This proposal introduces a set of optional, non-executional metadata extensions to OSI semantic models to improve how data is interpreted, presented, and consumed by downstream tools and agents. While OSI effectively standardizes structural and logical semantics, there is currently limited support for conveying interpretability context such as units of measurement, display conventions, default aggregation behavior, KPI polarity, sorting preferences, and alignment to external semantic concepts. As a result, these details are often redefined or inferred inconsistently across developers, BI tools, and AI systems.
The proposed metadata fields (e.g., measurement, display_format, semantic_type, default_aggregation, desired_direction, default_sort, and semantic_mappings) are designed to be fully optional, backward compatible, and non-impacting to execution semantics. They provide a lightweight, extensible layer that enables more consistent rendering, more accurate automated reasoning, and reduced duplication across consuming systems, while aligning with ongoing OSI discussions around semantic typing (#42), metrics and aggregation (#43), and metadata extensibility (#55).
Detailed proposal here: https://docs.google.com/document/d/1DLuteF0WwlVbpD5kmHhcqa6vItqNFL4MoAQrImy6dKo/edit?tab=t.0
This proposal introduces a set of optional, non-executional metadata extensions to OSI semantic models to improve how data is interpreted, presented, and consumed by downstream tools and agents. While OSI effectively standardizes structural and logical semantics, there is currently limited support for conveying interpretability context such as units of measurement, display conventions, default aggregation behavior, KPI polarity, sorting preferences, and alignment to external semantic concepts. As a result, these details are often redefined or inferred inconsistently across developers, BI tools, and AI systems.
The proposed metadata fields (e.g., measurement, display_format, semantic_type, default_aggregation, desired_direction, default_sort, and semantic_mappings) are designed to be fully optional, backward compatible, and non-impacting to execution semantics. They provide a lightweight, extensible layer that enables more consistent rendering, more accurate automated reasoning, and reduced duplication across consuming systems, while aligning with ongoing OSI discussions around semantic typing (#42), metrics and aggregation (#43), and metadata extensibility (#55).
Detailed proposal here: https://docs.google.com/document/d/1DLuteF0WwlVbpD5kmHhcqa6vItqNFL4MoAQrImy6dKo/edit?tab=t.0