Release v0.9.0 · yimindev/dati

Schema-scoped data access — data sources can now be scoped to specific schemas (schema_scope), so table access, metadata and MCP exposure are managed per schema instead of per whole database. Frontend table-access UI reworked accordingly.
MCP metadata tools cover the schema scope — list_tables now merges imported tables with physical tables from authorized schemas (imported metadata wins, deduped by schema+name, capped by dati.mcp.list-tables.max-physical-tables-per-datasource). Tables created through the schema scope are auto-registered into table_info / column_info with best-effort semantic indexing.
Semantic & subject consistency — closed consistency gaps between subjects, terms, semantic index and ACLs (ACL cleanup on delete, correct audit fields, preserved index createdTime across updates and resync).

MCP tool updates are truly partial (omitted fields are no longer reset).
Correct preview SQL rendering and runtime parameter validation; documented the separation between SQL binding and preview rendering.
Datasource metadata and connection validation hardened.
Auth hardened: password policy 8–72 chars, case-insensitive usernames (canonical casing returned by /auth/me), explicit AUTH002 / AUTH009 error bodies with WWW-Authenticate for unauthenticated requests, and JWT_SECRET is now required outside the dev profile.
Search aliases are allowed again; term relation columns are validated, ACLs cleaned on delete, guest API keys rejected.
Actionable tool parameter validation messages.

MANAGED_SCHEMAS renamed to schema_scope.
Legacy term-relation delete route (with the '_' sentinel) removed; delete by relation id only.

docker pull ghcr.io/yimindev/dati:0.9.0

Full Changelog: v0.8.1...v0.9.0

Navigation Menu