Speak in business terms
Express fields, enums, and calculations as entities, dimensions, measures, and business definitions.
Foggy uses TM to turn physical structure into domain semantics, QM to compose multiple TMs into a focused business model, and MCP plus Query DSL to deliver that model to LLMs.
What should be delivered to an LLM is not a complete database schema of columns and foreign keys. It is an organized set of business concepts, metric definitions, analytical relationships, and visibility boundaries. Foggy turns that translation into a maintainable modeling workflow.
Express fields, enums, and calculations as entities, dimensions, measures, and business definitions.
Compose multiple TMs through QM so AI reasons over one domain instead of assembling raw joins.
Make models discoverable, describable, validatable, and callable as a stable query boundary.
TM establishes domain semantic units. QM composes those units into a business model for real questions. The LLM primarily sees QM, not the raw database tables.
Map physical tables, fields, and basic relationships into entities, dimensions, measures, and business definitions.
Read the model reference ↗Compose multiple TMs, encapsulate semantic joins, and define the dimensions and measures available for analysis.
Read the model reference ↗Let AI discover accessible QMs, read model descriptions, and submit structured queries.
Read the model reference ↗Foggy does not ask the LLM to remember how orders, order items, customers, and products connect. It only needs to understand the business concepts exposed by the Sales Analysis QM.
t_order.customer_idt_order_item.amountt_order.created_att_product.categoryThe data source provides facts, but should not become the LLM’s cognitive interface.
Translate physical structure into entities, dimensions, measures, and definitions.
Compose across TMs, hide physical joins, and narrow the analysis surface.
Let the LLM discover models, read descriptions, and submit structured queries with evidence.
Handle permissions, dialects, query execution, and traceability.
Inventory tables, fields, enums, and raw relationships to define the business domain.
↗02Turn source objects into explainable entities, dimensions, measures, and foundational definitions.
↗03Connect TMs semantically and shape a unified model for sales, customer, or operational analysis.
↗04Let AI discover and read the QM, then submit a structured query through MCP and Query DSL.
↗05Runtime applies governance, executes the query, and returns explainable evidence.
↗The v1.0 and v2.0 public contracts remain stable. v3.0 is still Draft / Not Frozen. Implementation manuals and version materials will be published only within the confirmed public scope.