improve: add AGENTS.md and harden flatten list typing

AGENTS.md

@@ -0,0 +1,13 @@
+# json2sql — Agent Guide
+
+## Repo overview
+This repo converts JSON to SQL. Primary implementation is under `src/json2sql/`. CLI entrypoints live in `cli.py` and `__main__.py`.
+
+## Commands
+- Install/dev: `pip install -e .`
+- Tests: `pytest`
+- Lint/type checks (if configured): use tooling in `pyproject.toml`
+
+## Do not break
+- Do not remove or weaken existing tests.
+- Keep public CLI behavior stable unless an issue explicitly requests changing it.

src/json2sql/cli.py

@@ -1,9 +1,9 @@
 """CLI interface for json2sql using Typer."""
 
 import sys
+from pathlib import Path
 
 import typer
-from pathlib import Path
 
 # Lazy imports — converter/dialects pulled on command execution
 # to reduce cold start from ~340ms to ~160ms.
@@ -64,8 +64,6 @@ def convert(
     ),
 ):
     """Convert a JSON file to SQL INSERT statements."""
-    from .converter import JSONToSQLConverter
-    from .dialects import Dialect
 
     # Validate dialect
     try:

src/json2sql/converter.py

@@ -84,7 +84,7 @@ def _convert_objects(self, objects: list[dict], table_name: str) -> str:
             # Process nested arrays into child tables
             for obj in objects:
                 for key, value in obj.items():
-                    if isinstance(value, list) and value and isinstance(value[0], dict):
+                    if isinstance(value, list) and value and all(isinstance(v, dict) for v in value):
                         self._flatten_nested(table_name, key, value, obj)
         else:
             columns = self._infer_columns(objects)
@@ -163,7 +163,7 @@ def _infer_columns_flattened(
                             inferred = sql_type_for(sub_value, self.dialect)
                             if columns[flat_key] == "TEXT" and inferred != "TEXT":
                                 columns[flat_key] = inferred
-                elif isinstance(value, list) and value and isinstance(value[0], dict) and self.flatten:
+                elif isinstance(value, list) and value and self.flatten and all(isinstance(v, dict) for v in value):
                     # Skip - goes to separate table
                     pass
                 else:
@@ -216,5 +216,5 @@ def _process_flatten(self, objects: list, table_name: str) -> None:
             return
         for obj in objects:
             for key, value in obj.items():
-                if isinstance(value, list) and value and isinstance(value[0], dict):
+                if isinstance(value, list) and value and all(isinstance(v, dict) for v in value):
                     self._flatten_nested(table_name, key, value, obj)