Viash component authoring

Viash component authoring

A Viash component = a config.vsh.yaml (metadata: arguments, resources, engines)

• a script (Python/R/Bash). Viash compiles it into a CLI, a Docker image, and a Nextflow module.

Steps

1. Identify the stage and its interface. OpenProblems components conform to a comp_* API merged in via __merge__: /src/api/comp_<stage>.yaml. Find the right interface for the pipeline stage you're implementing (see context/ist-preprocessing-pipeline.md) and match its inputs/outputs exactly.

2. Write config.vsh.yaml. Use the modern flat config (no functionality: wrapper) with engines: / runners: (not the legacy platforms:). Follow OpenProblems conventions: merge the stage API, build on the shared base image with a setup partial, include the standard metadata block, and label the nextflow runner with resource directives. This mirrors a real task_ist_preprocessing component:

   __merge__: /src/api/comp_<stage>.yaml
   
   name: my_method
   label: "My Method"
   summary: "One-line biological purpose."
   description: "Longer description with method context."
   links:
     documentation: "https://github.com/openproblems-bio/task_ist_preprocessing"
     repository: "https://github.com/openproblems-bio/task_ist_preprocessing"
   references:
     doi: "10.xxxx/xxxxx"        # the method's paper
   
   arguments:
     - name: "--some_param"
       type: double
       default: 1.0
       description: "Biologically meaningful parameter."
   # Note: --input/--output are usually inherited from the merged comp_<stage> API.
   
   resources:
     - type: python_script
       path: script.py
   
   engines:
     - type: docker
       image: openproblems/base_python:1        # shared base, not raw python:*
       __merge__:
         - /src/base/setup_spatialdata_partial.yaml   # extra deps via setup partial
     - type: native
   
   runners:
     - type: executable
     - type: nextflow
       directives:
         label: [ midtime, lowcpu, lowmem ]     # resource labels
   

   Confirm the exact __merge__ paths (the stage API and the base setup partial) and the current Viash schema against the live repo before building — keys evolve between major Viash versions.

3. Write the script with the dev parameter block:

   import spatialdata as sd
   
   ## VIASH START
   par = {"input": "resources_test/.../dataset.zarr", "output": "output.zarr", "some_param": 1.0}
   ## VIASH END
   
   sdata = sd.read_zarr(par["input"])
   # ... implement; keep coordinate systems consistent ...
   result.write(par["output"])
   

   Fail loudly: log the error and sys.exit(1) on failure.

4. Validate before building. Run analyze_workflow_configuration on the config.vsh.yaml to catch structural/dependency issues.

5. Build & test with the user's local Viash (the server does not run it):

   • viash ns build
   • viash run config.vsh.yaml -- --input <test>.zarr --output tmp/out.zarr
   • unit test: add a test_resources script (e.g. test.py) to the config and run viash test config.vsh.yaml — every OpenProblems component ships a test.
   • integration: nextflow run main.nf -profile test,docker

Don't

• Don't invent create_spatial_component / build_viash_component tool calls — they're roadmap, not implemented. Generate files and run viash in the terminal.