juniper-cascor-worker

Distributed candidate-training worker for the Juniper Cascade-Correlation platform.

juniper-cascor-worker connects outbound to a running juniper-cascor instance, receives candidate-unit training tasks, trains them, and returns the trained candidates for selection into the growing network. Run several workers to parallelise candidate-pool training across hosts.

Part of the Juniper platform. juniper-cascor-worker is the distributed candidate-training worker for juniper-cascor in Juniper — a multi-package ML research platform built around constructive (Cascade-Correlation) and recurrent neural networks. It is managed by cascor over a wire protocol, with no code-import dependency between the two.

Install

pip install juniper-cascor-worker

Run

The default WebSocket mode connects to a juniper-cascor instance:

juniper-cascor-worker \
  --server-url ws://<cascor-host>:8200/ws/v1/workers \
  --auth-token <worker-token>

…or via the canonical environment variables:

export JUNIPER_CASCOR_WORKER_SERVER_URL=ws://<cascor-host>:8200/ws/v1/workers
export JUNIPER_CASCOR_WORKER_AUTH_TOKEN=<worker-token>
juniper-cascor-worker

Legacy --legacy mode (a multiprocessing.managers worker) is deprecated and kept only for transitional deployments.

Configuration

Key environment variables (WebSocket mode); the full surface and the deprecated CASCOR_* legacy aliases (pre-CFG-06) are documented in AGENTS.md:

Docker

docker build -t juniper-cascor-worker:latest .
docker run --rm \
  -e JUNIPER_CASCOR_WORKER_SERVER_URL=ws://<cascor-host>:8200/ws/v1/workers \
  -e JUNIPER_CASCOR_WORKER_AUTH_TOKEN=<worker-token> \
  juniper-cascor-worker:latest

The Dockerfile defaults to WebSocket mode against ws://juniper-cascor:8200/ws/v1/workers. For the full stack, see juniper-deploy.

Status

Live on PyPI. The current version is shown by the badge above; see CHANGELOG.md.

License

MIT — see LICENSE.