How do I contribute to or modify the snippets it keeps suggesting?

I've been evaluating Codeium for a month now, primarily for its autocomplete in our CI/CD pipeline code (Terraform, Jenkinsfiles, Python scripts). The suggestions are generally decent, but they keep pushing a specific, outdated pattern for AWS secret retrieval in our codebase. It's becoming a nuisance.

I need to either correct this pattern or contribute a better one. The documentation mentions "learning from your code," but it's unclear on active curation. Here's what I'm dealing with and what I've tried:

The problematic suggestion it loops on:
```python
# Codeium keeps suggesting this
secret = boto3.client('secretsmanager').get_secret_value(SecretId='arn:aws:secretsmanager:...')
# Hardcoded ARN, no error handling, no caching
```

What we actually use:
```python
# Our standard pattern
def get_secret(secret_name: str, region: str = 'us-east-1'):
client = boto3.session.Session().client('secretsmanager', region_name=region)
try:
resp = client.get_secret_value(SecretId=secret_name)
except ClientError as e:
logger.error(f"Secret retrieval failed: {e}")
raise
return resp['SecretString']
```

My attempts so far:
* Repeatedly accepting and using our pattern, hoping it learns.
* Using the 'thumbs down' feedback on every bad suggestion.
* Searching for a local config file (`.codeium`, `codeium.json`) to define or override snippets.

Questions for the community:
* Is there a way to directly edit or contribute to the snippet library, especially for organization-specific patterns?
* Does the 'thumbs down' actually train the model for my workspace, or is it just generic feedback?
* Are there config files or workspace settings where I can inject preferred code patterns, similar to VS Code snippets but for the AI?
* Has anyone successfully suppressed an unwanted suggestion permanently?

I'm looking for concrete, actionable answers, not just "it learns over time." We need predictability for security-related code patterns.