1. Always add the following note to the code you generate:

# NOTE: This <file/function/class> was partially generated using AI assistance.

Adjust the comment style to the language generated. This way we let people know the file was generated, which helps reviewing it with a different lens (i.e. not relying on a senior colleague having put a lot of thought into it but being a bit more critical / reflecting about it)

Coding style

We use Google style python docstrings
we use pytest for testing
we use pyspark for large scale data preprocessing
our pipeline leverages kedro for structuring our code
we use pandera for data validation
we use fastapi for API building
we use pydantic for data validation and settings management
we use uv for building and packaging
we like functional programming
we love performance and use joblib to cache expensive functions (.cache folder by default)
when writing unit tests, organise the test in a "GivenWhenThen" format, making the purpose of the test explicit
when writing unit tests, do not generate unit tests that hardly test anything of value. Only generate them when the full test is clean, intuitive and valuable. Else just generate a stub describing what to test so the developer completes the test themselves.
comments should never describe changes to the code, but why behind the changes. When editing code, do not add comments explaining a change between two versions but only use comments to explain a broader why of the code as it is now.

LiteLLM provider changes

When adding or editing LiteLLM providers, follow
```
docs/src/infrastructure/LiteLLM-New-Provider-Guide.md
```
so secrets, ExternalSecret mappings, and
```
proxy_config.model_list
```
stay in sync.

1. Always add the following note to the code you generate:

# NOTE: This <file/function/class> was partially generated using AI assistance.

Coding style

We use Google style python docstrings
we use pytest for testing
we use pyspark for large scale data preprocessing
our pipeline leverages kedro for structuring our code
we use pandera for data validation
we use fastapi for API building
we use pydantic for data validation and settings management
we use uv for building and packaging
we like functional programming
we love performance and use joblib to cache expensive functions (.cache folder by default)
when writing unit tests, organise the test in a "GivenWhenThen" format, making the purpose of the test explicit
when writing unit tests, do not generate unit tests that hardly test anything of value. Only generate them when the full test is clean, intuitive and valuable. Else just generate a stub describing what to test so the developer completes the test themselves.
comments should never describe changes to the code, but why behind the changes. When editing code, do not add comments explaining a change between two versions but only use comments to explain a broader why of the code as it is now.

LiteLLM provider changes

When adding or editing LiteLLM providers, follow
```
docs/src/infrastructure/LiteLLM-New-Provider-Guide.md
```
so secrets, ExternalSecret mappings, and
```
proxy_config.model_list
```
stay in sync.

1. Always add the following note to the code you generate:

1. Always add the following note to the code you generate:

Coding style

LiteLLM provider changes

Related Skills

Markdown Converter

Nano Banana Pro

1password

1. Always add the following note to the code you generate:

Coding style

LiteLLM provider changes