flyrain opened a new issue, #4522:
URL: https://github.com/apache/polaris/issues/4522
### Is your feature request related to a problem? Please describe.
Polaris already governs physical table metadata via the Iceberg/Polaris REST
Catalog API. The semantic layer above those tables is currently scattered
across dbt projects, Cube definitions, BI-tool-specific YAML, and vendor-owned
semantic views. Each consumer (BI dashboard, LLM agent, notebook) re-derives
definitions like "monthly recurring revenue" from incomplete inputs. By hosting
OSI models in Polaris, we can support following use cases:
1. AI agent grounding. An LLM agent answering "what were our top three
metrics by region last quarter?" loads the relevant OSI models, reads
ai_context for descriptions and synonyms, walks dataset.source to find the
underlying tables, and constructs an accurate query. Polaris is the system of
record for "what metrics exist and what do they mean."
2. BI catalog browsing. A BI tool lists all semantic models under a
namespace via the Polaris API, populates its metric grid from the returned
identifiers and ai_context, and lets users drag metrics onto dashboards. The
same definitions feed dashboards in tools that don't share a backend.
3. Multi-engine metric consistency. A data team defines
monthly_recurring_revenue once as an OSI metric in Polaris. Spark, Trino, and
others clients each read the same definition (with their respective SQL dialect
picked from expression.dialects[]) and produce identical numbers. Without this,
each engine's BI layer re-derives the metric independently and the numbers
drift.
### Describe the solution you'd like
_No response_
### Describe alternatives you've considered
_No response_
### Additional context
_No response_
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
