app: implement verification module

[gekiclaws]

Responsibilities:

• Automatically evaluate whether the generated visualization aligns with:
• User context or question
• Expected chart structure
• Statistical correctness (e.g., axes, proportions)

Approach:

• Use CoVe or a similar module to analyze the generated visualization
• Compare output against known reference templates or accepted visualization patterns
• Apply rule-based or ML-based quality checks over:
  • The chart image itself
  • Metadata (e.g., axis ranges, data mappings, labels)
  • Optional text outputs (e.g., the prompt, code, or stats summary)

Purpose:

• Provide a quality assurance layer to catch misaligned or misleading visualizations
• Surface issues (e.g., wrong chart type, mislabeled axis, off-scale values)
• Enable optional downstream actions:
  • Flag results for human review
  • Feed into an automated re-prompting/refinement loop (future extension)
  • Offer user-facing warnings or correction suggestions