federated-learning-enterprise 0.1.0

Enterprise-grade federated learning package (refactor of fl_framework)

Federated Learning — Enterprise Rewrite

This folder contains an enterprise-style refactor of the fl_framework package. The package is provided as federated_learning and preserves the original framework API while adding enterprise-oriented structure, packaging, and getting-started documentation.

Installation

1. Create a virtual environment and install runtime dependencies:

python -m venv .venv
source .venv/bin/activate  # on Windows use `.venv\Scripts\Activate.ps1`
pip install -r requirements.txt

Quickstart — programmatic (Server)

Minimal example to start a server programmatically:

from federated_learning.config import FLConfig
from federated_learning.fl_server import FLServer

cfg = FLConfig(
	backend_url="http://localhost:8000",
	sqs_queue_url="https://sqs.us-east-1.amazonaws.com/…/server-queue",
	sqs_server_queue_url="https://sqs.us-east-1.amazonaws.com/…/server-queue",
	client_id="fl-server",
	model_id="my_model_v1",
	min_clients_for_aggregation=2,
)

server = FLServer(cfg, client_queue_urls=[
	"https://sqs.us-east-1.amazonaws.com/…/client-1",
	"https://sqs.us-east-1.amazonaws.com/…/client-2",
])
server.start()  # Blocking; use start_async() for background thread

Quickstart — programmatic (Client)

Client example (you must provide a model implementing FLModel):

from federated_learning.config import FLConfig
from federated_learning.fl_client import FLClient
from my_models import MyModel  # your implementation of FLModel

cfg = FLConfig(
	backend_url="http://localhost:8000",
	sqs_queue_url="https://sqs.us-east-1.amazonaws.com/…/client-queue",
	sqs_server_queue_url="https://sqs.us-east-1.amazonaws.com/…/server-queue",
	client_id="edge-node-1",
	model_id="my_model_v1",
)

model = MyModel()
client = FLClient(model=model, config=cfg, train_data_fn=lambda: load_train_df(), eval_data_fn=lambda: load_eval_df())
client.start()

CLI

A minimal CLI is provided for development:

python -m federated_learning.cli --role server --model-id my_model_v1

Testing

Run the basic import test:

pip install -r requirements.txt
pytest -q

Configuration

Most runtime options can be set via environment variables. Important keys:

• FL_BACKEND_URL — Backend HTTP API base URL
• FL_SQS_QUEUE_URL — SQS queue URL for client
• FL_SQS_SERVER_QUEUE_URL — SQS queue URL for server
• FL_USE_MONGODB — when true, use MongoDB storage backend
• FL_MONGO_URI, FL_MONGO_DB — MongoDB connection details

Contributing

This repository is a starting point. Suggested next steps:

• Add structured logging and Prometheus metrics.
• Expand unit tests for aggregator, transport, and sqs_listener.
• Add CI (GitHub Actions) and packaging (wheel) pipelines.

See the package source in the federated_learning package for implementation details.