Replace custom term_validator with linkml-term-validator

[cmungall]

Problem

The validation pipeline references a communitymech.validators.term_validator module that does not exist (src/communitymech/validators/term_validator.py is missing), while __init__.py tries to import it. The just validate-terms target fails immediately.

Meanwhile, the plan docs and README already mention linkml-term-validator as the intended tool — we should use it directly rather than writing a custom Python wrapper.

What needs to happen

1. Add linkml-term-validator as a dependency

# pyproject.toml
dependencies = [
    ...
    "linkml-term-validator>=0.1.0",
]

2. Add schema-level bindings or dynamic enums

linkml-term-validator validate-data validates against dynamic enums and bindings — LinkML features that constrain which ontology terms are valid in a given slot. The current schema uses a generic Term class with plain string id/label fields, so the validator has nothing to check against.

Options (not mutually exclusive):

Option A: Bindings on descriptor slots — bind each descriptor's term.id to a prefix-constrained enum:

classes:
  TaxonDescriptor:
    attributes:
      term:
        range: Term
        bindings:
          - binds_value_of: id
            range: NCBITaxonEnum

Option B: Dynamic enums with reachable_from — if we want to constrain to specific branches:

enums:
  NCBITaxonEnum:
    reachable_from:
      source_ontology: obo:ncbitaxon
      source_nodes:
        - NCBITaxon:2  # Bacteria
      relationship_types:
        - rdfs:subClassOf

Option C: At minimum, id_prefixes — add id_prefixes to Term subclasses to at least validate the CURIE prefix is correct (e.g. NCBITaxon for taxa, CHEBI for metabolites). This is simpler but only checks prefix, not term existence/labels.

3. Update justfile targets

Replace the broken custom module calls with direct linkml-term-validator CLI:

# Validate ontology terms in data files
validate-terms FILE:
    uv run linkml-term-validator validate-data {{FILE}} -s src/communitymech/schema/communitymech.yaml --labels

validate-terms-all:
    #!/usr/bin/env bash
    for file in kb/communities/*.yaml; do
        echo "\nValidating terms in $file..."
        uv run linkml-term-validator validate-data "$file" -s src/communitymech/schema/communitymech.yaml --labels
    done

# Validate schema-level enum meanings
validate-schema-terms:
    uv run linkml-term-validator validate-schema src/communitymech/schema/communitymech.yaml

4. Fix the broken __init__.py import

Remove the TermValidator import from src/communitymech/validators/__init__.py since we're using the external tool instead of a custom module.

5. Add to QC pipeline

Update the qc target to include term validation:

qc: validate-all validate-terms-all lint test

Context

We just found 5 completely wrong NCBITaxon IDs and 3 label mismatches in EcoFAB_Ring_Trial_SynCom17.yaml through manual OLS verification. Examples:

• NCBITaxon:69459 was "Dicraspidia" (a plant), not Lysobacter
• NCBITaxon:164543 was "Opisthoteuthis massyae" (an octopus), not Marmoricola

Automated term validation would have caught all of these.

References

• linkml-term-validator docs
• LinkML semantic enumerations
• LinkML bindings
• Using ontologies as values