Fix Snowflake sync

[ClaireGz]

Problem
nao sync was not creating any .md files for Snowflake databases even when nao debug correctly detected multiple schemas (e.g. 4 schemas).
This was because:

• Wrong API usage: The sync used conn.list_databases(), which returns databases, not schemas. For Snowflake, the connection is already scoped to a database, so we need to list schemas within that database.
• Wrong attribute: The code used db_config.schema instead of db_config.schema_name (the actual field on SnowflakeConfig).

Solution

• Schema discovery: When no schema is configured, we now get schemas by querying INFORMATION_SCHEMA.SCHEMATA (with fallbacks: raw SQL, then parsing from list_tables(), then list_databases()).
• Table listing: We try list_tables(database=schema) first, then fall back to listing all tables and filtering by schema, handling both schema.table and database.schema.table formats.
• Config: Use db_config.schema_name when a single schema is configured.

With these changes, nao sync creates the expected markdown files (columns, preview, description) for all schemas and tables in the Snowflake database.…schema

[claude]

Code review

No issues found. Checked for bugs and CLAUDE.md compliance.

[claude]

Code review

Found 2 CLAUDE.md compliance issues:

1. Excessive comments describing WHAT instead of WHY

Location: cli/nao_core/commands/sync/providers/databases/snowflake.py, lines 34-102

The new code contains numerous inline comments that describe what the code does rather than why, which violates CLAUDE.md's comment minimization rule:

Minimize comments — only comment complex or ambiguous logic with short JSDocs or python docstring; never describe function inputs/outputs

Examples of unnecessary comments:

• Line 37: # Try to access INFORMATION_SCHEMA.SCHEMATA table - the code itself makes this obvious
• Line 42: # Fallback: try using raw SQL if available - the try/except structure makes this clear
• Lines 56-57: # Extract schema names from fully qualified table names... - the code logic is self-explanatory
• Line 62: # Format: database.schema.table or schema.table - describing the format being parsed
• Lines 79-80, 84, 86-87, 92, 96, 100: Similar descriptive comments

Recommendation: Remove these comments and let the code be self-documenting through descriptive variable/function names.

Reference:

nao/CLAUDE.md

Line 32 in 713b9d5

- **Minimize comments** — only comment complex or ambiguous logic with short JSDocs or python docstring; never describe function inputs/outputs

---

2. Single Responsibility Principle violation - Extract helper functions

Location: cli/nao_core/commands/sync/providers/databases/snowflake.py, lines 33-70 and 78-102

The schema discovery logic (lines 33-70) and table listing logic (lines 78-102) should be extracted into separate helper functions. These complex, deeply nested blocks violate CLAUDE.md's principles:

Follow SOLID principles — single responsibility
Write small, focused functions — each does one thing; extract early rather than inline

The schema discovery block has 4 levels of nested try/except handling multiple concerns:

• Accessing INFORMATION_SCHEMA
• Raw SQL fallback
• Parsing table names to extract schemas
• Default fallback behavior

Similarly, the table listing block (lines 78-102) combines table name parsing, schema matching, and filtering logic.

Recommendation: Extract to helper functions such as:

• _discover_schemas(conn, db_config, db_name) -> list[str]
• _list_schema_tables(conn, schema) -> list[str]
• _parse_qualified_table_name(table_name) -> tuple

This would improve readability, testability, and adherence to single responsibility.

Reference:

nao/CLAUDE.md

Lines 28 to 30 in 713b9d5

	- Follow **SOLID principles** — single responsibility, depend on abstractions
	- Place **high-level functions first** in each file, then private/helper functions below
	- Write **small, focused functions** — each does one thing; extract early rather than inline

[Bl3f]

Thank you, but closing this as it will be fixed by #166.