- Add extract_keywords() to LLMService with conversation context support,
using process_stream + start_log/end_log for logging consistency
- Add expand_with_terminology() to expand keywords with synonym matching
- Add fusion scoring in calc_table_embedding(): α * vec_score + (1-α) * keyword_score
- Add calc_keyword_score() with multi-level matching:
table name exact (1.0), comment exact (0.9), comment substring (0.7),
table name partial (0.25-0.5), field comment exact (0.5),
field comment substring (0.4), field name partial (0.3)
- Scoring considers keyword coverage rate to differentiate tables
with different numbers of matched keywords
- Extracted keywords persisted to ChatRecord for reuse across endpoints
- get_terminology_template() supports comma-separated keyword input
with per-keyword matching for higher precision
- Add TABLE_EMBEDDING_ALPHA and TABLE_EMBEDDING_KEYWORD_ENABLED config
All user-controlled values interpolated into SQL WHERE clause fragments
in transTreeItem() and getSysVariableValue() are now passed through
_escape_sql_value() which doubles single quotes (standard SQL escaping)
and escapes backslashes. The logic operator in transTreeToWhere() is
also validated against a whitelist.
Affected injection vectors:
- item["enum_value"] list (enum filter)
- item["value"] string (normal filter)
- user variable values (variable filter)
- current_user name/account/email (system variable filter)
- tree["logic"] operator (tree structure)
- Add SQLite support to data sources (file-based connection)
- Add Hive support to data sources
- Add SSL toggle option for MySQL and Doris databases
- Add sample data (3 rows JSON format) to help AI better understand table schema
- Fix SQLite schema table name issue (empty schema prefix)
- Add SQLite.yaml template for SQL generation
- Add Hive to haveSchema list in frontend
Co-Authored-By: Claude Opus 4.6 <[email protected]>
junjun