juniper-cascor-worker

Remote candidate training worker for the JuniperCascor cascade correlation neural network service.

Overview

This package enables distributed candidate training on remote hardware.

It supports two worker modes:

• WebSocket mode (default): CascorWorkerAgent connects to the juniper-cascor /ws/v1/workers endpoint and exchanges structured JSON and binary tensor frames.
• Legacy mode (--legacy): CandidateTrainingWorker connects to a CandidateTrainingManager over multiprocessing.managers (deprecated).

Ecosystem Compatibility

This package is part of the Juniper ecosystem. Compatible with:

juniper-data	juniper-cascor	juniper-canopy
0.4.x	0.3.x	0.2.x

Installation

pip install juniper-cascor-worker

Note: This package requires the JuniperCascor source code to be importable on the worker machine (the worker runs CasCor's training code locally).

CLI Usage

# Default mode (WebSocket)
juniper-cascor-worker \
    --server-url ws://192.168.1.100:8200/ws/v1/workers \
    --auth-token my-worker-token

# WebSocket mode with mTLS and custom heartbeat
juniper-cascor-worker \
    --server-url wss://cascor.example.com/ws/v1/workers \
    --auth-token my-worker-token \
    --heartbeat-interval 15 \
    --tls-cert /etc/juniper/worker.crt \
    --tls-key /etc/juniper/worker.key \
    --tls-ca /etc/juniper/ca.pem

# Legacy mode (deprecated)
juniper-cascor-worker \
    --legacy \
    --manager-host 192.168.1.100 \
    --manager-port 50000 \
    --authkey my-secret-key \
    --workers 4 \
    --mp-context forkserver \
    --log-level INFO

Python API

import asyncio

from juniper_cascor_worker import CascorWorkerAgent, WorkerConfig

config = WorkerConfig(
    server_url="ws://192.168.1.100:8200/ws/v1/workers",
    auth_token="my-worker-token",
)

agent = CascorWorkerAgent(config)
asyncio.run(agent.run())

Environment Variables

Variable	Mode	Description	Default
CASCOR_SERVER_URL	WebSocket (default)	Worker endpoint URL (ws:// or wss://)	(required)
CASCOR_AUTH_TOKEN	WebSocket (default)	Token sent as X-API-Key header	empty
CASCOR_HEARTBEAT_INTERVAL	WebSocket (default)	Seconds between heartbeat messages	10.0
CASCOR_TLS_CERT	WebSocket (default)	Client certificate path for mTLS	unset
CASCOR_TLS_KEY	WebSocket (default)	Client private key path for mTLS	unset
CASCOR_TLS_CA	WebSocket (default)	Custom CA bundle for TLS verification	unset
CASCOR_MANAGER_HOST	Legacy (--legacy)	Manager hostname	127.0.0.1
CASCOR_MANAGER_PORT	Legacy (--legacy)	Manager port	50000
CASCOR_AUTHKEY	Legacy (--legacy)	Manager authentication key	(required in legacy mode)
CASCOR_NUM_WORKERS	Legacy (--legacy)	Worker process count	1
CASCOR_MP_CONTEXT	Legacy (--legacy)	Multiprocessing method	forkserver

Juniper Ecosystem

This package is part of the Juniper Cascade Correlation Neural Network Research Platform.

Package	Description	Install
juniper-data-client	Dataset service client	pip install juniper-data-client
juniper-cascor-client	Neural network service client	pip install juniper-cascor-client
juniper-cascor-worker	Distributed training worker (this package)	pip install juniper-cascor-worker

License

MIT License - see LICENSE for details.