Glossary, Asset Documentation, and Metrics Library

Company Glossary

What it is: An automatically generated business glossary that maps your organization's business terminology and jargon to the technical data schema.

How it's built: Solid ingests wikis, internal documentation, BI report definitions, and query metadata. It extracts terminology and maps each term to:

• A plain-English definition
• The associated technical schema elements (tables, columns)
• Any synonyms or alternative phrasings used in the organization

Why it matters: Without a glossary, different users use different terms for the same concept (e.g., "revenue" vs. "bookings" vs. "ARR"), and the LLM cannot reliably translate natural-language questions into correct SQL. Solid's glossary is the bridge.

MCP access: The glossary_search MCP tool gives agents direct access to the glossary at runtime.

Asset Documentation

What it is: Per-table and per-column auto-generated descriptions with quality ranking.

What's generated per asset:

• A business description of the table/column
• Usage frequency (how often it's queried)
• BI query patterns (how it appears in BI reports)
• Column statistics
• An asset quality score / ranking

Quality assessment: Solid ranks assets by data quality signals, giving modelers and agents insight into which sources are most reliable.

Metrics Library

What it is: A centralized, reusable store of business metric entities. Extracted once from query history; referenced across any number of semantic models.

How it works:

1. Solid auto-extracts metric definitions from qualified query history
2. Metrics are stored in a central library, not duplicated per model
3. At semantic model creation time, relevant metrics are automatically assigned to the new model
4. The modeler reviews and validates metric assignments

Why "extract once, reuse everywhere" matters: it ensures that "Monthly Recurring Revenue" means exactly the same thing whether it appears in the Marketing semantic model, the Finance semantic model, or the Executive Dashboard semantic model — no metric drift, no conflicting definitions.