Always provide comprehensive API documentation that clearly specifies: 1. **Parameter types** - Document all acceptable parameter types, not just specific implementations. When a parameter accepts any object that satisfies certain behavior (like iterables), explicitly state this rather than referencing a specific class. 2. **Function aliases** - When implementing function aliases (like `acos` for `arccos` or `pow` for `power`), clearly document their relationship to the original function and explain why the alias exists. 3. **Standards compliance** - Reference any industry standards or specifications that the API follows, especially when they might differ from what users expect. Example: ```python def process_data(data_source, options=None): """Process data from the provided source. Parameters ---------- data_source : iterable object Any iterable that yields data batches. This can be a DataLoader instance, a list of arrays, or any custom iterable implementing the required interface. options : dict, optional Processing options. Notes ----- This function follows the data processing standards specified in https://example.org/standards/data-processing. """ ``` Or when documenting aliases: ```python acos = arccos acos.__doc__ = """ Trigonometric inverse cosine, element-wise. Notes ---------- `acos` is an alias for `arccos`. It is a standard API in https://data-apis.org/array-api/latest/ instead of an official NumPy operator. >>> np.acos is np.arccos True """ ```