proc.Start()

Then run your python file to register it with orca-core:

 ORCA_CORE=grpc://localhost:32770 PROCESSOR_ADDRESS=172.18.0.1:8080 python main.py

Replace the contents of ORCA_CORE and PROCESSOR_ADDRESS with the output of orca status.

6. Emit a window to orcacore

Check out more examples here.

7. Sync local registry of algorithms to Orca Core

When building algorithms that depend on the result of algorithms in other proccessors it's neccessary to reference these algorithms as python objects in the depends_on argument of the algorithm decorator function.

To do this, sync your python project with Orca Core:

orca sync

Run this command at the root level of your python project. This will generate a .orca folder that contains type stubs for the algorithms stored in other processors.

Then, configure the settings of your language server and/or linter to include these type stubs. E.g. for pyright, you would include these settings in your pyrightconfig.json file at the root of your python project:

{
  "stubPath": ".orca",
}

Now, when importing modules from orca_python.registry.algorithms you will see algorithms configured by remote processors.

For example:

from orca_python.registry.algorithms import <my_remote_algorithm>

This stubbed algorithm can be used as an algorithm in the depends_on argument of the algorithm decorator function.

from orca_python import Processor
from orca_python.registry.window_types import every_30_seconds
from orca_python.registry.algorithms import my_remote_algorithm

proc = Processor("ml")

@proc.algorithm(
    "MyAlgo",
    "1.0.0",
    every_30_seconds,
    depends_on=[my_remote_algorithm]
)
def my_algorithm(params: ExecutionParams) -> StructResult:
    ...

Configuration

Some configuration is required to register a process then start accepting processing requests. If using the internal orca stack for development (via orca start), then simply initialise the repository with an orca.json file:

orca init

You can additionally provide these environment variables to override, or set, certain configurations:

• ORCA_CORE - the address to reach the Orca-core service
• PROCESSOR_ADDRESS - the address needed by Orca-core to reach the processor, of format <address>:<port>
• PROCESSOR_EXTERNAL_PORT - an optional alternative port that should be used by Orca-core to contact the processor. Useful in scenarios like deploying the processor behind a managed service.
• ENV - when set to production the processor will serve using TLS

🧱 Key Concepts

Checkout the Orca docs for info on how Orca works.

👥 Community

GitHub Issues: https://github.com/orc-analytics/orca-python/issues

Discussions: Coming soon!

📄 License

This SDK is part of the Orca ecosystem, but licensed under the MIT License.

See the full license terms (here)[./LICENSE].