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.
Before using this SDK, you should install the Orca CLI and start Orca Core.
- Install the Orca CLI Ensure that Docker is installed on your system.
Linux / macOS
curl -fsSL https://raw.githubusercontent.com/Predixus/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.
- Start the Orca Server
orca start
- Print out the server details
orca status
- Install the Orca sdk into your python project:
pip install orca-time
- 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:
ORCASERVER=grpc://localhost:32770 HOST=172.18.0.1 python main.py
Replace the contents of ORCASERVER
and HOST
with the output of orca status
.
- Emit a window to orcacore TBD
Check out more examples here.
Processor: A container for algorithms, exposing them to the Orca Core service.
Algorithm: A Python function decorated and registered for DAG execution.
Window: A temporal trigger used to activate algorithms.
Algorithm and Window names must be in PascalCase.
Versions must follow semantic versioning (e.g., 1.0.0).
Dependencies must be declared only after their algorithm is registered.
Algorithms cannot depend on others from a different window type (enforced by Orca Core).
GitHub Issues: https://github.com/predixus/orca-python/issues
Discussions: Coming soon!
This SDK is part of the Orca ecosystem, but licensed under the MIT License.
See the full license terms (here)[./LICENSE].
Built with ❤️ by Predixus