feat(python): add import_paths config for custom code integration

[aditya-arolkar-swe]

Description

Refs: Requested by @aditya-arolkar-swe

Adds a generalized import_paths config option to the Python SDK generator that auto-loads user-defined files when the SDK is imported. This follows the same pattern as the Ruby SDK generator (PR #11264), but uses Pythonic naming (import_paths instead of requirePaths).

Link to Devin run: https://app.devin.ai/sessions/9e243bc5375e42018712dffa54f7546f

Changes Made

• Added import_paths config option to SDKCustomConfig (array of strings)
• Added _write_import_paths_hook() function in module_manager.py that writes import hooks at the end of root __init__.py
• Updated Project class to pass import_paths and package_name to ModuleManager
• Updated abstract_generator.py to read import_paths config and pass to Project
• Added get_import_paths() method to CoreUtilities
• Added test fixture exhaustive:import-paths to validate the feature
• Added changelog entry for v4.53.0

Usage

Generator config:

generators:
  - name: fernapi/fern-python-sdk
    version: 4.53.0
    config:
      import_paths:
        - sentry_integration
        - custom_logging

Generated code in root __init__.py:

# Load user-defined files if present (e.g., for Sentry integration)
# Files are loaded from <package>/ if they exist
for _path in ["sentry_integration", "custom_logging"]:
    try:
        import_module(f".{_path}", "<package>")
    except ImportError:
        pass

Example user-defined sentry_integration.py:

# src/<your_package>/sentry_integration.py
# Add this file to .fernignore to preserve across regeneration

import sentry_sdk

def initialize_sentry():
    sentry_sdk.init(
        dsn="https://[email protected]/123456",
        traces_sample_rate=1.0,
    )

# Call at module level - runs once when SDK is first imported
initialize_sentry()

Human Review Checklist

• Verify the generated __init__.py has the import hook at the end - check seed/python-sdk/exhaustive/import-paths/src/seed/__init__.py lines 64-70
• Confirm this approach aligns with Ruby PR feat(ruby): add opt-in requirePaths config for auto-loading user files #11264 pattern (using try/except for graceful handling of missing files)
• Verify the import_module import is available (already imported for lazy imports, conditionally added otherwise at line 236 of module_manager.py)
• Verify package name computation handles package_path config correctly
• Confirm config uses snake_case (import_paths) not kebab-case

Testing

• Seed test exhaustive:import-paths passes
• Pre-commit hooks pass (ruff formatting applied)
• All CI checks pass
• Generated SDK not fully tested (seed ran with --skip-scripts)

Updates Since Last Revision

• Renamed require_paths to import_paths for more Pythonic naming (per reviewer feedback)
• Removed kebab-case support (require-paths) - only snake_case (import_paths) is supported
• Renamed fixture folder from require-paths to import-paths
• Updated changelog to use import_paths terminology

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment and CI monitoring

[tjb9dc]

let's remove all the kebab-case support. We don't do it elsewhere in the python generator so I think it's just confusing

[tjb9dc]

Do you think require_paths is the right term here? It was a bit confusing to me that we import from a require_path (require IMO is not a python term) and then idk how package_name relates

[devin-ai-integration]

Good point! I used require_paths to match the Ruby SDK's requirePaths config option for consistency across generators (as requested in the original task to duplicate Ruby PR #11264).

However, you're right that "require" is more of a Ruby term. A more Pythonic name would be import_paths or auto_import_paths.

Regarding package_name - it's used in the generated import statement: import_module(f".{_path}", "seed") where "seed" is the package name. The .{_path} is a relative import within that package.

Would you prefer I rename this to import_paths for better Python semantics? Happy to make that change if so.

[tjb9dc]

yes, import_paths is PERFECT!!

[devin-ai-integration]

Done! I've renamed require_paths to import_paths throughout the codebase:

• Updated all Python generator files
• Renamed the fixture folder from require-paths to import-paths
• Updated the changelog to use import_paths

The config option is now import_paths (snake_case, Pythonic naming).