version: "1.0.0" name: houdini-copernicus description: >- Typed COP/Copernicus tools for image-processing networks: create a COP network, build a bounded filter chain, cook it, verify the output artifact, inspect connections, and validate cook errors before handing the result to a render or compositing workflow. license: MIT compatibility: "meshy-dcc-houdini 0.1+, Houdini 20.5+, meshy-dcc-core 0.2.0+" allowed-tools: Bash Read Write Edit metadata: meshy-dcc: dcc: houdini layer: domain stage: authoring version: "1.0.0" tags: [houdini, cop, copernicus, compositing, image, filter, raster, cook, output, validation] search-hint: "cop copernicus image compositing raster blur colorcorrect file filter network cook output artifact" tools: tools.yaml

houdini-copernicus

Typed Copernicus network authoring and inspection (legacy COP2 is rejected). The skill uses Houdini node types and parameter readback instead of embedding a large Python recipe, so agents can compose the same image-processing graph with the existing node and render skills.

Supported filter aliases

file, blur, colorcorrect, ramp, and null use their Copernicus names. composite maps to blend, and output maps to rop_image. A raw Houdini node type may also be used when it matches the namespaced node-type contract. Unsupported or unavailable node types return a structured error.

Tracer-bullet flow

1. create_cop_network(parent_path="/obj/geo1", network_name="copnet1")
2. create_cop_node(network_path="/obj/geo1/copnet1", filter_type="file", parameters={...})
3. create_cop_node(..., filter_type="blur", input_nodes=["file1"], parameters={...})
4. inspect_cop_network(network_path="/obj/geo1/copnet1")
5. validate_cop_network(network_path="/obj/geo1/copnet1")

The response exposes node paths, actual connection endpoints, parameter readback, and cached cook errors. Missing parameters fail the request; new nodes are removed on failure. Inputs must belong to the same network. Inspection does not cook, so validation reports only cached diagnostics. It does not claim a rendered image until a downstream render skill verifies an output artifact.

Cooking and output verification

Rendering to disk

This package deliberately does not ship its own render tool. A Copernicus output node is a ROP, so houdini_render__render_rop already drives it, with background jobs, cancellation and progress queries:

1. create_cop_node(network_path=..., filter_type="output") → rop_image
2. load_skill("houdini-render") → houdini_render__render_rop(rop_path=...)
3. inspect_cop_output(node_path=...) → verify the artifact landed on disk

Use build_composite_chain when the whole chain is known up front and create_cop_node when iterating step by step. For a validated recipe that spans node categories, use houdini_automation__build_node_chain — this tool stays Copernicus-specific so it can apply the filter aliases and the Cop category check.

Parameter policy

Parameter overrides are applied through parameter_edit: a name the node does not expose fails the call and rolls back, so create_cop_node and build_composite_chain have no skipped_parameters output. A failed chain step destroys every node the request created.

Legacy COP2 networks remain unsupported and are rejected with a structured error; Copernicus (copnet) is the only supported compositing context.