Fix: create_config TypeError when read_timeout is passed via config

[evanerwee]

Fixes #361.

Problem

create_config() hardcodes read_timeout=600 as a keyword argument AND spreads user config via **config_args. When a user passes read_timeout in their config JSON, Python raises:

TypeError: botocore.config.Config() got multiple values for keyword argument 'read_timeout'

Same issue affects retries and user_agent_appid.

Fix

Use setdefault() for all hardcoded values — the same pattern already used for max_pool_connections on the line immediately above. User-provided values take precedence; defaults apply when absent.

# Before (breaks):
return Config(read_timeout=600, **config_args)  # duplicate kwarg

# After (works):
config_args.setdefault('read_timeout', 600)
return Config(**config_args)                     # no conflict

Impact

• 4-line change, no new logic
• All existing defaults preserved (non-breaking)
• Users can now configure read_timeout, retries, and user_agent_appid
• Critical for production deployments behind API Gateway / ALB / Lambda timeouts

[mykola-pereyma]

Thanks, we will review you PR shortly.

[mykola-pereyma]

Thanks for the fix — the setdefault approach is clean and correct. Asking for regression tests before merge since this is a bug fix and best practice is to have a test that would have caught the original issue.

[mykola-pereyma]

Fix looks correct and follows same pattern already used for max_pool_connections above. Could you add regression tests for the three fields that were breaking? The existing TestCreateConfig class in test_neptune_graph_stores.py already has this pattern for max_pool_connections so should be straightforward. Without regression tests there is nothing preventing someone from refactoring this back to kwargs and reintroducing the TypeError.

[marora90]

Thanks for opening this @evanerwee I’d like to help get this over the line. Since the requested changes have been waiting for a couple weeks, I’m going to open a follow-up PR that cherry-picks your commits and addresses the review feedback. Happy to close mine if you’d prefer to continue this one.