[tteok1]

Description

SQLDatabase.insert_into_table("records", ...) raises KeyError when SQLDatabase has a configured schema: SQLAlchemy reflects the table under a key such as analytics.records, but insertion looks up records.

Resolve the schema-qualified metadata key first, while retaining exact-key lookup for callers already passing qualified names. This is a self-discovered bug; no existing issue is being claimed.

New Package?

• Yes
• No

Version Bump?

Not applicable to llama-index-core.

Type of Change

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

How Has This Been Tested?

• Added unit tests covering this change.
• Five SQLite cases cover the default schema, an explicitly selected default schema, an attached schema, and existing qualified-name calls. Each verifies that only the intended same-named table receives the row.
• Before the fix: 2 failed, 3 passed. After the fix: all 5 passed.
• SQL wrapper and structured SQL index tests: 32 passed, including the 5 new cases; one event-loop deprecation warning was emitted.
• Core Mypy: no issues in 479 source files.
• Repository pre-commit checks on both changed files and git diff --check: passed.

Validation used Windows/Python 3.12 and in-memory SQLite; no external database or model API was used.

AI assistance

AI assisted with identifying the bug, implementing the metadata lookup, and writing regression tests. The checks above were run locally.