Problem
Currently the renderer has two outcomes: validation passes and HTML is generated, or validation fails with errors. There is no way to inspect the computed layout without generating output.
For agents iterating on a diagram, being able to "preview" the layout would dramatically reduce trial-and-error.
Suggested Improvement
Add a --dry-run or --validate flag:
node renderers/architecture/render-architecture.mjs input.json --dry-run
Output should include:
- All computed component coordinates (id, x, y, width, height)
- All computed label positions
- All boundary rects
- All connection paths
- Validation results (pass/fail)
This lets the agent inspect the layout numerically before committing to a render — catching issues like "component X is too far right" without a full render cycle.
Acceptance Criteria
--dry-run exits 0 on validation pass, non-zero on failure.- Output is machine-readable (JSON or structured text).
- No HTML file is generated.
- Works for all 5 diagram types.
Problem
Currently the renderer has two outcomes: validation passes and HTML is generated, or validation fails with errors. There is no way to inspect the computed layout without generating output.
For agents iterating on a diagram, being able to "preview" the layout would dramatically reduce trial-and-error.
Suggested Improvement
Add a
--dry-runor--validateflag:Output should include:
This lets the agent inspect the layout numerically before committing to a render — catching issues like "component X is too far right" without a full render cycle.
Acceptance Criteria
--dry-runexits 0 on validation pass, non-zero on failure.