Replace all JSON file storage (connections, secrets, history, AI config,
conversations, sidebar layout) with a single SQLite database (dbx.db).
Auto-migrates existing JSON data on first launch, renaming old files to
.bak. Remove keyring dependency — passwords now stored in SQLite.
Refactor all database drivers, query execution, schema operations, and
business logic into a shared `dbx-core` crate. The Tauri desktop app
becomes a thin shell delegating to dbx-core, enabling future web and
CLI frontends to reuse the same core.
PostgreSQL list_indexes was returning all_cols (key + included) in IndexInfo.columns,
causing DDL and UI duplication. Now only key columns are returned in columns, with
included columns correctly split via indnkeyatts.
SQL Server varbinary now handles Some(-1) to display varbinary(max).
- Filter empty strings from STRING_AGG/LISTAGG splits in SQL Server, Oracle, MySQL
- Emit INCLUDE clause only for PostgreSQL and SQL Server (not MySQL/SQLite)
- Emit WHERE clause only for PostgreSQL, SQL Server, SQLite (not MySQL)
- Replace index type free-text input with dropdown for all 5 database types
- Fix hardcoded column count in resize handle (use indexColLabels.length - 1)
- Fix pg_get_indexdef type mismatch: cast k.n from bigint to integer
The numeric_scale extraction used (& 2047) - 1024 which produces negative
values; fixed to & 65535. Added 12 tests for index SQL generation: INCLUDE,
USING, WHERE, COMMENT ON INDEX, type prefix, and combined scenarios.
- Add filter, index_type, included_columns, and comment fields to IndexInfo
- PostgreSQL: query pg_am for index type, use indnkeyatts to split key/included columns, retrieve index comments via obj_description()
- SQL Server: add type_desc, filter_definition, included columns via is_included_column
- MySQL: add INDEX_TYPE from information_schema.STATISTICS
- Oracle: add INDEX_TYPE from ALL_INDEXES
- DDL generation: add USING (PostgreSQL), type prefix + INCLUDE + WHERE (SQL Server), COMMENT ON INDEX (PostgreSQL)
- Frontend: update types, state mapping, SQL generation, and i18n translations
- PostgreSQL: rewrite column query from information_schema to pg_attribute for accurate type info via format_type(atttypid, atttypmod); extract character_maximum_length from atttypmod for varchar/bpchar; fix execute_query with persistent(false) and empty result fallback
- SQL Server: reconstruct column types with precision (varchar(max), decimal(p,s), datetime2(p)); stop dividing nvarchar/nchar CHARACTER_MAXIMUM_LENGTH by 2
- MySQL: use COLUMN_TYPE instead of DATA_TYPE for full type definitions; add CHARACTER_MAXIMUM_LENGTH; fix ONLY_FULL_GROUP_BY with MIN(NON_UNIQUE)
- Oracle: add CHAR_LENGTH/DATA_LENGTH for column type precision; fix INDEX_OWNER join to prevent cartesian product
- SQLite: fix PRAGMA injection via double-quote escaping; detect primary key indexes via origin column
- Auto-focus DBX window when MCP opens a table or executes a query
- Ensure connections are loaded before handling MCP events
- Add /execute-query bridge route and dbx_execute_and_show tool
- Format all tool responses as markdown tables (less tokens, more readable)
- Extract shared bridgeRequest helper to reduce duplication
- Add ssh_key_passphrase field to ConnectionConfig model with full keyring lifecycle
- Pass passphrase to russh load_secret_key for encrypted SSH keys
- Add native file picker (FolderOpen icon) next to SSH key path input
- Add password-type input for key passphrase in connection dialog
- Make connection dialog form scrollable for short windows
- Fix alignment of key path input and browse button
- Add i18n keys (en + zh-CN)
During development, every Rust recompile changes the binary signature,
causing macOS to invalidate Keychain "Always Allow" and prompt for the
password on every hot reload. Switch to a JSON file store in debug
builds so dev workflow is uninterrupted; release builds still use the
system keychain.
- Add rust_decimal feature to sqlx and tiberius for proper Decimal type handling
- Fix try_get_unchecked<String> causing NULL or garbled output for numeric types
- Add numeric_precision and numeric_scale to ColumnInfo for all database drivers
- Update frontend to display precision (e.g., numeric(20,6)) in column headers
- Fix typeColorClass to handle type names with precision suffix
Root cause: PostgreSQL binary protocol sends NUMERIC as PgNumeric binary format,
not UTF-8 text. try_get_unchecked<String> attempted to decode binary bytes as UTF-8,
causing decode failures (NULL) or garbage characters (乱码).
t8y2
Agent-43
yavon007
SuLea-IT
skyler
SuLe
kingcanfish