Can AI read a database schema from a whiteboard?

Database schema design often starts on a whiteboard. Engineers sketch tables as boxes with column lists, draw lines between them to show relationships, annotate primary keys with 'PK' and foreign keys with 'FK', and write cardinality at the ends of relationship lines (1:1, 1:N, M:N).

This whiteboard ER diagram is a high-value artifact — it captures the first agreed data model. But it's fragile: it exists only as long as someone doesn't erase it, and a photo of a whiteboard schema is not searchable or editable.

What BoardSnap reads from a whiteboard schema:

• Table names
• Column names and basic annotations (PK, FK, NOT NULL)
• Relationship lines and their cardinality labels
• Index or partitioning notes written in margins
• Open questions and tradeoffs noted during the design session

The output: BoardSnap returns a written summary describing the tables, their columns, and their relationships in plain text. This is useful for:

• Seeding a Notion or Confluence data model doc
• Giving the summary to a developer to implement in SQL or a migration script
• Reviewing the schema design with someone who wasn't in the room
• Asking follow-up AI questions: "What normalization issues exist in this schema?" or "What's missing for a multi-tenant setup?"

Limitations: BoardSnap outputs structured text, not SQL DDL or a database-importable format. If you need a DDL file, use the summary as the spec and write the DDL from it. Schemas with very small text or overlapping relationship lines may require a closer snap or a cleaner board layout.

When to use what: Use BoardSnap when the schema was designed collaboratively on a whiteboard and you need a portable record. Use a dedicated schema design tool (dbdiagram.io, Vertabelo, etc.) when the schema is ready for formal documentation.