fix(agents): return tables from all databases in get_usable_table_names

[chon3806]

Description

In MindsDBQuery.get_usable_table_names, the result_tables list is reset to [] on every iteration of the for db_name in self.tables.databases: loop. As a result, when an agent is configured with tables from more than one data source (e.g. data = {\"tables\": [\"db1.*\", \"db2.*\"]}), only the tables from the last iterated database are ever returned. Tables from earlier databases are silently dropped.

This affects:

• AgentsController.check_agent_data (called via get_usable_table_names(lazy=False) during agent creation), which can spuriously fail validation.
• The agent runtime's view of which tables it has access to whenever a wildcard is used across multiple data sources.

The fix is a one-line change: move the result_tables = [] initialization out of the per-database loop so matching tables are accumulated across all configured databases.

A unit test (tests/unit/interfaces/agents/test_sql_toolkit.py) is added covering:

• Single-database wildcard expansion
• Multi-database wildcard expansion (regression test for this bug)
• Lazy mode returning items as-is
• Empty tables returning an empty list

Type of change

• 🐛 Bug fix (non-breaking change which fixes an issue)

Verification Process

• Test Location: tests/unit/interfaces/agents/test_sql_toolkit.py
• Verification Steps:
  1. pytest tests/unit/interfaces/agents/test_sql_toolkit.py -v
  2. Manual: create an agent with data = {\"tables\": [\"db1.*\", \"db2.*\"]} and confirm tables from both databases are listed.

Checklist

• My code follows the style guidelines (PEP 8) of MindsDB.
• I have appropriately commented on my code, especially in complex areas.
• Necessary documentation updates are either made or tracked in issues.
• Relevant unit tests are added.

[github-actions]

All contributors have signed the CLA ✍️ ✅
_{Posted by the CLA Assistant Lite bot.}

[entelligence-ai-pr-reviews]

EntelligenceAI PR Summary

Removes dead code in the SQL toolkit and adds unit test coverage for MindsDBQuery.get_usable_table_names.

• Deleted unused result_tables = [] variable in mindsdb/interfaces/agents/utils/sql_toolkit.py
• Added tests/unit/interfaces/agents/test_sql_toolkit.py with four test cases: single-database wildcard, multi-database wildcard (regression), lazy mode, and empty input
• Test infrastructure uses _make_query and _make_handler factory functions with unittest.mock.Mock to isolate tests from SessionController and live DB dependencies

---

Confidence Score: 5/5 - Safe to Merge

Safe to merge — this PR removes a dead result_tables = [] variable in sql_toolkit.py that was never read or returned, and adds four focused unit tests for MindsDBQuery.get_usable_table_names covering the single-database wildcard, multi-database wildcard regression, lazy mode, and empty input cases. No logic changes are introduced to production code paths, and the new test file is cleanly isolated using unittest.mock.Mock factory helpers. No review comments were raised and no pre-existing unresolved issues touch the modified code.

Key Findings:

• The only production code change is deletion of the unused result_tables = [] assignment in sql_toolkit.py, which has zero runtime impact since the variable was never referenced after initialization.
• The new tests/unit/interfaces/agents/test_sql_toolkit.py correctly exercises the multi-database wildcard path, which appears to be the regression scenario motivating this PR, providing a safety net against future regressions.
• Test isolation via _make_query and _make_handler Mock factories is appropriate for unit-level coverage and avoids any dependency on live database connections.
• No automated review comments were generated and heuristic analysis found zero medium, significant, or critical issues across both changed files.

Files requiring special attention

• mindsdb/interfaces/agents/utils/sql_toolkit.py
• tests/unit/interfaces/agents/test_sql_toolkit.py

[chon3806]

I have read the CLA Document and I hereby sign the CLA

[torrmal]

thank you!! can you please create a pull request in github.com/mindsdb/engine