How It Works¶
Neksus JobSpec follows a simple pipeline:
Pipeline stages¶
- Parse: YAML files are loaded into dictionaries.
- Validate: Pydantic validates dictionaries into typed
JobSpecmodels. - Inspect: Metadata is extracted for human and JSON inspection outputs.
- Render: Valid models render to markdown, html, or json outputs.
- Check: Project-level checks validate config, directories, file validity, and duplicate IDs.
Request lifecycle example¶
Input file jobspecs/backend-engineer.jobspec.yaml:
Behavior:
- YAML is read and parsed.
- Model validation runs.
- Warning checks run (duplicates, short title, location detail checks).
- In
--strict, warnings can fail the command. - Output is emitted as human text or stable JSON (
--json).
Then render:
neksus-jobspec spec render jobspecs/backend-engineer.jobspec.yaml --format html --theme modern --output dist/backend-engineer.html
Behavior:
- Validated model is passed to renderer.
- Format-specific renderer outputs deterministic content.
- CLI writes output file if
--outputis provided.
Boundaries¶
- CLI layer: thin command orchestration and output formatting.
- Core layer: parsing, validation, checks, and rendering logic.
- Planned MCP/API: future network wrappers around core logic; not implemented now.