A good semantic should describe by itself.
I'd like to discuss whether ai_context could eventually become optional by making our core metadata more self-descriptive.
Current situation
ai_context is a pragmatic field today – it provides synonyms, custom instructions, and extra hints for AI tools. I understand its value, but I believe a well-designed semantic model should carry enough meaning in name and description alone.
Suggestions for stronger naming & descriptions
Naming conventions:
-
Use business terms (e.g., Revenue not TotalSalesAmount).
-
Avoid abbreviations except common ones (e.g., ID).
-
Be consistent: use date suffixes, is prefixes for booleans.
-
Name the business concept, not its implementation.
-
Use singular for models, plural for datasets.
Descriptions should cover:
-
Grain (what each row represents)
-
Source
-
Business definition / formula
-
Differences from similar fields
-
Edge cases or filters
-
Synonyms/aliases (which alone can replace many uses of ai_context)
Conclusion
If we invest in precise, structured descriptions and clear naming, we can significantly reduce the need for ai_context. I'd suggest we treat this as a long‑term goal – perhaps documenting best practices and eventually deprecating ai_context when the metadata is rich enough.
A good semantic should describe by itself.I'd like to discuss whether ai_context could eventually become optional by making our core metadata more self-descriptive.
Current situation
ai_context is a pragmatic field today – it provides synonyms, custom instructions, and extra hints for AI tools. I understand its value, but I believe a well-designed semantic model should carry enough meaning in name and description alone.
Suggestions for stronger naming & descriptions
Naming conventions:
Use business terms (e.g., Revenue not TotalSalesAmount).
Avoid abbreviations except common ones (e.g., ID).
Be consistent: use date suffixes, is prefixes for booleans.
Name the business concept, not its implementation.
Use singular for models, plural for datasets.
Descriptions should cover:
Grain (what each row represents)
Source
Business definition / formula
Differences from similar fields
Edge cases or filters
Synonyms/aliases (which alone can replace many uses of ai_context)
Conclusion
If we invest in precise, structured descriptions and clear naming, we can significantly reduce the need for ai_context. I'd suggest we treat this as a long‑term goal – perhaps documenting best practices and eventually deprecating ai_context when the metadata is rich enough.