-
Notifications
You must be signed in to change notification settings - Fork 1.4k
Description
Summary
I'm running Wren OSS (Docker-based deployment). I have a large Persian (Farsi) schema with ~113 tables. Column names, table names and many enum values are in Persian or use non-user-friendly technical names. Wren works functionally, but it cannot reliably map Persian natural-language queries to the correct tables/columns because the semantic metadata (display_name, description, synonyms) is missing or hard to edit at scale.
In Wren Cloud there are GUI import/export and model-editing flows that make adding display_name and description easy. In OSS (self-hosted, Docker) I could not find an official export/import or API to bulk update model metadata. There is no documented, supported CLI/API for programmatic bulk updates to model metadata (display_name / description / synonyms / enum labels) that the query engine and LLM can use.
Why this matters
For non-English schemas (Persian, Arabic, Chinese, etc.) or databases with technical column names, adding localized display_name, description and synonyms is essential for accurate natural-language-to-SQL behavior.
A bulk import/update mechanism would allow teams to generate metadata from a spreadsheet/translation pipeline (or from an automated script that uses an LLM) and apply it reliably to an OSS deployment.