Skip to content

Status: Needs Review

This page has not been reviewed for accuracy and completeness. Content may be outdated or contain errors.

Pipeline & Graph API

Pipelines in this branch are defined by checked-in YAML configs, built through the core pipeline builder, and restored through the shared restore utilities.

Current Pipeline Config Shape

Use the current pipeline schema keys only:

metadata:
  name: MyPipeline
  description: Current pipeline example
  author: cuvis.ai

nodes:
  - name: source
    class_name: cuvis_ai.node.data.CU3SDataNode
    hparams:
      processing_mode: Raw

connections:
  - source: source.outputs.cube
    target: some_node.inputs.data

Key points:

  • class_name identifies the importable node class.
  • hparams carries node constructor arguments.
  • source / target define port-to-port edges.

Shipped Pipeline Families

Current checked-in pipeline configs live under configs/pipeline/ and are grouped as:

  • RX: configs/pipeline/anomaly/rx/
  • Deep SVDD: configs/pipeline/anomaly/deep_svdd/
  • AdaCLIP: configs/pipeline/anomaly/adaclip/
  • SAM3: configs/pipeline/sam3/

Restoration And Remote Execution

These interfaces are part of the current pipeline surface:

  • restore-pipeline
  • restore-trainrun
  • cuvis_ai.utils.grpc_workflow

See: