openproblems-submission

OpenProblems submission

Goal: a Viash component that conforms to a task's API, builds cleanly, passes the test profile, and is reproducible — ready for a PR.

Checklist

1. Repo setup. Clone the target task_* repo and initialize the common/ git submodule (git submodule update --init --recursive). Read the repo's CONTRIBUTING / README for current conventions.

2. Conform to the component API. Your config.vsh.yaml must __merge__ the correct comp_<stage>.yaml interface and match its inputs/outputs and info/metadata fields. Run analyze_workflow_configuration on the config.

3. Data contract. Inputs/outputs are SpatialData (zarr) / AnnData as the stage requires. Validate test data with validate_spatial_data and confirm the elements the stage needs are present (check_spatial_data_compatibility for multi-file). State raw-vs-normalized expectations explicitly.

4. Reproducibility. Build on the shared base images (openproblems/base_python:1 / base_r:1) with __merge__-ed setup partials rather than raw python:*/r-base images. Pin base-image tags and package versions — no unpinned latest. Provide the standard metadata (label, summary, description, links, references/doi) and document every parameter with a default and biological rationale.

5. Build & test locally (server doesn't run these — use the terminal):

   • viash ns build
   • viash run config.vsh.yaml -- --input <test>.zarr --output tmp/out.zarr
   • viash test config.vsh.yaml — a unit test (test_resources) is expected for every component
   • nextflow run main.nf -profile test,docker

6. Dependency sanity. Run analyze_workflow_dependencies across the workflow files to catch missing/conflicting containers or libraries.

7. PR hygiene. Include test data + expected output, document parameter choices, and confirm CI (GitHub Actions) builds and runs the component.

Don't

• Don't claim the server can build/benchmark/validate-submission automatically — build_openproblems_method, run_openproblems_benchmark, and validate_openproblems_submission are roadmap. Use the validation/analysis tools that exist, and run the build/benchmark steps via local CLIs.